DEV Community: Brian Pawl

What "AI-ready data" actually means (in plain English)

Brian Pawl — Fri, 21 Aug 2026 16:37:34 +0000

Originally published on the NuWay Biz Solutions blog.

✦ Cover image: Made with ChatGPT (OpenAI) — we're transparent about AI. See the exact prompt on the original post.

"Get your data ready for AI."

You've heard it from every software demo, every LinkedIn thought-leader, every consultant angling for a discovery call. And not one of them tells you what ready means.

So the phrase just sits there, sounding like homework you'll never finish. Some giant, expensive project you're supposed to complete before the AI you already pay for will do anything useful.

It's not that. "AI-ready data" is four specific things, and you can check for all four in an afternoon. Even better: you don't need them everywhere. You need them for the data behind the one question you actually want answered.

If you've watched a tool stall the moment you ask it something that matters, this is the thing sitting underneath that. Here's each of the four, in plain English: what it means, what it looks like when it's missing, and a ten-second way to tell where you stand.

Connected: your tools feed one place, on their own

Connected means your systems hand information to each other without a person in the middle.

You'll know it's missing the second you catch someone retyping. An order lands in the webstore, and a human keys it into the accounting software by hand. A lead fills out the form, and someone copies the details into the CRM. A quote gets approved in email, and the numbers get re-entered into the job sheet.

Every one of those hops is a place a number gets dropped, fat-fingered, or forgotten. And an AI can't sit in that gap and catch it. It can only work with data that actually arrives somewhere it can read.

Ten-second test: count the times this week someone on your team moved the same information from one screen into another. Retyping is just the obvious version. It also counts when they export a spreadsheet out of one tool to upload into the next, re-key an invoice that already exists somewhere, or save a report so another system can load it back in. All of it is copying by hand, whatever it's called. That number is your disconnection tax, and you've been paying it in salary.

Clean: one customer, one record, one spelling

Clean means each thing in your business exists once, spelled one way.

It breaks in the least dramatic way imaginable. "Bob's Plumbing," "Bob Plumbing LLC," and "bobs plumbing" quietly become three different customers. The same invoice shows up twice. A phone number lives in one system and nowhere else. None of it ever feels urgent enough to fix.

Then you ask the AI how much Bob spent with you last year, and it answers with total confidence, using one of the three Bobs. The number is a third of the real one. It has no way of knowing the other two exist.

Ten-second test: search your customer list for one of your best customers. If more than one version of them comes back, your data isn't clean yet.

✦ Made with ChatGPT (OpenAI) — we're transparent about AI. See the exact prompt on the original post.

Trustworthy: one number everyone agrees on

Trustworthy means that when two systems disagree, you already know which one is right, because there's a single source of truth instead of four competing ones.

The tell is the meeting. The sales total in the CRM doesn't match the revenue on the P&L, and the room burns twenty minutes arguing about which report to believe instead of what to actually do about the number.

Point an AI at that and it doesn't referee the argument for you. It picks one of the numbers (you won't know which) and delivers it in a clean paragraph that sounds completely settled.

Ten-second test: ask two of your systems the same question. Last month's revenue, this quarter's new customers, whatever. If they don't give you the same answer, you don't have a source of truth. You have a debate.

Reachable: the AI can actually open it

Reachable is the one everyone forgets. Your data can be connected, clean, and trustworthy and still be useless, because it's sitting somewhere the AI simply can't go.

A scanned PDF that's really just a photo of a page. A report only one person's login can pull. The spreadsheet on the laptop that goes home every night. The pricing logic that lives entirely in your operations manager's head.

If the AI can't get to it, it may as well not exist.

This is the quiet reason behind half the "why doesn't it just know that?" moments. It would know. It can't reach it.

Ten-second test: pick the one document you'd need to answer your most important question. Could a tool open it on its own, without a human digging it out first? If not, it isn't reachable.

"Ready" doesn't mean "perfect"

Here's the part the scary version of this advice always leaves out.

AI-ready data isn't perfect data. It's data you'd stake a client meeting on.

You are never going to have all four, everywhere, forever. Nobody does. The goal was never a spotless company-wide database. It's four properties that hold true for the specific slice of data behind the specific question you're asking right now.

That's why this is smaller than it sounds. "Ready enough to answer one question" is usually a weekend of plumbing, not a two-year infrastructure project.

Gartner: through 2026, 60% of AI projects unsupported by AI-ready data will be abandoned

Gartner, the technology-research firm a lot of corporate IT budgets get built on, expects organizations to abandon 60% of their AI projects through 2026 — not because the AI was bad, but because it was never handed data it could stand on. The four properties are how you land in the other 40%.

Where to start Monday

Pick the one question you keep wishing your AI could just answer. Which customers are about to leave. What actually made money last quarter. Who never got followed up with.

Now run the four tests on only the data behind that one question. Connected. Clean. Trustworthy. Reachable. Nine times out of ten, one of the four is the whole problem, and fixing that one is the entire job.

Start there. Prove it on a single answer you can actually trust. Then pick the next question and do it again.

If you'd rather not guess which of the four is your gap, start a no-pressure conversation and we'll map it with you: the one question worth starting on, and the one property worth fixing first, whether or not you ever hire us.

Practical AI. Clear process. Real business value.

— Brian, NuWay Biz Solutions

P.S. A test you can run in the next five minutes: ask whatever AI you already pay for, "which customer paid us the most last year?" If you can't trust the answer it gives back, you just found which of the four you're missing — and where to start.

Why doesn't ChatGPT mention my business?

Brian Pawl — Thu, 20 Aug 2026 17:16:40 +0000

Originally published on the NuWay Biz Solutions blog.

✦ Cover image: Made with ChatGPT Images 2.0 — we're transparent about AI. See the exact prompt on the original post.

Somebody sat down and typed this into Google, in a full sentence, the way you'd ask an actual person:

our website hasn't been updated in a while — could that be the reason ai tools don't mention us?

— A real search query, August 2026

You can hear it, can't you. That's not a keyword. That's somebody who's had a suspicion gnawing at them for a while and finally said it out loud to a search box at some point in the evening.

So let's answer it properly, because a lot of people have this backwards.

Short version: yes, a neglected website is a factor. It's also the smallest thing on the list, and going after it first is how businesses spend four grand on a redesign and stay exactly as invisible as they were.

The picture most owners have in their head is a bot that swings past the website, has a look around, decides the place seems abandoned, and moves on. A health inspector with a clipboard.

Nothing like that is happening.

What's actually going on is stranger, and once you can see it the whole thing stops being mysterious.

What the assistant actually reads

Say you ask ChatGPT to recommend a bookkeeper in your town. It doesn't sit there and think about bookkeepers. It runs a search, reads what comes back, and writes you a summary of that.

Google's AI Overviews do the same against Google's index. Google says as much in its own guide for site owners: the generative features are "rooted in our core Search ranking and quality systems," working from "publicly accessible, crawlable content."

Two things fall out of that. The second one surprises people.

First, if a machine can't reach your pages, nothing else on this list matters. Worth actually checking rather than assuming, because a lot of small-business sites quietly block crawlers through a security plugin or some firewall setting nobody remembers switching on.

Second, and this is the one: most of what the AI reads about you, you didn't write.

What owners picture

The AI visits your website
It reads your homepage and your services page
It judges whether the business looks current
A recent redesign moves the needle

What usually happens

It runs a search and reads the results
Most of those results are pages about you, not yours
It looks for several sources agreeing
Your homepage is one input among many, if it shows up at all

Pew Research Center actually went and measured this. They tracked the browsing of 900 US adults through March 2025, across 68,879 Google searches, and found the three most-cited sources in AI summaries were Wikipedia, YouTube and Reddit — together about 15% of everything those summaries pointed at.

Wikipedia. YouTube. Reddit. Your homepage isn't anywhere on that list.

One thing genuinely worth doing today, though. ChatGPT search runs its own crawler, and OpenAI is blunt about it in its documentation: "OAI-SearchBot is used to surface websites in search results in ChatGPT's search features," and "sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers."

Which means there's a plain text file sitting at the root of your website quietly deciding whether you exist inside the most-used assistant on earth. Go and look at it.

It has to be sure you are one business

Alright. Say the assistant does find you. Four pages mention your business: a directory listing, an old chamber-of-commerce page, a review site, and your own homepage.

Something now has to work out that all four are the same company.

Sounds like nothing. Go and look.

Search your own business the way a stranger would, and what you'll usually turn up is the name spelled three different ways, two phone numbers (one of which still rings a mobile you replaced in 2021), an address from the office you moved out of, and a description of what you do that stopped being true a couple of years ago.

Every one of those was correct the day somebody typed it.

Google asks for consistency in about as plain a sentence as Google writes, in its Business Profile guidelines: "Your name should reflect your business's real-world name, as used consistently on your storefront, website, stationery, and as known to customers."

Consistently is carrying an awful lot of weight in that sentence.

Our read, and I'll flag it as ours because nobody's published a study measuring this for businesses your size: this is where small companies actually come unstuck. Not from being too quiet. From being described four slightly different ways by four sources that never compared notes.

Which is the same thing we keep going on about, arriving from outside the building. The facts about your business are data too. They just live in systems you don't own, and they disagree with each other. Inside your company that gets you a wrong invoice. Outside it, it gets you silence.

✦ Made with ChatGPT Images 2.0 — we're transparent about AI. See the exact prompt on the original post.

The file everyone tells you to add

Now, there's a whole industry grown up around this, and a lot of what it sells you is files.

Schema markup. An llms.txt. Structured data packages, billed monthly.

Google's own documentation is ruder about this than I'd be.

From that same guide: "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add." On the AI-specific files: "You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search." It then says Google Search ignores them.

And one more, worth reading twice if anybody's offered to go and build you brand mentions at scale: "Seeking inauthentic 'mentions' across the web isn't as helpful as it might seem."

Structured data still earns its keep for ordinary search results, and we put it on client sites. It's just a poor answer to why an assistant skips past you.

So what does move the needle? The one properly peer-reviewed piece of work in this field is a 2024 paper out of Princeton, IIT Delhi and Georgia Tech called Generative Engine Optimization, presented at the KDD conference. They ran 10,000 queries and tested nine different ways of rewriting a page to see which got it quoted more often. Adding quotations came out top, then adding statistics, then citing your sources.

Keyword stuffing scored worse than changing nothing at all, which I enjoyed more than I probably should have.

That was measured against the engines of 2024 and I wouldn't bet the exact numbers still hold. The direction is worth having, though, because it's cheap and it repeats: pages that make specific, checkable, quotable claims get quoted. Pages that gesture don't. It's the outward-facing cousin of something we wrote recently about giving a reviewer something concrete to check against.

What a mention is actually worth

Before you go off and do any of this, the honest part.

In that same Pew study, of all the visits to a results page carrying an AI summary, people clicked a source link inside the summary on 1% of them. On those same pages, clicks through to an ordinary search result ran at 8%, against 15% where no summary showed up at all.

One percent. Do that math for a second: get yourself quoted in a hundred AI answers and roughly one person clicks through.

So if you were planning this as a traffic channel, plan for approximately no traffic.

It's still worth having, just for a different reason than you'd think. A mention behaves like a referral. Somebody asks for a recommendation, your name comes up next to two others, and later they look you up directly or walk into a conversation already knowing you exist. Our read, and it's a read rather than a finding: that visit lands as a direct one, and nothing in your analytics will ever tell you where it came from.

Which makes this a brand problem wearing a technical costume. What you're after is being the name that gets said, and it pays off somewhere you can't see it.

Where to start this week

Check that a machine can reach you. Open your robots.txt and see what it's blocking. Search your own business name and confirm your site actually comes up. Claim your Google Business Profile if you never got round to it, and get the details right, because for local recommendations it's the most authoritative description of you that exists anywhere.

Make your basic facts agree. Name, one-line description, address, phone. Write the correct version down once, then go and fix the places it's wrong. It's dull, and it's the highest-value hour on this list.

Give the web something worth quoting. One page that answers a real question properly, with actual numbers and named sources, beats ten pages of positioning. If your site says you deliver tailored solutions, there's nothing there for a machine to lift.

The ten-minute version

Ask ChatGPT, then Google, three things: what your business does, whether it can recommend a business like yours in your town, and who the good options are.

Then read the answers for what's wrong, rather than for whether you're in them. A wrong phone number. An old service list. A competitor described in the exact words you use about yourself. That's your to-do list, and you just got it for free.

None of this is quick. Entity information moves slowly, directories update whenever they feel like it, and there's no button. But it's finite work, most of it costs nothing, and unlike a website redesign it stays fixed once you've fixed it.

If you've gone and looked yourself up and didn't much care for what the machine thinks you are, that's worth an hour of conversation. Start a no-pressure conversation about which of your facts we'd straighten out first.

Practical AI. Clear process. Real business value.

— Brian, NuWay Biz Solutions

P.S. This website serves an llms.txt. Thirty-three entries, neatly organised, and I was quite pleased with myself, and Google's documentation says Google Search ignores it. So we're taking our own medicine here — working through this exact list on our own business, in public, while telling you to go and do it on yours. I'll report back on what actually moved.

The AI wrote it in ten seconds. Checking it takes an hour.

Brian Pawl — Thu, 13 Aug 2026 15:29:59 +0000

Originally published on the NuWay Biz Solutions blog.

✦ Cover image: Made with ChatGPT Images 2.0 — we're transparent about AI. See the exact prompt on the original post.

Last week I had an AI write up a traffic report for a website. Simple job: pull the last year, tell me what happened.

It came back looking excellent. Clean tables, a tidy summary at the top, numbers carried out to two decimal places. The headline finding was that traffic had fallen off a cliff.

It hadn't.

Most of the traffic in that report belonged to a completely different website. An older site had lived on that domain years earlier, and it had been quietly reporting into the same analytics account the whole time. Of the sessions in the report, 963 of 1,106 belonged to it rather than to the site anyone was asking about. Somebody else's visitors, from a site that no longer exists, faithfully counted and beautifully presented.

Here is what bothers me about it. Nothing in that report was sloppy. Every number was really in the data, pulled correctly, added up correctly, formatted correctly. They were just describing the wrong website.

It took three passes to catch. On the second pass, the corrected version made the same mistake again in a different spot, which I only found because someone went looking for it specifically.

If your team is buried under AI-generated work that nobody has the hours to properly check, you are not imagining it, and it is not a discipline problem on your end.

40% of US desk workers received AI-generated work that looked finished and wasn't, in a single month

Making got cheap. Checking cost exactly what it always did.

Writing a first draft of something used to take forty minutes. Now it takes about ten seconds. That is a real gain and I am not going to pretend otherwise.

Reading that draft closely enough to decide whether it is right takes as long as it ever did. Longer, actually, and there is a specific reason why.

When a colleague hands you a draft, you know them. You know they are excellent on the numbers and vague on the timeline, so you read hard in one place and skim the rest. That shortcut is most of what makes review survivable. AI output takes it away from you. It is uniformly polished and uniformly confident, so the wrong sentence looks precisely like the right one and the mistakes are scattered evenly through it instead of clustering where a tired person would put them.

Researchers at BetterUp Labs and Stanford's Social Media Lab put a number on how common this has become. In a September 2025 survey of full-time US desk workers, 40% said they had received AI-generated work in the previous month that looked good and lacked substance (their term for it is workslop), and sorting out each instance took about two hours. The write-up ran in Harvard Business Review.

Two hours. For one item. That is not a rounding error on a productivity gain, that is the gain, handed to somebody else.

Every hour AI saves at the making step lands on somebody at the checking step. That is the same hour. It just moved desks.

The checking is the expensive half, and nobody budgeted for it

Workday looked at the same problem from the company's side. Their January 2026 research found that nearly 40% of AI time savings are lost to rework — correcting errors, rewriting content, verifying outputs — and that only 14% of employees consistently get a clear positive result out of AI at all.

One thing to know before you apply that to yourself: every person in that study works at a company doing over $100 million a year. That is a much larger business than yours, and it matters, because it cuts the way you would not expect.

A company that size has slack. Somebody's afternoon absorbs the rework and it never shows up as a line item. You do not have that person. When the checking lands in a six-person shop it lands on the owner, at night, on top of everything else.

The boring reason it takes so long

To check anything quickly, you need two things: you need to know what a correct answer would look like, and you need something trustworthy to hold it up against.

Most businesses have written down neither. So checking collapses into reading it over and seeing whether it feels about right, which is proofreading with extra anxiety.

Go back to that traffic report for a second. Nobody involved was careless. What was missing was the one sentence that would have killed the error instantly: this report covers traffic to the current site, from the day its tracking was installed, and nothing before it. There was no line in the sand, so there was nothing for a wrong number to fail against.

This is the same thing we keep going on about, arriving from a direction that surprised me. Connected, clean, trustworthy data is usually sold as the thing that makes AI work. It is also the thing that makes checking AI cheap, and that turns out to be where the hours actually go.

It is a cousin of something we wrote about a few weeks back: a stopped automation looks exactly like a quiet week. Silence reads as fine. So does a confident paragraph.

✦ Made with ChatGPT Images 2.0 — we're transparent about AI. See the exact prompt on the original post.

Three rules that make checking fast

None of this is an argument for using less AI. The early wins are real and we have written about where they come from. This is about not handing the bill to whoever sits downstream.

Decide what right looks like before anything gets written. One sentence, written first, saying what the output covers and what a correct version would contain. It takes thirty seconds and it is the single highest-leverage thing on this list. A reviewer holding that sentence checks in minutes. A reviewer without it has to work backwards from the output to guess what was intended, which is most of the two hours.

Check against a number, not a feeling. If somebody can tell you what last month's figure actually was, verifying a summary takes a minute. If the only way to find out is to rebuild it from four exports, every check becomes a small project and, realistically, nobody does it. That is the one honest number problem, and it is why it is worth solving before the AI arrives rather than after.

Only point it at work you could verify quickly by hand. This is the rule that saves you from the worst version. If checking the output takes longer than doing the task yourself would have, you have not automated anything. You have bought a slower version of the job and added a step where a mistake can hide.

The measurement worth taking this week

Pick one thing your team produced with AI in the last fortnight. Ask whoever reviewed it two questions: how long did that take you, and how did you know it was right?

The second answer is the one to listen to. If it comes back as "it looked fine" or "I read it through," what you have is a reading step wearing a checking step's name. Reading catches typos. It sails straight past a number that describes the wrong month.

Worth saying plainly: this is the ops-side twin of something we wrote for creative teams about using AI without producing slop. That one is about the work you make. This one is about the work that lands in your inbox from everybody else.

Where you actually are right now

[ ] Somebody on the team regularly receives AI-written work they have to substantially redo
[ ] Nobody writes down what the output should contain before it gets generated
[ ] Checking a number means rebuilding it from exports rather than looking it up
[ ] The person doing the checking is the owner, and it happens after hours
[ ] Something AI-generated has reached a customer, or a decision, without a real check

Two or more of these and the checking step is where your time is going, not the making step. That is fixable, and the fix starts before anything gets written rather than after.

The businesses handling this well can answer "how would we know this is right?" without a pause, because they sorted out what they trust before they started pointing software at it. The tools they use are the same ones you have.

If your team is producing more than anyone can check, that is worth an hour of conversation. Start a no-pressure conversation about which of your numbers we would make trustworthy first — the one that would take the most checking off your plate.

Practical AI. Clear process. Real business value.

— Brian, NuWay Biz Solutions

P.S. While writing this, I went to double-check the workslop study's own figures. BetterUp's landing page says the survey covered 1,150 people and that each incident takes two hours to resolve. BetterUp's own blog post, about the same study, says 1,004 people and one hour fifty-one minutes. Same company, same research, two pages, two answers. I have used the round number and told you where it comes from, because I genuinely do not know which is right. Two hours of somebody's afternoon, and I could not verify it in twenty minutes.

Before you buy an AI agent, answer these five questions

Brian Pawl — Thu, 06 Aug 2026 19:43:11 +0000

Originally published on the NuWay Biz Solutions blog.

✦ Cover image: Made with ChatGPT — we're transparent about AI. See the exact prompt on the original post.

"It's an AI agent. It handles your inbound automatically."

"Handles it how?"

"Automatically."

That exchange is a fair summary of the market right now, and it is not really the salesperson's fault. The word arrived faster than the meaning did.

Gartner, the technology research firm a lot of corporate software budgets get built on, has a name for the resulting mess. They call it agent washing: taking an assistant, a chatbot or an old automation tool, and reissuing it with the new label. Of the thousands of vendors selling agents, they counted about 130 selling a real one in mid-2025.

Gartner counted about 130 real agent vendors in mid-2025, out of thousands

You do not need to work out which 130. You need five answers, and every one of them is a question about your business rather than a question about the software.

1. What does it do when nobody is watching it?

This is the question the label is hiding. Ask the vendor to name one thing their software does with no person approving it first.

If every answer comes back as it suggests, it drafts, it recommends, it surfaces, then you are being shown an assistant. Which is genuinely useful and worth paying for. It is just a different product at a different price, and you should know which one is on the invoice.

An assistant

Waits for you to ask
Hands the work back for you to approve
Its mistakes stay inside your building
Usually priced per person per month

An agent

Decides on its own that something needs doing
Does it, then tells you
Its mistakes reach customers before you see them
Usually priced per action, per outcome, or per run

Gartner's analyst on this, Anushree Verma, put the useful half of it more bluntly than a vendor ever will: “Many use cases positioned as agentic today don't require agentic implementations.”

Plenty of jobs are better done by something that asks first.

2. What can it do that I can't take back?

Everything an agent is allowed to touch falls into two piles, and the piles are not the same size.

One pile is reversible. It tagged the wrong lead, it filed a job under the wrong customer, it wrote a note nobody needed. You fix it in a minute and no one outside the business ever knew.

The other pile leaves the building. It emailed a customer. It issued a refund. It quoted a price. It cancelled tomorrow's appointment.

You cannot un-send an email to a customer, and the agent will not know it should have asked.

So get the list. The one that names each action it can take with no human in between, and where that action lands. Vendors lead with a longer and friendlier list, so ask for this one specifically.

Then say out loud which of those you are comfortable with on a Saturday when nobody is looking at a screen. That is the real permission you are granting, and it is worth writing into the contract rather than discovering later.

3. Which of my numbers does it read?

An agent acts on what it reads. That single fact is why this question sits in the middle of the list rather than at the end.

We have made this point about what AI-ready data actually means for a while, and about why the AI you already pay for stalled out. An agent raises the stakes on it. A dashboard built on numbers that disagree wastes your afternoon. An agent built on the same numbers sends the customer a bill for the wrong amount.

So ask which systems it reads from, and then ask yourself the harder half: would you personally bet on those numbers being right this morning?

If the honest answer is no, do that work first. It pays for itself whether or not you ever sign. Match the build to the mess, same as always.

4. What does this cost in my busiest month?

The software you already buy costs the same in February as it does in July. Agents mostly are not sold that way.

Intercom publishes its price plainly, which is more than most vendors do, so it makes a clean example: its Fin agent is $0.99 per outcome, charged once per conversation. An outcome counts when the customer confirms the issue is resolved, when the agent completes a workflow, or when the customer doesn't ask for more help after it responds.

Read that last one twice. Silence counts as success, which is a reasonable way to measure it and also exactly the number you want to watch.

Usage pricing is a perfectly reasonable way to buy this. It just asks one piece of arithmetic of you first: take your worst week last year, the one where the phone did not stop, and work out what that week would have cost at the quoted rate.

Then ask what happens when you hit the ceiling on your plan, because the answer is rarely nothing.

5. How would I know it did the job right?

We wrote recently that a stopped automation looks exactly like a quiet week. An agent inherits that blind spot and adds a worse one of its own.

An automation does one fixed thing every time its trigger fires, so at least you know what it was trying to do. An agent picked the thing. It can run perfectly, do exactly what it decided to do, and still have done the wrong thing, confidently, in your name.

There is a good measurement of how often an agent finishes the job at all. Researchers at Carnegie Mellon built a fake company, a simulated software business with its own internal websites and data, and turned the best agents available when they ran it loose on the work its staff would handle. The most competitive agent finished 30% of the tasks on its own. Their conclusion is worth keeping in your head during the sales call: simpler tasks often go fine, and longer multi-step ones are still beyond what these systems can do.

That number will improve. It has not improved enough for you to skip the check.

So ask what it produces that a person can look at afterwards. A record of what it decided and why, in language you can read on a phone, with a way to find the ones it got wrong before your customer does. A dashboard of how busy it has been will not tell you any of that.

Ask this one on the call

"Show me the last hundred things it did in an account like mine, and point at the ones it got wrong."

A vendor with a real product will have a screen for this, because their other customers ask. A vendor who offers you a demo environment instead has told you something useful, for free.

When the answer is yes

✦ Made with ChatGPT — we're transparent about AI. See the exact prompt on the original post.

None of this is an argument against buying one. There are jobs where an agent earns its keep quickly, and they have a recognisable pattern: the same decision, many times a week, against a rule you could write down on an index card.

Booking requests. First-pass triage on inbound. Chasing the same three pieces of missing information off every new job. Work where being right ninety-something percent of the time and flagging the rest is a genuine improvement on what happens today, which is that it waits until Thursday.

Verma's own advice on where to start says it better than we can: “They can start by using AI agents when decisions are needed, automation for routine workflows and assistants for simple retrieval.” Three different tools. Most of what gets pitched as the first one is doing the job of the other two.

The businesses that get value out of this are the ones that could answer the five questions before they signed.

[ ] I can name one thing it does with no person approving it
[ ] I have the list of actions it can take that reach a customer
[ ] I know which of my systems it reads, and I would bet on those numbers this morning
[ ] I have worked out the bill for my busiest week
[ ] I know what it produces that I can check afterwards

Four or more and you are ready to have a serious conversation with a vendor. Fewer, and the gaps are your agenda for the next call: whichever one you couldn't tick is the one that will cost you.

Take the list to the next call. You will know inside ten minutes whether the person on the other end has thought about their product as hard as you are about to think about your business.

And if you get the answers and still cannot tell whether the thing is worth buying, send us what you were pitched and we will tell you what it actually is. Sometimes the answer is that you already own something that does the job.

Practical AI. Clear process. Real business value.

— Brian, NuWay Biz Solutions

P.S. One more to try on the next demo, and it is the cheapest of the lot. Ask what the software does when it is not sure. A real answer sounds like a rule: it stops, it flags, it asks a human. If what comes back is a shrug, or that it uses its best judgement, you have just learned the most important thing on this list.

A stopped automation looks exactly like a quiet week

Brian Pawl — Fri, 24 Jul 2026 15:05:59 +0000

Originally published on the NuWay Biz Solutions blog.

✦ Cover image: Made with ChatGPT — we're transparent about AI. See the exact prompt on the original post.

Zapier will switch one of your automations off once it errors 95% of the time it runs, and only after it has run more than twenty times in the past week. That is a reasonable threshold for a company running a few hundred million of these things. Below it, the automation stays on.

Zapier turns an automation off after it errors 95% of the time it runs

Sit with what that means for one business. Something that fails half the time is, by that measure, working fine.

And that is the generous case, because at least a failure is an event. It gets counted. Zapier will email you about it immediately, every time, without you configuring anything, because that is the default setting.

The ones that hurt you never produce an event at all.

The automation stops running, and nothing errors

An error notification needs an error. An error needs a run. Follow that backwards and you get the whole problem: an automation that has stopped triggering generates no runs, so no errors, so no emails, so no alarm anywhere in the system.

It is not hiding the failure from you. There is no failure to report. As far as the tool is concerned, nothing happened, which is exactly what it would report on a genuinely slow week.

An automation that has stopped running looks exactly like an automation with nothing to do.

This happens more mundanely than people expect. Somebody at a software company changes something on a Tuesday and the trigger stops matching. A password change or an expired login quietly drops the connection. Somebody turns an automation off to test something and never turns it back on. Or you hit the ceiling on your plan, at which point Zapier holds actions across every automation in the account until the next billing cycle.

Notice that only some of those are anybody's fault, and none of them make a sound.

Sets off the alarm

A step runs and returns an error
The same step errors on nearly every run for a week
A connected account needs reconnecting

Sets off nothing

The trigger stops firing, so nothing runs at all
It runs and writes an empty field
It runs twice and creates the record twice
It fails a third of the time, forever

It runs, and writes the wrong thing

The second failure mode is worse, because it produces a success. Somebody renames a field on one side, and the automation keeps firing on schedule and writes a blank into the other side. Every run goes green. Your customer list slowly fills with records missing the phone number.

Nothing in that chain is broken in a way software can detect. The trigger triggered, the step ran, the record was created. A tool that moves data between two apps has no opinion about whether the data was any good, and asking it to have one is asking it to understand your business.

The third mode is the same idea pointing the other way: it runs twice and you get two of everything. We spent a whole article on what that costs you when it lands in your books, so it is enough to say here that a double is also a success, twice.

The pattern behind all three: your tools can tell you a step failed. They cannot tell you the result was wrong, and they cannot tell you about work that never happened.

Every one of these tools has the same gap

Zapier is named throughout because it publishes its thresholds and its defaults, which is more transparency than most tools offer. Every other tool that moves data when something happens has the same blind spot, whatever the logo.

The check is the thing you are missing

The fix is not a better platform. If your automations are mostly working, keep them. Moving everything to a new tool buys you a fresh set of quirks and the identical blind spot.

What you are missing is a check: something that confirms the work actually happened, and happened correctly, without relying on the tool that did the work to grade itself.

At the smallest scale that check is a person and a calendar reminder.

Count both sides over the same window. Forty-one people filled in the form last week. Are there forty-one new rows in the customer list? That single comparison catches all three failure modes at once, which no amount of staring at the automation history will do, because the history can only show you the runs that happened.

✦ Made with ChatGPT — we're transparent about AI. See the exact prompt on the original post.

The ten-minute version

Pick the automation that touches your money or your customers. Count what went in over the last full week at the source, then count what came out at the destination. Write both numbers down. If they match, you have learned your automation is alive, which you did not actually know this morning. If they do not, you have found the thing before your customer did.

Above a certain size, that check has to run on its own. Something that expects work to happen and complains when it sees none. The rule to hold onto is that no errors does not mean nothing is wrong: you want proof of work completed, not the absence of complaints.

Not everything needs this. If an automation moves something you would notice missing by lunchtime, your business is already the check, and adding a second one is theatre. The ones that need watching are the quiet ones: the automations feeding a number somebody trusts, where nobody is waiting on the result and nothing looks wrong until a decision has been made on it.

It's the same story as the spreadsheet that had been quietly wrong for a year. A number gets believed because it appeared where numbers appear. Automating the work does not remove that risk. It removes the person who used to trip over it.

[ ] I know which of my automations touch money or customer records
[ ] I would find out within a day if one stopped running
[ ] Somebody checks that both sides still agree, on a schedule
[ ] The error emails go to a person who reads them, not a shared inbox
[ ] I could rebuild the important ones if the person who set them up left
[ ] I know what each one is supposed to produce in a normal week

Four or more means your automations are supervised, which puts you ahead of most. Look at the ones you couldn't tick and start with whichever touches money.

Start with one automation. The one that would embarrass you in front of a customer if it had been dead since March.

Count both sides. Then decide whether you knew the answer before you counted.

If you've got a handful of automations running and nobody's sure which ones are still doing their job, tell us what they're meant to be doing and we'll tell you which ones are worth putting a check on. Some of them won't be.

Practical AI. Clear process. Real business value.

— Brian, NuWay Biz Solutions

P.S. This question gets much sharper the moment somebody tries to sell you an AI agent that runs on top of all this. That one's coming.

The spreadsheet had been quietly wrong for a year

Brian Pawl — Tue, 21 Jul 2026 04:19:30 +0000

Originally published on the NuWay Biz Solutions blog.

✦ Cover image: Made with ChatGPT — we're transparent about AI. See the exact prompt on the original post.

The first thing they showed us was a file.

A landscaping company, about a dozen crew, two office staff. We'd been brought in to look at whether any of their admin work could be handed to a machine, and the owner opened a laptop, clicked through three folders, and turned the screen around. Column after column of green. Somebody had clearly cared about this thing for a long time.

It was the monthly numbers. Every job, what it brought in, what the crew hours cost, what the materials cost, and at the bottom of the third tab, the number the whole company steered by: what each crew actually made last month.

The office manager built it fresh every month. First working day, coffee, door shut. Export from the accounting system, export from the scheduling app, paste both in, fix the job names that came through spelled differently, drag the formulas down, check the pivots didn't break. About three hours if nothing went wrong. Longer in January.

The owner told us he trusted that sheet more than anything else in the business.

Then we asked the office manager to walk us through one column, and she got about halfway and stopped. She didn't build that part. A bookkeeper had set the sheet up four years earlier and left, and she'd inherited it the way you inherit a house with a boiler you don't understand: it works, you don't touch the pipes, and you learn which noises are normal.

So we spent an afternoon rebuilding it from the raw exports, on our own, to see if we landed on the same answer.

Mostly we did. One line was off. A kind of subcontractor cost was landing just outside the block of cells the crew-profit formula added up, so every crew came out looking a little better than it was. Small on any one job. Not small across a year, which is roughly how long it had been that way. Long enough that the owner had already made a real call on the back of it, about which kind of work to chase more of.

Nobody had done anything careless. The sheet was good work.

It had just never been checked by anyone, including the two people who trusted it most.

The obvious advice, and why he was right to resist it

Walk into that room as a consultant and the pitch writes itself. Three hours a month. A fragile file. One person who understands it. Automate the whole thing, buy a dashboard, get the humans out of the loop.

The owner pushed back before we'd finished saying it, and his version was better than the pitch. That hour isn't data entry, he said. It's the only hour all month I actually look at my business. Touching every number is how I notice the job that went sideways, the supplier who raised prices without telling anyone, the crew whose hours stopped matching their output. Automate the typing and you automate away the noticing.

He was right, and it's the part the pitch never has an answer for.

Those are two different jobs wearing the same hour. Gathering the numbers is mechanical, and a machine could do it at 6am on the first. Reading them is judgment, and judgment is the part that catches things. One of those two jobs is worth defending.

Automate the gathering. Keep the reading. They only look like the same task because the same person does them back to back.

That would be the end of a shorter article. The complication is that reading only works if the thing you're reading is true, and this is where the manual sheet starts losing the argument it was winning.

A spreadsheet gets believed because it's a spreadsheet

There's a group of researchers in Europe who have spent years collecting the ways spreadsheets go wrong in real organisations. One of their case studies follows a small non-profit run by a volunteer committee, about the level of formality a lot of the businesses we work with have. Two volunteers built a cash-flow model in Excel to plan the next 35 years. It said the organisation was heading for trouble and couldn't afford to replace its aging equipment, so the board froze the spending.

The model had understated future cash flow by more than a factor of two. It also ran its whole calculation at a negative discount rate, which is a bit like doing your budget with a calculator that adds when you press minus.

When someone wrote up a step-by-step test script showing exactly how to reproduce the errors, and sent it to the chairman and the treasurer, nothing happened. No response. No evidence anyone ever ran it.

The researcher's line for this is the sharpest description of the problem I've read: once information is loaded into a spreadsheet, it acquires properties it may not deserve. He lists them. Believability. Correctness. Appropriateness. Objectivity. Authority.

The green cells do that. The grid looks like arithmetic, so it reads as fact.

And notice the failure is identical to the one everybody warns you about with dashboards. A dashboard keeps rendering a beautiful chart of a wrong number and you find out in March. A trusted spreadsheet does exactly the same thing, with the same confidence, in a font you happen to like better. The landscaping company's sheet was wrong for a year in a file the owner trusted more than anything else he owned.

A big company would have someone whose actual job is to check that file. A company this size has the person who built it, and she has four other jobs.

The two things worth fixing

So the sheet stays, and automating it is only half a job. What's left is smaller than either, and duller, and it's most of what we do when we walk into one of these.

One. A second person has to be able to rebuild it. That means rebuilding it from the raw exports, not following along while somebody talks them through it. Where does each number come from, which export, which app, what does this column mean, and what is the sheet for. Written down in the file itself, because a document about the sheet lives in a folder nobody opens. The test is blunt: if the only person who can reproduce your monthly numbers goes on holiday for three weeks, does the business still know what it made?

The rule of thumb we use: if a spreadsheet would take more than about two days of work to reproduce from scratch, it has stopped being a document and become a piece of your business. Treat it accordingly.

Two. One number gets checked by hand, every month, before you trust the rest. Cash in the bank. Total deposits. Something you can verify at the source in two minutes without the sheet's help. If it doesn't tie out, you stop and find out why before you read anything else.

This is the easiest step to skip, and it's the one that would have caught the landscaping company a year earlier. It matters more, not less, once you automate anything, because the two failure modes are opposites. A manual sheet fails by being late. An automated one fails by being confidently wrong and on time.

Then automate the boring half, with your eyes open

Only now is it worth touching the tooling, and the honest version is that every path charges a toll.

The most common advice is to use the tools built into Excel to pull and join the data for you. It genuinely works, and it will join almost anything you point it at. What people don't mention is that it often won't refresh on its own in a browser, and on a Mac it can't reach most of the apps you'd want. So the monthly job doesn't disappear. It moves onto one particular person's Windows laptop, which is the same dependency you started with wearing a better outfit.

The free dashboard tools will blend a handful of sources and are often the right call, up to the point where they stop letting you add more. The services that wire apps together are built for things that happen constantly, which makes them a poor fit for a job that runs once a month: something that fires twelve times a year can break in February and sit broken until May.

We put these in for people constantly. The point is to know which one you've picked, who owns it, and how you'd find out it stopped.

Start with the thing you already pay for. A surprising number of accounting and scheduling systems will email a scheduled report on the first of the month, and nobody ever turned it on. That plus a folder where the exports always land with the same filenames kills most of the three hours, costs nothing, and doesn't create a new machine only one person understands.

Why the sheet is there in the first place

It's easy to read all this as an argument against spreadsheets, and it isn't. When Intuit surveyed 1,500 US firms under 100 employees in 2022, reported in a National Bureau of Economic Research working paper, 46% said they use spreadsheets to manage or record their money.

46% of small US firms use spreadsheets to record their money

They could pick more than one answer, and more of them named accounting software than named spreadsheets. So in plenty of these businesses the sheet isn't standing in for the accounting system. It's sitting next to it, doing a job the accounting system won't do, which is usually joining money to something the money doesn't know about: a job, a crew, a channel, a customer. That joining is most of what "ready" data actually means in practice.

That's a gap, left where two systems wouldn't talk to each other, and a person stepped into it. The spreadsheet is the shape of that person's work.

Which is why the answer is usually to treat the sheet as a piece of the business rather than a document: somebody else can rebuild it, something checks it, and the machine does the parts that don't need a brain.

If it grows past what a sheet can carry, and for some businesses it does, that's when the conversation about connecting the tools properly earns its place. Plenty of businesses aren't there yet, and don't need to be.

The three-week test

If there's a file your business steers by, open it. Pick one number on it that matters, and try to answer two questions without asking the person who maintains it: where did this come from, and how would I know if it were wrong? If both answers are missing, that's a morning's work, not a project.

We finished up with the landscaping company in about a week. Two scheduled exports, a folder, a page of notes at the front of the workbook explaining every column, and one figure the office manager ties back to the bank before she sends anything to the owner. Still three tabs of green. Still hers.

The rebuild takes her about forty minutes now, and she still does it on the first working day of the month, with the door shut, because that's when she reads it.

If there's a file like that holding up your business and nobody's completely sure how it works, tell us what it does and we'll tell you whether it needs automating or just needs an owner. It's usually the second one.

Practical AI. Clear process. Real business value.

— Brian, NuWay Biz Solutions

P.S. If you went looking for that file and found four versions of it with different dates in the filename, pick the one everybody actually opens and delete the argument.

You don't have a dashboard problem. You have a data-readiness problem.

Brian Pawl — Fri, 17 Jul 2026 20:34:12 +0000

Originally published on the NuWay Biz Solutions blog.

✦ Cover image: Made with ChatGPT — we're transparent about AI. See the exact prompt on the original post.

You know your business, and you wanted one screen that finally showed it to you straight. Real profit by job. Which channels actually make money. The number you'd bet a new hire on, in one place, not scattered across six apps that each hold a slice. So you did the reasonable thing. You got a dashboard.

And it showed you six numbers that don't agree. Sales say one thing, the bank says another, the profit-per-job figure either looks wrong or won't pull at all. It's tempting to decide the dashboard was a waste of money.

It wasn't. A dashboard was the right instinct. You just skipped the part that makes one work.

A dashboard is a window onto your data, not a fix for it

This is the piece the software ads leave out. A dashboard, or a full BI tool, doesn't create the truth about your business. It displays whatever data sits underneath it. Feed it clean, connected, reconciled numbers and it gives you a crisp, trustworthy view. Feed it six disconnected apps whose numbers were never reconciled, and it does exactly what it's built to do: it shows you the disagreement, faster and in nicer colors.

So a better screen won't help you, because the data foundation under it was never built. That's the whole thing, and it's the same reason your AI stalled out if you've tried that: point either one at a mess and it faithfully reflects the mess back at you. The real work lives in the data, which is what we mean by data-readiness.

A dashboard is only ever as honest as the data feeding it. Fix the data and the same dashboard starts telling the truth.

Why the six numbers disagree in the first place

≈36 apps at companies under 50 people (Okta, 2024)

Start with how many places your money actually lives. Even the smallest companies, the ones with fewer than fifty people, run about thirty-six different apps, according to Okta's look at 18,000 businesses. So six is the polite version of this. And those apps disagree for concrete, unmagical reasons, which is worth seeing clearly, because it's exactly what a dashboard cannot fix on its own.

The same job is "Job #4021" in your field app and "the Smith kitchen" in your books, so nothing lines them up. One system counts a sale the day you invoice it, another the day the money lands, so the same week shows two revenue figures. The labor for a job lives in one tool and the materials in another, and no one ever joined them, so "what did we actually make on that job" isn't a number anywhere. Point ten dashboards at that and you get ten renderings of the same confusion.

It gets worse every time a number is copied by hand between tools, because people get simple data entry wrong somewhere between one and five percent of the time. We walked through that number in the last piece, and the errors compound, so a figure that passes through six hand-touched steps has six chances to drift. None of that is a visualization problem. It's a foundation problem.

✦ Made with ChatGPT — we're transparent about AI. See the exact prompt on the original post.

The number that isn't anywhere yet

An owner who tracks profitability job by job described the real frustration, and it isn't ignorance. She knows her business cold. But getting true per-job profit meant combining the labor from one tool with the materials from another by hand, every month, long after the jobs were done. The number she needed most wasn't hiding in a report she hadn't opened. It didn't exist yet in any single place, because nothing had connected the two systems that held its halves.

What actually gets you one trustworthy screen

Your instinct was right. You just need to build the thing that feeds it. Four steps, and only the third one waits on the calendar.

📊 This section has an interactive graphic — see the full version on the original article.

Connect the sources into one place. Stop asking six apps to each keep their own tally. Pipe them into a single home so the data actually sits together. For a lot of small businesses that home is the system most of your money already lands in, your accounting tool or your point-of-sale. For more complex ones it's a small warehouse that connectors feed. Either way, the move is the same: one place, not six.

Align them so they agree. This is the unglamorous step that does the real work, and it's the one nobody sells you. Getting the tools into one place isn't enough if they still call the same customer three different names and count revenue two different ways. Aligning means deciding, once, that a customer is one customer, a job is one job, and a dollar of revenue is counted the same way everywhere, then making every source follow that. Skip this and your shiny new central dashboard shows the same disagreement it always did.

Reconcile it to your books. Tidy isn't the same as true. Once the data is connected and aligned, you tie it back to your reconciled, closed books, so the picture on the screen matches the money that actually moved. This is where the accounting close earns its keep, as one step in the foundation rather than the whole story. A bookkeeper described her own monthly reconcile as "the actual accounting work takes maybe half a day", and turning it into a report her client could trust took another four days, because reconciling waits on late-arriving fees, refunds, and unpaid invoices. That's the one step that lags the calendar, and it's normal.

Now show it. With the data connected, aligned, and reconciled, the dashboard finally does the job you hired it for. One centralized view, one version of revenue, real profit sliceable by job and by channel, the number you'd put your name to. The tool didn't change. What changed is that it now has something honest to display. And the same clean foundation is what any AI you try next will stand on, which is the whole point.

Is your data ready for the screen you want?

Before you buy or build anything, it's worth a blunt self-check. This is what "ready" actually looks like under a dashboard that tells the truth.

[ ] Every tool that touches money feeds into one central place
[ ] The same customer or job has one identity across tools, not five spellings
[ ] Revenue is counted the same way everywhere (on invoice, or on payment, but consistently)
[ ] Labor, materials, and fees for a job can actually be joined into one figure
[ ] The central numbers tie back to your reconciled, closed books
[ ] You can get "profit by job" or "profit by channel" without rebuilding a spreadsheet

Tick five of these and your data is genuinely ready. A dashboard built on it will tell the truth, and so will anything you point at it next, including AI.

How big a build you actually need

How much you build depends entirely on how tangled your money is, and most owners over-buy here. The goal is a screen you trust, not a project we get to sell you.

If you're a genuinely simple business, one bank account, one sales platform, no inventory and no unpaid invoices sitting out, you don't need a dashboard or any of this. Your books are basically current, so a good bookkeeper and your P&L already give you the honest view. For everyone past that point, there are two sane paths, and the difference is complexity, not ambition.

Consolidate. The light path.

Most of your money already lands in one system
A handful of sources, one or two sales channels
That system's built-in reporting becomes your dashboard

Build a BI layer. The heavier path.

Money genuinely lives across several channels that don't reconcile
You need profit sliced by product, channel, and job
A real BI tool, fed by connectors, on a small warehouse

Most small businesses live on the left. The multi-channel seller, running Shopify and Amazon and a marketplace and ad platforms at once, is the clearest case for the right: no single tool can reconcile that, so the BI layer is doing genuine work, not adding shine. Either way, own the foundation you build. Plenty of owners handed their books and their data entirely to an outside service and got locked out overnight when it failed. Rent the labor if you like. Keep the central place, and the login, yours.

That's the whole shape of it. A dashboard was always the right idea, just pointed at data that wasn't ready. Connect your sources, align them, reconcile them to your books, and the same screen that showed you six arguments starts showing you one answer.

One question, six apps

Pick a single question you'd love a screen to answer: profit on last month's biggest job, or which channel actually paid off. Now try to answer it with the tools you already have, and count how many you have to open and cross-reference by hand. That count is the size of your data-readiness gap. It's also exactly the gap a dashboard can't close for you until the data underneath it is connected.

Getting a business to one trustworthy screen, so the dashboard tells the truth and whatever you point at your data next actually works, is the work we do. If you're tired of six apps that won't agree, tell us what you're running on and we'll tell you honestly how big a build it really needs, which is usually smaller than you'd think.

Practical AI. Clear process. Real business value.

— Brian, NuWay Biz Solutions

You already know what typing it in twice costs you. You're counting the wrong thing.

Brian Pawl — Thu, 16 Jul 2026 15:26:00 +0000

Originally published on the NuWay Biz Solutions blog.

✦ Cover image: Made with ChatGPT (OpenAI) — we're transparent about AI. See the exact prompt on the original post.

Somewhere in your business right now, a person is typing something into one screen that already exists on another screen.

You know about it. You've probably known for about a year. You might even have a rough number in your head for what it costs you: a few hours a week, somebody's Saturday, the job that always gets done later.

And you've decided to live with it.

I want to start by saying that decision is more reasonable than most people in my line of work will admit. Then I want to show you the part of the bill you almost certainly haven't seen.

Back in January an owner wrote up his weekend, and you can feel the whole problem in the title: "Spent my Saturday manually matching 47 invoices to bank payments, there has to be a better way."

Three clients had paid him in lump sums with no reference numbers. So there he sat with a calculator, working out which invoices added up to which deposits. Four hours. On a Saturday.

The replies did what replies do. Buy the accounting software, it's thirty bucks a month. What's your time worth?

But he'd already answered that, right there in the post. The software looked expensive and complicated, he said, for what was basically just a matching problem. He knew what the Saturday had cost him. He'd looked at the fix and decided the fix was worse.

He wasn't obviously wrong.

The case for leaving it alone is better than consultants admit

People who sell this stuff tend to skip this part, so let's not.

When you get two pieces of software talking to each other, the connection between them is fragile. Somebody at a company you've never heard of changes something on a Tuesday, and your setup quietly stops working. Nobody tells you. You find out three weeks later, when a number looks wrong.

A former Zapier employee said it plainly: these connections are brittle because the software underneath them keeps changing. And ask anyone who builds them for a living what actually breaks. It's never anything clever. It's someone typing a phone number with brackets around it.

A person never has this problem. A person sees the brackets, shrugs, and carries on. Your team quietly absorbs a hundred little changes like that every year, for free, and they will never once ring you at eleven at night because a supplier renamed something.

That's a real advantage. It's why plenty of sharp owners keep doing it by hand on purpose. If the whole mess adds up to four hours a month, the maths genuinely favours the calculator and the Saturday.

And it gets worse for my side of the argument. Connecting two systems badly can leave your books dirtier than leaving them apart, because the connection writes the same transaction in twice. This is common enough that Zapier publishes a help page called "Zap is creating duplicate data". Wire two systems together carelessly and you don't stop checking the numbers by hand. You just check them somewhere new.

So the maths on the hours holds up.

It's the rest of the bill that's missing.

The two costs that never land on your desk

When you worked out what this was costing you, you counted wages. Everyone does, because wages are the part you can see. Someone is sitting at a screen, and you know what that person costs an hour.

You didn't misprice the hours. You mispriced everything else on the bill.

There are two more costs, and neither of them ever arrives as a bill. Which is precisely why they're still there.

Owners do audit things, by the way. One of them went through his subscriptions last year and found his two-person company was paying £847 a month for software. Each one felt small, he said. Twenty-five here, thirty-five there. Together they'd quietly become a part-time salary.

That got caught because it turns up once a month with a total on it. The cost of your team retyping invoices turns up never, addressed to nobody. So it never has to compete for budget, and it never gets cancelled.

You can see it perfectly well. Your accounts have simply never heard of it.

The first one is the money that leaks out.

A contractor put it about as plainly as it can be put. A customer asks the crew on site for one extra thing. They say yes, they do the work. And then at the end of the month, writing up the invoice in a different system from the one where the job actually happened, he forgets to add it. His own verdict on his business: they were giving away labour and materials for free.

That's not wasted time. That's work you did, delivered, and never charged for, because the job lived over here and the invoice was written over there.

It runs the other way too. The same gap gives you the double charge, the wrong price, the stock count that says eleven when there are two. Every one of those is a phone call you'd rather not take.

The second one is trust.

An online retailer once described what quietly happens to a growing company, and I've never been able to shake it. The ops team believes the spreadsheet over the dashboard. Support keeps its own private tracker. The warehouse recounts the stock by hand before it ships anything. And the finance side exports everything into a spreadsheet before it believes a single number.

Sit with that last one. There is a dashboard. Nobody believes it.

✦ Made with ChatGPT (OpenAI) — we're transparent about AI. See the exact prompt on the original post.

So every real decision gets made twice: once by the system, and once by a person quietly double-checking the system. And in the gap between those two, you run your business on a number you're not quite sure about. Nobody bills you for that and nobody notices it, right up until the quarter you bet on the wrong one.

Typing it in twice doesn't just waste time. It plants wrong numbers.

There's one hard number in this whole article that I'd actually defend, and it surprised me.

Decades of studies: people get 1–5% of simple entries wrong

Ray Panko, a professor at the University of Hawaii who has spent a career cataloguing how often people get simple things wrong, gathered up the research. The studies land in a tight band. People make mistakes on roughly one to five percent of simple entries. Typing ten digits into a calculator goes wrong about five percent of the time.

Call it one slip somewhere between every twenty fields and every hundred. By a careful person. On a good day.

You can't train that away. Hiring someone more diligent moves it a little, not a lot. It's simply what it costs to move information with a human hand.

Which means every hop between your systems is quietly dropping wrong numbers into the exact places you bill from and decide from. Our man's four hours were the visible cost. The invoice he matched to the wrong deposit is the one that finds him in April.

The big scary numbers don't survive a click

Go looking for what all this costs and you'll turn up some enormous figures. It's worth knowing where they come from, because we went and followed them back.

The famous numbers, briefly

"Bad data costs the US $3.1 trillion a year." Every trail ends at a 2016 Harvard Business Review article that credits the figure to IBM and shows no workings for it. Go looking for IBM's analysis and there isn't one.

"McKinsey found small businesses recover 6.5 hours per employee per week." We couldn't find any such McKinsey report. The claim lives on the blogs of companies selling automation to small businesses, quoting one another.

"Data silos cost companies $12.9 million a year." That one's real, but it's Gartner talking about poor data quality, not silos, and the figure comes from customers of data-quality software guessing at their own losses. It's been quietly relabelled ever since.

We're an AI consultancy. Quoting one of those would have been the easiest paragraph in this article to write.

We went looking for a trustworthy number for a business your size and couldn't find one that survives its own footnotes. There's an irony in an industry that sells you data you can trust, running on data it can't.

So don't take a number from us. Go and get your own.

Look at your spreadsheets, not your software

The instinct is to list your apps. Skip that. Your spreadsheets are where the truth is.

Every spreadsheet in your business exists because two systems wouldn't talk to each other and a human stepped into the gap. Which makes that folder an unusually honest map of where your software is letting you down, drawn over years by the people doing the work.

The load-bearing spreadsheet audit (one afternoon, costs nothing)

List every spreadsheet your business actually uses. Next to each one, write down three things: who owns it, where its numbers come from, and what breaks if it disappears tomorrow.

Most of them, it turns out, are dead. Old exports nobody opens. Delete those and enjoy yourself.

The rest are holding the place up. Somebody rebuilds them by hand every month. Somebody's job stops without them. One of them is called some version of DO NOT DELETE.

Those are your answer. Each one marks a spot where your systems don't join up, annotated by the person who's been covering for them. That's your to-do list, and you got it without hiring anybody.

Now take only the sheets that touch money, or a number you'd act on, and price those three ways: the hours, the money that leaks, the trust. That's your real bill. Not ours. Not Gartner's. Yours.

Where the line actually is

Now the part where I talk you out of it, at least some of the time.

If a manual step never touches money and never touches a number you'd make a decision on, leave it alone. Copying a delivery address by hand is a chore, and a chore is allowed to just be a chore. Automating it will cost you more than it saves, and it'll break in March anyway.

If you're one or two people and the retyping comes to a few hours a month, the honest advice is a cheap tool and a good bookkeeper. Both beat me on price and neither needs a consultant. The people telling our Saturday man to just buy the accounting software weren't wrong. They were answering the smaller question.

And often the best move is to own fewer tools rather than connect the ones you've got. Glue five separate apps together and you've got five things that can be wrong and a connection between each pair that can also be wrong. Buy the one system that already does that whole job, and you've got one. A three-person landscaping crew turning over $280k a year runs on two: everything about the work in one, the books in the other, and the two talk. That's the whole setup. He never had to connect his way out of the problem, because he refused to buy his way into it.

The line gets crossed somewhere else. It's crossed when somebody is retyping invoices, prices, or stock counts, because that's where those small slips start landing on things that bill. It's crossed when your team has quietly started keeping their own version of the numbers. And it's crossed the day someone asks a question you ought to be able to answer in a minute, like what did we actually make on that job, and the truthful answer is that you'd have to go and reconcile three systems to find out.

Past that line, the hours stop being the point. What you're buying back is the ability to believe your own books.

That's the same foundation everything else sits on, incidentally. It's why the AI you pay for stalls the moment you ask it something that matters, and it's the connected half of what "AI-ready data" actually means. You can't build anything clever on numbers you don't trust. Nobody can.

So go and run the audit. It takes an afternoon and it costs nothing, and you'll come out of it holding a real number instead of a borrowed one.

If that number turns out to be small, I've just saved you a consulting bill. If it turns out to be the size most owners find, start a no-pressure conversation and we'll go through those sheets with you and tell you which single connection is worth making first. Even if the answer is that you should go and hire a bookkeeper instead.

Practical AI. Clear process. Real business value.

— Brian, NuWay Biz Solutions

P.S. If you only do one thing: open last month's invoices and find a job where the work and the bill came from two different places. Then check whether everything you actually did made it onto that bill. Most owners find their answer on the first try, and they don't enjoy it.

Why your AI stalled out — and it's not the AI's fault

Brian Pawl — Thu, 09 Jul 2026 04:13:45 +0000

Originally published on the NuWay Biz Solutions blog.

✦ Cover image: Made with Google Gemini — we're transparent about AI. See the exact prompt on the original post.

The first week with a new AI tool feels like magic. It drafts the email, summarizes the long thread, answers the quick question in seconds. You think: finally.

Then you ask it something that actually matters. Which of our jobs last quarter made money? Which customers are about to leave? And the magic stops cold. It stalls, it hedges, or it hands you a number that's confidently, provably wrong.

So you assume you bought the wrong tool, or that you're using it wrong. Usually it's neither.

Your AI isn't broken. It ran out of data it could trust.

So why did the magic stop?

Because the easy wins and the hard questions pull from completely different places.

Drafting an email needs nothing but the prompt in front of it. Telling you which jobs were profitable needs your sales numbers, your labor costs, your invoices, and what actually got paid. On most teams, those four things live in four systems that quietly disagree.

MIT 2025: 95% of corporate AI pilots showed no measurable P&L impact

This is the part most of the coverage gets wrong. A 2025 MIT study found that 95% of corporate AI pilots delivered no measurable impact on the bottom line. The headline blamed the AI. Read the details and a quieter story shows up: the pilots that stalled did so on brittle workflows, poor fit with how the business actually runs, and data the tools couldn't rely on. Data is one cause among several, and it's the one that sets the ceiling. How far AI can climb depends on how much of your data it can trust.

What does the ceiling actually look like?

It shows up as small, specific friction you've probably stopped noticing.

An order comes in on the webstore, and someone re-keys it into QuickBooks by hand.
The total in your CRM doesn't match the total on your P&L, and no one can say which is right.
Your customer list exists in three places, in three slightly different versions.
Every month, a report you depend on gets rebuilt from scratch in a spreadsheet.

Each one is survivable on its own. Together they mean there's no single version of the truth for an AI to reason from. Or a new hire. Or you, at eleven at night, trying to answer a question you should be able to answer in a minute.

Barr Moses, who named this problem "data downtime," describes it as "periods of time when your data is partial, erroneous, missing or otherwise inaccurate." The question she uses to make it land is the one every owner already feels: "If this chart is wrong, what other charts are wrong?"

Point an AI at data like that and it does exactly what it's built to do. It gives you a fluent, confident answer from shaky inputs. Garbage in, garbage out — except now the garbage sounds articulate.

✦ Made with ChatGPT — we're transparent about AI. See the exact prompt on the original post.

What does "AI-ready data" even mean?

Strip the jargon and it's three things. Not perfect. Just trustworthy enough for the job in front of you.

Data your AI can't use

Scattered across tools that don't talk to each other
Inconsistent: the same customer spelled three different ways
Nobody's quite sure which number is the real one
Half of it lives in an inbox or a spreadsheet on someone's laptop

AI-ready data

Connected: your tools feed one place automatically
Clean: one customer, one record, one spelling
Trustworthy: a single source of truth people agree on
Reachable: the AI can actually get to it, not hunt for it

That's the hurdle. It's boring. It's also the difference between AI that saves you a day a week and AI that embarrasses you in a meeting. If you want each of those four in plain English, with a ten-second test for your own data, here's the full breakdown.

You don't need a giant data project to start

Here's where a lot of consultants would tell you to stop everything and build a data warehouse. Ignore that.

Modern AI can work with data spread across the tools you already use. You do not have to move everything into one place before you get value, and waiting on a big infrastructure project is its own way of falling behind. The owner who ships one genuinely useful thing this afternoon is right to.

So start where your data is already good enough. Pick the one workflow with one clean system behind it, point AI at that, and let it earn its keep. The bigger foundation work earns its place later, the moment the ceiling starts capping what you can trust. For a growing business running five or ten systems, that moment tends to come fast.

Can't I just wire it together myself?

You can try, and plenty of owners do, with a weekend and a chain of Zapier zaps.

It holds until a tool changes its export format. Then it breaks quietly, and you find out when a number is wrong in front of a client or a lender. Connecting two apps is easy. Keeping them connected, and keeping the data honest as both sides change underneath you, is a different job: the unglamorous, ongoing kind.

That same MIT study found something worth sitting with. AI projects built by bringing in an outside specialist succeeded about 67% of the time. The ones built in-house succeeded only a third as often. The boring foundation is exactly the kind of work that goes better with someone who has built it before.

And an AI agent bolted on top of messy data doesn't rescue you. It automates being confidently wrong, faster.

How do I know if I've hit the ceiling?

Run an honest check on the last month.

[ ] You couldn't answer a basic question about the business without opening three different apps
[ ] The same customer or job exists in two systems with two different sets of numbers
[ ] A report you rely on gets rebuilt by hand in a spreadsheet every month
[ ] Someone asked "is this number actually right?" and nobody was sure
[ ] You tried an AI tool for something that mattered and quietly gave up on it

Two or more, and you've hit the ceiling: the useful AI you want is capped by data it can't trust. Skip the two-year overhaul. Pick the single workflow where the payoff is highest, make that data trustworthy, and let AI compound on it. That's the cheapest, highest-return place to start.

Here's the honest version of the whole thing. AI is a multiplier, nothing more and nothing less. Point it at data you trust and it compounds. Point it at a mess and it multiplies the mess.

The businesses pulling real, growing value out of it have one thing in common: they gave the AI something solid to stand on, one workflow at a time. Start where your data is already good enough. Fix the foundation where it counts. Let the AI compound from there.

If you want to know where your own ceiling is, start a no-pressure conversation. We'll find the one workflow worth starting with and the one data gap worth fixing first, whether or not you ever hire us.

Practical AI. Clear process. Real business value.

— Brian, NuWay Biz Solutions

P.S. Yes, we're an AI company, and we just spent a whole article telling you AI isn't your first problem. We know how that sounds. It's also why you'd want us on it: the value in this work lives in the boring foundation underneath the demo, and boring, done right, is the part that actually pays off.

How to actually use AI in your creative work (without becoming slop)

Brian Pawl — Sun, 05 Jul 2026 12:42:22 +0000

Originally published on the NuWay Biz Solutions blog.

✦ Cover image: Made with ChatGPT — we're transparent about AI. See the exact prompt on the original post.

Let's start with the part nobody selling you an AI tool wants to sit in.

If you make things for a living, the last two years have probably included some version of this. You post something you drew or designed by hand, the way you've worked for a decade, and a stranger assumes a machine made it. A client forwards a logo their nephew generated for free and asks if you can just clean it up. Work that looks like yours, except hollow, fills every feed you scroll.

Designers are saying it out loud everywhere: sixteen years of craft, and now people see it and think it came out of a prompt. That one stings in a specific way, because being mistaken for the thing flooding the internet is its own small grief.

There's an official name for the flood now. Merriam-Webster made it the 2025 word of the year: slop.

And the fear underneath all of it is quieter and worse: that if you start using these tools in your own work, you'll turn into the thing you can't stand.

So this one is for the designer who actually wants to use AI in the craft, not just the back office, and is scared of exactly that. The manifesto in this series said to keep AI off the art. This is the companion piece. Here's how to let it into the work without losing the part that's yours.

The line isn't where AI touches the work. It's whether your judgment is driving.

Slop isn't a property of the tool. You can make slop with a pen. What makes the AI version of it so abundant is that the tool will happily supply the one thing it was never yours to outsource: the judgment about what's good.

Tanya Donska, writing in Built In about the AI design slop era, put the failure mode exactly: AI slop "resembles everything and argues for nothing." It's the average of every design ever fed to the model, rendered competently enough to clear any review it'll ever face. Competent. Positionless.

The thing that keeps your work on the other side of that line is taste, and taste is also the thing AI can't hand you. As designer Elizabeth Goodspeed wrote, "AI itself can't generate good taste for you." Brian Collins says the same thing from the other direction: the part a machine will never have is judgment, "the ability to know what not to make."

The art was never in the making. It's in the deciding. That's the part to keep both hands on.

Here's the catch the upbeat takes skip: taste is expensive. It's built over years of looking, copying, failing, and looking again. Which means the real danger runs deeper than one bad final edit. These tools let you ship work above the taste you've actually built, then quietly stop you from building any more. Practitioners have a name for it: the taste-skill discrepancy. The tools raise the floor, not your ceiling.

So keeping your judgment in the driver's seat means three places where your hands stay on the wheel: the front, the middle, and the end.

Keep your hands on the front: decide what you're making before the machine offers

The most common way good designers slide into slop is upstream of any prompt. They open the tool first, before they've decided anything, and let its first plausible suggestion become the brief.

Start in the analog. A sketch, a sentence, a scribbled point of view: set the logic of the thing before the machine gets a vote. The intent is the part a client is actually paying you for, and it's the one input AI genuinely cannot originate. AI that finishes a thought you brought is an amplifier. AI that supplies the thought is a replacement, and you'll feel the difference in how little the result is yours.

The illustrator Gary Hanna is a useful model here. He trains models on his own sketches and prompts them with almost nothing, because the point isn't to borrow a look off the internet. As he puts it, he treats his sketches like a scene he built and AI like the renderer. The vision is upstream, in his hand. The machine just does the rendering he aimed it at.

✦ Made with ChatGPT — we're transparent about AI. See the exact prompt on the original post.

Keep your hands in the middle: don't outsource the muscle that makes you good

This is the one the all-upside crowd never mentions, and it's the one that actually erodes a career.

Skill comes from friction. The hours of doing the hard part badly, then less badly, are how the eye gets built. Hand every hard part to the tool the moment it's hard and the friction disappears, and so, eventually, does the eye. You don't notice it going. You just slowly lose the ability to tell your own good from your own slop.

The working rule that protects you: let AI do the parts you've already mastered, not the parts that are still teaching you. The roto you could do in your sleep, the fortieth export resize, the denoise on a noisy plate: hand those over and buy back your hours. The thing you're still learning to see, the composition you can't quite land yet, the type pairing that's almost right: stay in that. That's not grunt work. That's the gym.

Keep your hands on the end: the art is in the choosing

If AI is good at anything in the craft, it's volume. Twenty directions in the time you'd have sweated out one. That's a real gift, and it's also where most people stop, ship the first decent one, and call it done. Stopping there is the whole difference.

Justin Ahrens, writing in PRINT, framed the discipline cleanly: "Let AI widen the field. You narrow it." The widening is cheap now. The narrowing is the job.

The selection is not a clerical step after the creative work. It is the creative work. Which one, why that one, what to fix in it, what to throw out, what to put your name on: that's taste applied, and it's the thing nobody can prompt their way into. Generate broadly. Then cut ruthlessly, refine by hand, and own what's left.

Where AI actually earns its place in the craft

Plenty of working designers use AI in the actual work and don't make slop, because they aim it at assistance, not authorship. A few that hold up:

The graphics team on a network late-night show uses Runway to handle rotoscoping, the kind of tedious masking that used to eat hours. One artist did in five minutes a shot what used to take five. Nothing about the creative call changed. The drudgery did.

Designer Paul Hatton uses AI to upscale for print, kill noise, and recover detail in soft areas, while keeping, in his words, the reins. Product designer Nikita Samutin treats it as a "visual sparring partner": a source of alternatives to react to, not a source of finished screens. Even Aaron Draplin, no soft touch on this stuff, shrugged that when a friend needed vertical art made square and AI built the missing edges, "that's good".

The pattern across all of them: AI on the mechanical, the exploratory, the parts that were never the point. The judgment stays on the person. Reference, not output.

And where it reliably turns into slop

The other side is just as consistent. Slop is what you get when raw output becomes the finished bar.

The prompt-generated logo is the clearest case. As one creative director put it, generative tools are "very efficient copying machines", assembling from everything that came before without a thought behind any of it, which is also why AI logos drift toward looking like every other logo in the category. Untouched AI copy reads grammatically perfect and completely forgettable. The generic AI hero image is visual elevator music. None of it is wrong, exactly. It just resembles everything and argues for nothing, and your audience clocks the hollowness even when they can't name it.

The test is simple. If you couldn't defend each choice in the piece to another designer, you didn't make those choices. The model did, and you shipped its average.

The part nobody mentions: you can't sell what you don't own

There's a hard business reason to keep your hands on the work, not just a soul reason. In the US, purely AI-generated output isn't copyrightable, and a prompt alone doesn't make you the author. The Copyright Office and the Thaler ruling are both clear: human authorship is the bedrock requirement.

Work where a human makes real creative selection, arrangement, and modification can be registered, but only the human-authored parts are protected. So a logo conjured from a prompt may be something neither you nor your client can actually own or defend, the opposite of what a client is paying for. Your hand in the work makes it good. It also makes it ownable.

The reason any of this matters more now, not less, is that the hollow stuff made the real thing scarce. Companies are already hiring people to fix the slop they generated; one marketplace reported demand for that kind of cleanup jumping 250% in six months. The taste you've spent years building is the appreciating asset. Don't trade it for speed.

And if the reason you never get the hours to keep building it is that your week is buried in intake, proposals, and invoice-chasing, that's the half we'd actually touch. We automate the business around the work, on purpose, so the craft stays yours and you get more time inside it, not less. If that's the trade you want, start a no-pressure conversation, or read what to automate first.

Practical AI. Clear process. Real business value.

— Brian, NuWay Biz Solutions

P.S. A test, if you want one. Take the last thing you made with AI in the loop and try to defend every choice in it to another designer, out loud. The parts you can defend are yours. The parts where you hear yourself say "that's just what it gave me" are the parts to go back and make yours. That's the whole method.

The proposal that was already lost

Brian Pawl — Sat, 04 Jul 2026 04:25:34 +0000

Originally published on the NuWay Biz Solutions blog.

✦ Cover image: Made with ChatGPT — we're transparent about AI. See the exact prompt on the original post.

You spent the better part of a day on it.

A real proposal. The brand work scoped out properly, a timeline that made sense, a couple of options, a fair number you talked yourself into not lowering. It looked like something a serious studio sends. You read it twice and hit send a little proud of it.

Then nothing. A day. Three days. You start refreshing your inbox like it owes you money.

When the reply finally comes, it's kind, and it's short. “Thanks so much for putting this together, it looks great! We've decided to go a more budget-friendly route for now, we're going to try a few things in-house first. We'll absolutely keep you in mind.” In-house. You know what that means this year.

So you do the thing everyone does. You assume the number was too high. Next time you'll trim it, or you'll write three more paragraphs explaining why it's worth it. You file it under priced myself out and move on, quietly a little smaller than you were that morning.

The price was never read on its own

Here's what actually happened to that proposal. It got opened, alone, on a screen, days after you last spoke, by someone half-deciding between you and two other tabs. And on the whole document, the one concrete, comparable thing was the price.

Everything else, the craft, the thinking, the outcome, was words. Words are easy to skim and impossible to weigh. The number is right there, exact, and it has a tab open next to it with a smaller number on it. So that is the only contest that happens. Not your work against their other option. Just your number against a cheaper one, and lately, against a twenty-dollar subscription that promises ninety percent of the thing for five percent of the cost.

The proposal didn't lose on price. It lost because a number showed up before the value did.

A proposal that opens with a deliverables list and a fee is, functionally, a menu. And nobody reads a menu and thinks about value. They read it and they price-shop, because that is the only thing a menu is built to let you do.

The sale doesn't happen in the document

The people who have spent careers pricing creative work, Blair Enns, Jonathan Stark, the ones who wrote the actual books on this, all say a version of the same blunt thing: the proposal should never be the first time money comes up. By the time you are writing one, the deal is mostly already decided. The proposal is the receipt, not the pitch.

Which means the part you have been treating as paperwork, the conversation before the proposal, is where the whole thing is won or lost. That call is not a formality on the way to sending a PDF. It is the sale.

On it, you do three things a document cannot do for you. You get them to say out loud what the outcome is actually worth to their business, and what it costs them if it goes wrong, the rebrand that confuses their customers, the site that quietly leaks every third lead. You find out what they can actually spend, before you have written a single line for free. And you put a number in front of them while you are still in the room to frame it, watch their face, and answer the flinch.

The only order that works

Value, then price, then scope. In that order, every time.

Establish what the outcome is worth before you say a number. Say the number while you can still frame it. Pin down exactly what's in and what's out last, once they've already said yes to the value. Do it in any other order and you have handed someone a menu to shop.

✦ Made with ChatGPT — we're transparent about AI. See the exact prompt on the original post.

So what is the proposal even for?

It documents the thing you already agreed to. Done right, it opens with their problem in their words, not your service list, so the first thing they read is proof you understood the stakes. The number sits where it belongs, after the value, framed by it, never naked on the page.

Give them options if it fits, usually three. The expensive one is not there to be picked. It is there to make the middle one look like the reasonable adult in the room. Some studios do better with a single, confident price and no menu at all. The count is not the point. The point is that the number never shows up without the value standing right next to it, holding its hand.

A good proposal can't win what the conversation didn't. It can only keep from losing it.

Where AI fits, and where it doesn't

Because the question is coming: can't AI just write these now? Some of it, yes, and the honest answer is worth saying plainly. AI will brand your template, beat the blank page, and turn the notes from your call into a competent first draft in about a minute. If you are still building each proposal from scratch in a blank document, that part is a solved problem, and you should let a tool do it.

What it cannot do is the part that actually closes anything. It cannot run the value conversation, cannot read the room, cannot decide what your work is worth or stand behind the number when the client pushes. It automates the document around the proposal. The judgment inside it, the thing the client is genuinely paying for, stays yours. That is the whole position: hand the paperwork to the machine, keep the part that only a person who has done the work can do.

Some clients were gone before you spoke

One hard truth, because pretending otherwise would make this whole piece a lie. Some of the clients you are losing were never yours to lose. They had already decided, before they ever emailed you, that AI or the cheapest freelancer was good enough, and they were really just collecting a number to confirm it. No proposal on earth reframes that. You will usually never even get the chance to try. They just go quiet.

That is not a proposal problem, and it is not a price problem. It is a positioning problem: somewhere upstream, you became interchangeable with a subscription. The fix for that one lives in a different decision entirely, about being the studio a client cannot simply swap out. A sharper proposal wins back the client who is still deciding. It cannot reach the one who already has.

What this looked like for one studio

We worked with a small studio that was losing a painful share of its proposals and blaming the prices on them. We barely touched the prices. We moved the money conversation onto the discovery call, where it belonged, and rebuilt the proposal to lead with the client's outcome and put the number after it, framed, with their budget already known going in.

The proposals stopped being the place a deal went to die alone in an inbox. They started landing as a confirmation of something already decided out loud, on a call, by people who had already heard the price and not flinched. The fee on the page did not change much. How often a yes came back did.

If your proposals keep coming back as a polite note about budget, the document is not the problem and neither is your number. The conversation that was supposed to happen first never did. Start a no-pressure conversation and we'll fix the order: where the value gets established, where the number goes, and what the proposal is actually allowed to carry. You keep doing the work. The number stops being the first thing they see.

Practical AI. Clear process. Real business value.

— Brian, NuWay Biz Solutions

P.S. That's the front of the business walked end to end now: the “AI can do it free” objection, the line scope creep crosses, the inquiry that goes cold, and the proposal that prices it. The last one in the series puts them in order, what a creative studio should actually automate first. See you there.

Automate your business, never your art

Brian Pawl — Fri, 03 Jul 2026 12:30:09 +0000

Originally published on the NuWay Biz Solutions blog.

✦ Cover image: Made with ChatGPT — we're transparent about AI. See the exact prompt on the original post.

Let's get the awkward part out of the way. We set up AI systems for a living, and we think most of what AI is being sold to creatives right now is a trap.

This has nothing to do with AI being evil, and everything to do with which half of your work everyone keeps pointing it at.

They're aiming at the wrong half.

The trap everyone is selling you

Open any feed and the pitch never changes. AI to make more, faster. Ten logos in ten seconds, a month of posts in an afternoon, 10x your output. At 11 p.m. with three deadlines stacked up, it sounds like mercy.

Here's the catch.

Your output is the exact thing that just got commoditized. The market is already waist-deep in generated work that, as people keep saying, feels like no one made it. Every "10x your designs" tool pours more into that flood, and the flood is what you're bidding against.

It trains your clients, too. Run the work through a generator and the whole market absorbs the lesson: this is a commodity, it should be cheaper, the nephew with a free tool counts as competition. Point AI at your craft and you start underbidding a version of yourself that works for nothing and doesn't care how it turns out.

There's a quieter tax on top of that. The tools all drift toward the same look. Same polish, same safe gradients, same blobby mascots nobody chose. Lean on them long enough and your portfolio starts to resemble everyone else's who leaned on the same tool, and the thing that made your work recognizably yours sands down a little with every piece.

You feel it before you can name it.

So that's the trap, and it's a clean one. The tools sold to rescue creatives mostly speed up the one thing quietly killing their rates.

Draw a line down the middle of your business

So draw one. Right down the middle of what you do.

One side is the craft. The taste, the judgment, the calls only you would make, the reason a client picked you and couldn't say why. That is your moat.

Keep AI's hands off it. It's the only thing you own that doesn't turn into a commodity the second a machine can do everything else.

The other side is the business. Intake emails. Proposals. Scheduling. The follow-ups you keep meaning to send. Invoices. The contract you've rewritten by hand a hundred times.

None of that is why you became a designer. All of it can run on its own this week, without anyone laying a finger on the actual work.

Here's why the craft is the real moat: you can't write it down. Hand a machine a perfect, detailed brief and the client still won't get what they wanted, because what they wanted lived in a hundred tiny judgment calls they could never put into words and you make without thinking. That gap is your whole business.

Automating the thing that lives in the gap is the one move that makes no sense at all.

The craft (keep AI's hands off it)

The creative judgment and taste
The final call on what ships
The directions only you would think of
The thing clients can't name but pay for

The business (automate it without guilt)

Intake, qualifying, and first replies
Proposals and the cost-of-getting-it-wrong framing
Scheduling, reminders, and follow-ups
Invoicing and the hundredth contract rewrite

Automate the business. Never the art. That's the whole strategy.

✦ Made with ChatGPT — we're transparent about AI. See the exact prompt on the original post.

This is the version that makes money

In case the high road wasn't motivation enough on its own: this is also the version that makes more of it.

The craft stays scarce, so it stays worth paying for. The clients who matter hire you because the work isn't machine-made, and they can tell faster than you'd think. Meanwhile the admin you finally automated was never billable anyway.

You're cutting the hours you give away and guarding the hours you charge for.

Run the math on your own week. The time you lost to intake and scheduling and chasing invoices last month was unpaid, all of it. The time you spent designing was the part that earned. Buying back the first to make room for the second is the best trade in your business, and it costs you nothing of what makes the work yours.

The market is already turning your way

As the feeds fill up with generated everything, "made by an actual person" is quietly becoming a selling point, the way "handmade" did for furniture once the factories got good. For the clients who can tell the difference, the premium on obviously-human work is climbing, not falling.

A four-person brand studio came to us last year half-convinced they had to start running production through generators just to survive on price. We talked them out of it. We didn't touch the design. We automated the parts they hated anyway, the intake, the proposals, the chase-the-invoice emails, and they kept doing the work by hand and kept charging for it. A year on, the thing keeping them booked is the exact craft they almost handed to a machine.

Look at who's actually struggling right now.

It's the ones who automated the art and still do the admin by hand. Run it the other way around.

What using AI should actually look like

We're the worst possible people to tell you never to touch AI, so we won't. Use it. Use a lot of it.

Just stay honest about which side of the line each use is on, because one side compounds your value and the other quietly sells it off.

On the craft side, if you reach for it, treat it like a reference photo or a rough thumbnail. A private sketchpad nobody else ever sees. The minute AI output leaves the building with your name on it, you've handed the client their reason to stop paying you.

Own the final mile. Sign only what's yours.

There's a real gap between starting from a generated sketch you redraw by hand and sending a generated file you nudged twice. The first is a tool doing tool things. The second gets your contract cancelled the week the client works out they can nudge it themselves.

Whatever carries your name has to carry your hand.

On the business side, let it rip. The studios climbing through all of this are the ones who guard the work like a vault and let the back office run itself.

Where the line actually falls

The line sounds clean until you're standing on it, because most of the hard calls live in the gray. So here's the test we hand people. Would you be comfortable telling the client this part was AI? Easy yes, it's the business side, automate it. If it makes you wince, it's craft, and it stays your hands.

A few that come up on every project. Moodboards and reference-gathering: business side, automate freely, nobody is paying for your Pinterest habit. Cutting one approved design into fifty social sizes: pure mechanics, let the machine have it. The rough first pass a junior used to do: careful, because the second it starts shaping the final direction it has crossed into the craft. The concepts you present as your thinking: never, not once, not even the one you were sure nobody would clock.

The wince is a good instrument. Trust it.

One more thing, because we're not purists about this.

If you're a creative who genuinely wants to push AI further into how you work, not just your back office, we're not going to talk you out of it. That's your call to make, not ours. Come talk to us and we'll help you build something that fits the way you actually work. The one rule we'd hold to is the one that protects you: your name only ever goes on what's yours.

That line is where we work, and the only place we work. We build the systems on the business side, the intake and the follow-up and the proposals and the invoicing, so the admin stops eating your nights. We never touch the art. That's the whole point.

If you want to see where the business side of your studio is bleeding time, and which single system to fix first, start a no-pressure conversation. The teardown is free, and we will never ask to see your sketchbook.

Practical AI. Clear process. Real business value.

— Brian, NuWay Biz Solutions

P.S. Next time someone sells you AI that "designs for you," ask them who's going to pay a premium for it. Then go automate your invoicing instead.