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.
Top comments (0)