TL;DR: a $50k hire costs $80-90k in year one. A single-process automation costs $2,000-8,000 to build and $100-500/month to run. But that comparison only holds for repetitive work with consistent inputs. For anything involving judgment, relationships or novel situations, hiring is still correct and automating is an expensive way to find that out.
The public version of this argument is technology against jobs. That is not the question in front of a founder with a bottleneck. The question is narrower and more useful: what does it cost to get this specific work done, and which path stops it from being the thing that caps the business.
I run an automation studio, so you should read the rest of this knowing which side of the table I sit on. I have also told clients to hire instead, and the section on when to do that is the one I would read first if I were you.
What a hire actually costs
Almost every founder undercounts this, and not by a little. The salary is the visible part and it is roughly two thirds of the real number.
Base salary: $40,000-80,000/year depending on role and market
Employer taxes, superannuation or benefits: add 25-35% on top
Recruitment: 15-25% of first-year salary through an agency, or several weeks of your own time doing it yourself
Ramp: two to four months before someone is fully productive, paid in full throughout
Management: three to five hours a week of your time, per direct report, indefinitely
Tools, equipment, desk: $2,000-8,000/year per person
Turnover: average tenure in operational roles runs two to three years, so the recruitment and ramp costs recur
Work it through. A $50,000 salary becomes $65,000-75,000 once taxes and benefits land. Add recruitment and the unproductive ramp period and year one is $80,000-90,000. That is north of $7,000 a month for a mid-level operations hire, and for the first quarter of it you are paying for someone who is still learning where things are.
The cost that never appears on a spreadsheet is your attention. Three to five hours a week of founder time, on a founder whose time is the scarcest input in the company, is not a rounding error. It is often the real reason a small team resists hiring even when the arithmetic says otherwise.
What you get from a person that you do not get from a system
Worth stating plainly before the comparison, because anyone selling you automation has an incentive to skip it.
Judgment on situations nobody anticipated, which is most of what goes wrong
The ability to notice that something is off without being told what to look for
Relationships with customers and suppliers, which compound and do not transfer to software
Capacity that redeploys: the same person can do a different job next quarter
Someone who can tell you the process itself is wrong
A system does exactly what it was built to do, at volume, without complaining, and it does not notice when the world changes underneath it. That is simultaneously its advantage and its limit.
What automation actually costs
A well-built system for a single business process - order processing, lead qualification, first-line support, invoice matching - runs $2,000-8,000 to build and $100-500/month to operate. The build is one-off. The running cost is not, and it has three parts worth separating.
Platform fees: $9-50/month for Make or n8n at small business volumes, predictable and stable
AI API usage: $20-100/month typically, and this one scales with how hard the system works, so success makes it larger
Maintenance: a few hours a quarter for most systems, triggered by an integration changing its API or your own process changing
The honest caveat on the second year: the build cost does not recur, but nothing else stops. Budget the running cost as a permanent line, and budget some hours for the moment a vendor you integrate with changes something without asking you.
The comparison, worked
Put the two next to each other over a realistic horizon rather than in the first month, where the build cost makes automation look worse than it is.
Year one, same volume of work:
Hire: $80,000-90,000, productive from roughly month three
Automation: $5,000-10,000 to build, $1,500-3,600 to run, productive from roughly week four
Year two, hire: $65,000-75,000 and rising
Year two, automation: $1,500-3,600 plus occasional maintenance
Payback on an automation that genuinely absorbs the same volume typically lands at two to four months. The caveat carrying all the weight is 'genuinely absorbs'. If the system handles 70% of cases and a person still reviews the rest, you have not removed the headcount, you have made one person able to carry far more. Which is usually the better outcome anyway, and is the next section.
Do this arithmetic on your own numbers
Take the hours the process eats each month. Multiply by a loaded hourly cost, meaning salary plus the 25-35%, not the salary alone. Compare that against the build spread over twelve months plus the monthly running cost. If it does not clear comfortably inside a year, the problem is not the price of automation, it is that this process is not the bottleneck you thought it was.
Where automation wins, and where it does not
The dividing line is not how hard the work feels to a human. It is how consistent the inputs are and how expensive a wrong answer is.
Wins clearly
Repetitive, rules-based work with consistent inputs, where a mistake is visible and recoverable. Order processing, email classification, data entry, report generation, qualification against known criteria, first-line support on questions that repeat.
Struggles
Novel situations needing judgment, relationship-heavy work, negotiation, and any process where exceptions are the rule rather than the tail. If a person doing the job would say 'it depends' to most questions about it, automating it will cost far more than the quote and satisfy nobody.
Functions we have seen automated well, with the realistic share:
First-line customer support on FAQ-type queries: 60-80% of volume handled without a human
Order processing and fulfilment coordination: near complete, with exceptions escalated
Invoice matching and accounts payable: high, once the document formats are known
Lead qualification and first-touch outreach: high on qualification, human on the close
Content and social scheduling: near complete on mechanics, human on judgment
Internal reporting and data aggregation: complete, and usually the cheapest win available
The answer most businesses actually land on
Not automation instead of hiring. Automation that makes one hire do the work of three or four. The system carries volume, the person carries exceptions, relationships and judgment. You still get a human who can notice things, and you stop paying for a human to retype order details.
Simbago is the clearest version of this we have shipped. It is an online store running on one warehouse worker. Orders arrive from WooCommerce and are processed automatically, supplier reorders fire on stock thresholds, and a chatbot handles the bulk of support. The warehouse worker does physical work only. Without the system that operation needs three to four people, and the difference over a year is six figures.
Note what did not happen there: nobody was replaced. The business grew into a shape that would otherwise have required hiring, and did not have to.
When you should hire and not automate
We turn down projects on these grounds, and I would rather you read it here than find out after paying for a build.
The process runs a handful of times a month. Automating it is a hobby, not an investment.
The process is about to change. Build the automation after the change, not before it.
The process is broken. Automating a broken process gives you a broken process running faster and with less visibility into why.
Nobody in the business can say precisely how the work is done today. That is a sign the knowledge lives in one person's head, and the first job is getting it out, not encoding it.
The work is the relationship. Sales conversations with a handful of large accounts do not want a system in the middle.
The question founders ask last and worry about first
What happens to the person whose work gets automated. In practice, in small businesses, almost always the same thing: they stop doing the part of the job they disliked and take on work that was not getting done at all. Nobody in a ten-person company is short of things that need doing. The cases where automation genuinely removes a role tend to be ones where the role existed only because the process was manual, and those are usually roles the business struggled to fill anyway.
It is still worth saying out loud to the person before the build starts rather than after. We have watched projects go badly for no reason other than that nobody told the operations manager what was being built.
How to decide this month
Pick the process that eats the most hours, write it out step by step, and count how many of those steps have a consistent input and a rule you can state in a sentence. If most of them do, you have an automation candidate and you can get a scoped quote in a day or two. If most of them do not, you have a hiring candidate, and knowing that is worth the hour it took.
Four questions, two minutes, no call: 2pizza.team/audit tells you which part of your process is worth automating first. If the honest answer is hire someone, it will say that instead.
Originally published at 2pizza.team. We build AI and automation systems for small teams - fixed price, two to six weeks. See the work.
Top comments (0)