AI automation vs. hiring: do the math first
A new hire costs a lot more than the salary. Here's the fully-loaded math — and how to tell whether you're facing a hiring problem or a wiring problem.
5 min readBy Nick Hayghe
"We're drowning. We need to hire someone."
Maybe. Or maybe you're paying a human being $60,000 a year to do clerk work that a workflow could do for the price of a used truck, once.
Here's the reframe: that's not a hiring problem. It's a wiring problem. Before you post the job ad, do the math. You might actually need a person. But you'd be surprised how often you don't.
This post is part of AI automation for small business: the plain-English guide — the full framework this math fits into.
What a hire actually costs
A $45,000 salary is not a $45,000 decision. Let's run the real numbers for an office/admin hire — illustrative, but close to what most 10-person shops see:
| Cost | Per year |
|---|---|
| Salary | $45,000 |
| Payroll taxes + workers' comp | $5,500 |
| Benefits (health, PTO, retirement) | $7,000 |
| Recruiting + onboarding time | $3,000 |
| Equipment, software, desk | $2,000 |
| Management time (yours) | $5,000 |
| Total, fully loaded | ~$67,500 |
And that's the good case — someone who stays, learns fast, and doesn't make expensive mistakes in month two. Add turnover risk: if they leave in year one, you pay most of this twice.
I'm not saying don't hire. I'm saying know what "hire" costs before you compare it to anything.
What the same work costs automated
Take the work that's actually drowning you. Usually it's some version of: answer the leads, book the jobs, chase the paperwork, follow up the quotes. Say it's 25 hours a week of clerk work spread across your team.
An automation that handles that chain — lead capture to qualification to booking to follow-up — is a fixed-price build. Our Automate builds start at $7,500. One time. You own it. Then it runs for tens of dollars a month in AI and message fees.
| Hire | Automation | |
|---|---|---|
| Year 1 cost | ~$67,500 | ~$8,500 (build + run) |
| Year 2+ cost | ~$67,500/yr | ~$500/yr run cost |
| Works at 11pm | No | Yes |
| Forgets follow-ups | Sometimes | Never |
| Quits in month 8 | Possible | No |
| Handles a furious customer | Yes | No — routes to you |
The payback on the build is measured in months, not years. That's not hype; it's just arithmetic with the fully-loaded number instead of the salary number.
That's not a hiring problem. It's a wiring problem.
What "run cost" actually looks like
Owners hear "AI" and picture a meter running. At your size, the meter is a trickle. A lead-response workflow handling a few hundred conversations a month typically costs $20–60/month in AI usage plus a few cents per text message. The CRM and calendar connections are usually tools you already pay for.
So the honest year-two comparison is roughly $67,500 for the hire versus a few hundred dollars for the workflow. That's not a typo. The build is the cost; the running is pocket change. Which is exactly why the build has to be right — there's no subscription revenue smoothing over a bad design, and any builder worth hiring knows their work has to stand on its own after handoff. You own it, it keeps running, and if we ever part ways nothing stops.
When you actually need the human
Honesty is the whole brand here, so: plenty of situations need a person, and automation won't touch them.
- The work is mostly judgment. If the job is reading customers, negotiating, making calls nobody wrote rules for — hire. Automation handles the clerk work around the judgment; it doesn't replace it.
- You're understaffed on skilled labor. No workflow installs a water heater. If you need techs in trucks, you need techs in trucks.
- The process doesn't exist yet. If nobody can describe what should happen when a lead comes in, there's nothing to automate. Hire the person, let them build the process, then automate the repetitive parts.
- Volume is tiny. Ten leads a month? Answer them yourself. The math only works when the task repeats enough to matter.
The test is simple: if the work is repetitive, time-sensitive, and rules-based, it's a wiring problem. If it needs judgment, skill, or a pulse — hire.
The move most owners miss
You don't have to pick one. The best setup is usually: automate the clerk work, then hire (or keep) the person for the work that matters.
Your office manager stops spending her mornings on follow-up emails and spends them on the upset customer, the vendor who shorted you, the estimate that needs a human eye. She doesn't get replaced — she gets promoted to the work she was hired for. That's the version where automation and hiring aren't enemies. It's the version where the $67,500 you spend on a person buys $67,500 worth of human judgment instead of $67,500 worth of copy-paste.
Do the math on your own numbers
Grab a sheet of paper. Write down the task that's drowning you. Hours per week. Fully-loaded hourly cost of whoever does it. Multiply by 50. That's your annual cost of not fixing this — whether you fix it with a hire or with wiring.
If the number's big and the task is repetitive, you already know the answer. If the number's big and the task needs a human, hire well and automate everything around them.
Get a free tool-stack audit
We'll look at your actual workload, run this math with your real numbers, and tell you straight: hire, automate, or both. If automation doesn't pay here, we'll say so — we'd rather lose a sale than build you something useless. Book a free tool-stack audit.
