By Aryeh Robinson, Founder, Cognito AI Solutions
How to work out what AI workflow automation would actually pay back at your business — the calculation, the candidate processes, and why you should distrust the percentages you'll find elsewhere.
"We don't have time to figure out AI" is something I hear a lot from small business owners about automation — and it's a reasonable objection. Between customer work, phone calls, and running the business itself, evaluating new technology feels like one more thing competing for hours you don't have.
But that objection usually assumes automation is a long-term bet with a fuzzy payoff. For the kind of repetitive back-office work most small businesses deal with, it doesn't have to be. You can work out the payback yourself, in about ten minutes, before you talk to anybody. Here's how.
"ROI" gets thrown around vaguely in a lot of AI marketing. The more useful question for a small business owner is simpler: how many weeks until this pays for itself, and what does it save every month after that?
That reframing matters because it turns automation from a leap of faith into a calculation you can check — implementation cost, hours saved per week, and a break-even date. Three numbers. Two of them are already sitting in your own business.
Search for this and you'll find plenty of them. First-year returns of 200 to 400 percent. Breakeven in sixty days. Reported returns as high as a thousand percent.
I went looking for where those came from. The trail runs from one vendor's blog post to another vendor's blog post and then stops. I couldn't find the study underneath. That doesn't prove the numbers are wrong — but a number you can't trace is a number you can't use, and I'm not going to hand you one and let you make a spending decision on it.
This is the same habit I teach in every session, and it applies to me too: if a figure can't be sourced, it doesn't get repeated. So here's the number I'll give you instead. Yours.
Pick the one process that eats the most time. Then work out four things.
1. What the task costs you now. Hours it takes per week, times what an hour of that person's time actually costs you — salary, plus payroll taxes and benefits, divided by working hours. Not their billing rate. What they cost.
Say it's ten hours a week at a loaded cost of forty dollars an hour. That's $400 a week, or about $20,800 a year, spent on one repetitive task.
2. What's left after automation. Be honest here. It's rarely zero. Something still needs reviewing, and exceptions still get handled by a person. If ten hours becomes two, you saved eight, not ten. Use the pessimistic figure — if the math works at eight hours it will work at nine.
Eight hours a week at forty dollars is $320 a week.
3. What it costs to build and to run. A one-time build cost, plus whatever it costs per month to keep running — software, API usage, hosting.
4. Weeks to breakeven. Build cost divided by weekly saving. If the build is $6,000 and you're saving $320 a week, that's about nineteen weeks. Then subtract the monthly running cost from the saving to get what it actually returns after that.
Those are arithmetic, not results — plug in your own numbers and you'll get a different answer. The point is that it's an answer you can defend to yourself, which is more than any percentage you'll read in a blog post can give you.
If the breakeven lands past a year, it's the wrong process. Pick another one.
Not every process is a good automation candidate. The best ones share three traits: they're repetitive, they follow a fairly consistent set of rules, and they currently eat a meaningful number of hours per week. For most small businesses, the shortlist usually includes:
If a task on this list is currently done by a person copying information from one system into another, it's very likely a fast-payback candidate. If the bottleneck is less about repetitive tasks and more about a flooded inbox, see our related post on AI email triage for small businesses.
The math above assumes the automation is built correctly — connected to the tools your business already uses, not a separate system your team has to remember to check. That's the difference between automation that runs in production and a project that quietly dies after three months.
A properly scoped automation project starts by mapping the current process step by step, identifying exactly where the manual work happens, and then replacing only those steps — leaving the rest of your workflow untouched. No rip-and-replace, no new platform for your team to learn from scratch.
You don't need six automation projects to see a return. You need one — the highest-volume, most repetitive process currently eating your team's time — done well. That single project can be enough to prove the model and fund the next one out of the time it saves.
As engagements complete, I'll publish real payback numbers from them here, with the process and the assumptions attached so you can judge whether they transfer to your business. Until there are some, the calculation above is the honest version of this post.
If you're not sure which process on your team is the best candidate, that's the kind of question worth a short conversation rather than months of research.
Bring the process that's currently eating the most hours — we'll walk through the real payback math for your business before you spend a dollar.
Book My Free Call