The ROI calculator

What would AI actually save you?

Pick one repetitive task. Put in a few rough numbers. See the hours and the money AI could hand back, with the math shown and nothing hidden. Your numbers appear right here, no signup needed.

One task at a time. Be specific, it makes the number more honest.

Salary plus overhead, roughly. An estimate is fine.

50% AI does most, human reviews

AI produces the work, a person reviews everything before it counts.

Add running and build cost (optional)

On the napkin

0 hours back / year
$0 value of that time / year
Cost to run the tool$0 / yr
Net value, year one$0

This is a napkin model. The real number depends on the four ingredients, which is what the emailed report walks through for your task.

Email me the full report

The report restates your numbers, then shows what the napkin leaves out (adoption, review time, data readiness) and what a realistic first project on your task looks like.

Show the math (no black box)

Hours back / year = people × hours per week × 46 working weeks × hand-off %

Value / year = hours back × hourly cost

Running cost / year = monthly tool cost × 12

Net value, year one = value − running cost − one-time build

46 weeks, not 52: a conservative allowance for holidays, ramp-up, and the weeks nobody touches the task. Change any input and the numbers update live.

Before you ask

Straight answers on AI ROI.

How do I calculate the ROI of an AI tool?

Estimate the hours a repetitive task takes each week, multiply by the number of people doing it and the working weeks in a year, then by the share you could realistically hand to AI. Multiply those hours by what that time costs, and subtract what the tool costs to run. That gives you a rough annual value. It is a napkin model to decide whether something is worth a closer look, not a full business case, because it leaves out adoption, review time, and data readiness.

What is a realistic first AI project?

A realistic first AI project is small and specific: one repetitive task, with clear inputs and a clear definition of done, that you can prototype in a couple of weeks and try for real. The best candidates are tasks where the information already exists but is slow to pull together, and where a person stays in the loop to check the output. Start there, prove it, then widen.

Why can AI not take over 100% of a task?

Because someone still owns the outcome. AI is a magnifier of expertise, not a replacement for judgment, so a person reviews the output, handles the exceptions, and stays accountable, especially early on. This calculator lets you slide the hand-off all the way to 100% so you can see the number, but it flags anything above 90%: in practice the last stretch is review, edge cases, and accountability, and the strongest systems keep a person in the loop on purpose.

Ran the numbers? Let's pressure-test them.

A free, 20-minute readiness call. Bring the task you costed out. If it's not a fit, you still leave with a clear first project.