Skip to content

Will It Pay Off? A Simple ROI Worksheet for AI Automation

By USA AI Corp · · 6 min read

The short answer

To calculate the ROI of AI automation, add up what it gives back (staff hours saved at their full hourly cost, errors avoided, revenue recovered), subtract what it costs to run, including your team's review time, and divide that net gain by what you spent, setup included. Count value only in the months the work happens, and costs in every month you're billed. Then measure after launch and redo the math.

The sticky note on the office manager's monitor says “AI recall thing: worth it?” In this hypothetical three-chair dental practice, the dentist who owns it has a vendor quote, a hunch that the overdue-patient list is a mess, and no time to think about either. The office manager doesn't need a financial model to answer. She needs one page with eight lines on it.

The whole calculation in three lines

Monthly value minus monthly running cost equals monthly net. Setup cost divided by monthly net equals the months until the project pays for itself. First-year ROI is the year's net gain divided by what you spent on it. Everything below is about filling in those lines honestly.

The worksheet, line by line

  1. 01A. Hours saved per month. Time five real examples of the task and average them. Multiply by how many times it happens in a month, then by the share the system will genuinely handle. Count routine cases only; exceptions stay with people.
  2. 02B. Full hourly cost of the person doing it. Wage plus payroll taxes and benefits. If you don't have the figure, the Bureau of Labor Statistics' June 2026 compensation survey put benefits at 30.0 percent of private-industry employer costs, so roughly 1.4 times the hourly wage is a reasonable starting point. That's an all-industry average; your own payroll figure is better.
  3. 03C. Labor value = A × B. Count it in full only if the freed hours go somewhere useful. If they'd simply become slack, count part of it, or none.
  4. 04D. Errors avoided per month. How many mistakes the task produces now (wrong entries, missed follow-ups, double-bookings) multiplied by what each typically takes to fix.
  5. 05E. Revenue recovered per month. Work you lose today that the system would win back: appointments rebooked, invoices paid sooner, calls that used to hit voicemail. Use a cautious rate and your own average ticket.
  6. 06F. Running cost per month. Subscriptions, usage at your real volume, and upkeep. The cost breakdown lists what belongs here.
  7. 07G. Review time per month. Hours your team will spend checking the system's work, at their full hourly cost. Expect more in the first months.
  8. 08H. Monthly net = C + D + E − F − G. Payback in months = setup cost ÷ H, provided the value arrives every month you pay.
  9. 09I. First-year ROI = (12 months of value − 12 months of running cost − setup) ÷ (setup + 12 months of running cost). Value means C + D + E − G. For seasonal work, count value only in the months the work happens, and running cost in every month you're billed.

Most of these numbers already exist somewhere in your business. You just have to know where to look:

Where to find each number on the worksheet
LineWhere to look
A. Hours savedA stopwatch on five real examples, plus a count from your inbox, call log or task list
B. Full hourly costYour payroll report, or your accountant; otherwise the wage times about 1.4
D. Errors avoidedWrite-offs, redo notes and complaints from the last three months
E. Revenue recoveredYour practice-management, POS or job-history reports for average ticket and volume
F. Running costEach vendor's written quote, priced at your own monthly volume
G. Review timeThe vendor's estimate, doubled for the first month until you've measured it

A worked example with made-up numbers

Here's the worksheet for a hypothetical six-person accounting firm weighing automated document requests and reminders for tax season. Every figure is invented for illustration. Swap in your own.

Hypothetical worksheet: document chasing at a six-person accounting firm, January through April
LineAssumptionMonthly value
A. Hours savedAdmin spends 40 hours a month chasing documents; the system handles about three-quarters30 hours
B. Full hourly cost$28 wage × 1.4About $39 an hour
C. Labor value30 hours × $39; the admin moves to onboarding new clients$1,170
D. Errors avoidedTwo returns a month stalled by a missed document, about 2 hours each to untangle at the preparer's full hourly cost of $60$240
E. Revenue recoveredNot counted: the returns get done either way$0
G. Review time4 hours a month of the admin's time−$156
Value before running costsC + D + E − G$1,254

The table stops before line F, and it covers only January through April. About $1,250 a month is what the task is worth in busy season, but a subscription usually bills all year. Over twelve months the task is worth about 4 × $1,254 = $5,016, so a quote billed every month has to stay under roughly $5,016 ÷ 12 ≈ $418 a month just to break even, before the setup fee.

Comparing a year-round fee with the busy-season figure would make the tool look three times more affordable than it is. If a vendor bills only for the months you use it, compare against the $1,254 instead.

To finish the example, say a vendor quotes $1,500 to set up and $250 a month all year (invented numbers, not real prices from us or anyone). First-year cost: $1,500 + 12 × $250 = $4,500, so first-year ROI is ($5,016 − $4,500) ÷ $4,500 ≈ 11 percent. In year two, with no setup fee, it's ($5,016 − $3,000) ÷ $3,000 ≈ 67 percent. The project pays back within its first year, but only just; one slow busy season and it wouldn't.

Working out the ceiling first is the most useful habit on this page. It turns “is this worth it?” into “what is this worth to us?”, a number you can take into any sales conversation.

Notice the empty revenue line, too. For the accounting firm, the gain is time and fewer stalled returns, not new money. For the dental practice with the sticky note, revenue recovered is usually the biggest line: each overdue patient who rebooks a cleaning puts production back on the schedule.

Say 200 overdue patients have agreed to be contacted, one in ten rebooks over two months, and the practice's own average hygiene visit is $150 (all invented). That's 20 visits and about $3,000 of production, or roughly $1,500 a month for those two months. Production isn't profit, though: if the hygiene schedule was already full, a rebooked patient may simply replace someone else.

Flowchart

Turning the worksheet into a yes or no

  1. StartFill in lines A to G with your own counts, rounding value down
  2. StepWork out a year of value, counting only the months the work happens
  3. DecisionIs that more than a year of the quote's running costs?
    No
    StepDon't buy as quoted: ask for seasonal billing or a narrower scope
    Yes
  4. StepPayback = setup ÷ (yearly value − yearly running cost)
  5. DecisionDoes it pay back within the period you set?
    No
    A personOwner decides: wait for a better quote or pick a different task
    Yes
  6. A personOffice manager records two weeks of baseline counts before launch
  7. DoneBuy, knowing the numbers you'll check at eight weeks
Two checks decide any quote: does a year of value cover a year of cost, and does the setup fee come back fast enough on cautious numbers?

Before and after

Deciding by gut vs. deciding by worksheet

Gut feel

  1. 01Vendor says practices “typically” see big gains
  2. 02Hours saved guessed as a round number
  3. 03Hourly wage used instead of full cost
  4. 04Review time and upkeep left out
  5. 05No baseline recorded before launch

ResultA yes or no that nobody can check later

Worksheet

  1. 01Five recall calls timed and averaged
  2. 02Full hourly cost used, review time included
  3. 03Revenue counted only from patients who opted in
  4. 04Value ceiling compared against each quote
  5. 05Baseline saved so results can be measured

ResultA decision with its reasons written down

The same sticky-note question, answered two ways.

Five ways ROI estimates go wrong

  • Counting every minute as saved. The system takes the routine cases. Exceptions still need a person, and they're the slow ones.
  • Counting hours that go nowhere. Saved time becomes money only if it cuts overtime, avoids a hire or moves to work that earns.
  • Double counting. If a rebooked appointment is counted as revenue, don't also count the front-desk minutes it would have taken as saved.
  • Judging on month one. Early weeks carry extra review and fixes. Base the payback on a settled month, not the first.
  • Ignoring seasons. Document chasing is heavy from January to April and light in July, but a subscription bills every month. Compare a year of value with a year of cost, as the example does.

Measuring after launch

The worksheet is a forecast. To turn it into a record, keep a five-column weekly tally in a spreadsheet: items the system handled, items it passed to a person, errors caught in review, errors that reached a client or patient, and minutes spent reviewing. That takes two minutes a week, and after a quarter you can rerun the worksheet with real figures.

Use the tally on a fixed schedule. Record two weeks of baseline counts before launch. At eight weeks, compare measured monthly value with the forecast; if it's under half, find which worksheet line fell short (hours, errors or revenue) and decide whether to adjust, wait a month or cancel. After that, once a quarter is enough.

If you'd like a second set of eyes on the assumptions before you buy, we're glad to look over your numbers on a first call. And if the worksheet says a project won't pay, that's a good result too: it just saved you the setup fee.

Sources

Common questions

Quick answers.

How do you calculate time saved by automation?

Time five real examples of the task and average them, multiply by how often it happens each month, then by the share of cases the system will fully handle. Subtract the time people will spend reviewing its work. What's left is your net hours saved.

What is a good ROI for AI automation in a small business?

There's no universal number. A practical test: using cautious assumptions, the project should pay back its setup cost within a period you're comfortable with, and still come out ahead in a slow month. If it only works with generous estimates, choose a different first project.

Should I include my own time in the ROI calculation?

Yes, on both sides. Count the owner's hours the system saves, and the owner's hours spent testing, reviewing and managing it. Owner time is often the scarcest resource in a small business, so value it at what an hour of your attention is worth to the business.

How long should I wait before judging whether automation paid off?

Give it about two to three months at normal volume. The first weeks include setup fixes and extra review, so they understate the value. Seasonal businesses should judge over the busy season, when the task actually happens.

Put it to work

AI consulting in New York

AI consulting for small and mid-sized businesses: we map your workflows, rank what's worth automating and hand you a written plan, with us or without us.

See how it works

Most relevant for accounting firms, dentists.

Keep reading

Ready when the work is.

Show us what is slowing your business down.

Book a call

Not ready to talk?

Written brief

Skip the call.
Send a brief.

A few sentences is plenty. We read every brief and reply by email within one business day.

Try it free. If you love it, keep it. If it's not for you, we take it back, no questions asked.

We only use this to reply to you. Privacy policy