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
- 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.
- 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.
- 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.
- 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.
- 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.
- 06F. Running cost per month. Subscriptions, usage at your real volume, and upkeep. The cost breakdown lists what belongs here.
- 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.
- 08H. Monthly net = C + D + E − F − G. Payback in months = setup cost ÷ H, provided the value arrives every month you pay.
- 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:
| Line | Where to look |
|---|---|
| A. Hours saved | A stopwatch on five real examples, plus a count from your inbox, call log or task list |
| B. Full hourly cost | Your payroll report, or your accountant; otherwise the wage times about 1.4 |
| D. Errors avoided | Write-offs, redo notes and complaints from the last three months |
| E. Revenue recovered | Your practice-management, POS or job-history reports for average ticket and volume |
| F. Running cost | Each vendor's written quote, priced at your own monthly volume |
| G. Review time | The 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.
| Line | Assumption | Monthly value |
|---|---|---|
| A. Hours saved | Admin spends 40 hours a month chasing documents; the system handles about three-quarters | 30 hours |
| B. Full hourly cost | $28 wage × 1.4 | About $39 an hour |
| C. Labor value | 30 hours × $39; the admin moves to onboarding new clients | $1,170 |
| D. Errors avoided | Two 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 recovered | Not counted: the returns get done either way | $0 |
| G. Review time | 4 hours a month of the admin's time | −$156 |
| Value before running costs | C + 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
- StartFill in lines A to G with your own counts, rounding value down
- StepWork out a year of value, counting only the months the work happens
- DecisionIs that more than a year of the quote's running costs?NoStepDon't buy as quoted: ask for seasonal billing or a narrower scopeYes
- StepPayback = setup ÷ (yearly value − yearly running cost)
- DecisionDoes it pay back within the period you set?NoA personOwner decides: wait for a better quote or pick a different taskYes
- A personOffice manager records two weeks of baseline counts before launch
- DoneBuy, knowing the numbers you'll check at eight weeks
Before and after
Deciding by gut vs. deciding by worksheet
Gut feel
- 01Vendor says practices “typically” see big gains
- 02Hours saved guessed as a round number
- 03Hourly wage used instead of full cost
- 04Review time and upkeep left out
- 05No baseline recorded before launch
ResultA yes or no that nobody can check later
Worksheet
- 01Five recall calls timed and averaged
- 02Full hourly cost used, review time included
- 03Revenue counted only from patients who opted in
- 04Value ceiling compared against each quote
- 05Baseline saved so results can be measured
ResultA decision with its reasons written down
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.