To estimate automation ROI before you buy, multiply the hours a task takes each week by 52 and by the loaded hourly cost of the person doing it, take the share of that work the automation would remove, then subtract what the automation costs to run. Divide the build price by the monthly saving and you have the payback period in months. If the payback is under a year and the task will still exist in two years, the case is usually strong.
That is the whole method. The rest of this post is about getting each number right, because most bad ROI estimates come from one inflated input, not from the math. We read the Bureau of Labor Statistics Employer Costs for Employee Compensation release for June 2026 on October 29, 2026 for the labor cost figures below.
The formula
| Step | Calculation |
|---|---|
| Yearly cost of the task | Hours per week x 52 x loaded hourly cost |
| Yearly saving | Yearly cost x share of work removed |
| Net yearly saving | Yearly saving minus yearly running cost |
| Payback in months | Build price / (net yearly saving / 12) |
Our free tools page runs the first three steps for you in the browser. Nothing you type is sent or saved.
Input 1: hours per week, measured
Do not guess. For one normal week, have the person who does the task keep a tally: each time they do it, how long it took. Include the small parts that do not feel like the task, such as finding the file, checking the other system and fixing the typo afterwards. One honest week beats any estimate from memory.
If volume swings with the season, measure in a busy week and a quiet week and use the average.
Input 2: the loaded hourly cost
Wages are only part of what an hour of staff time costs. The BLS reported that in June 2026, private industry employers spent an average of $46.89 per hour worked on compensation: $32.82 in wages and $14.07 in benefits, so benefits made up 30% of the total. For full-time private industry workers the average was $54.00 an hour.
A practical shortcut: take the person's hourly wage and divide by 0.7 to get a rough loaded cost. A $25 an hour coordinator costs closer to $36 an hour once benefits are counted. Use your own payroll numbers if you have them.
Send the hours and who does the work. We'll reply with one idea worth building first.
Input 3: the share of work removed
This is the input people inflate. An automation rarely removes 100% of a task. Someone still reviews exceptions, answers the odd question and checks the output. In our experience a well-scoped workflow removes most of the routine work but not all of it, so pick a share you would defend to your accountant and write down why.
Our case studies show the kind of change real builds produced. A prior authorization workflow cut each request from 12 minutes to 2 (case study). Intake moved from 20 minutes to 6 minutes per patient (case study). Annual review prep went from 2.5 hours to 20 minutes per client (case study). Those are reductions of roughly 70% to 87%, and each still left a person in the loop. Our results benchmarks page collects the numbers from all 26 studies in one table.
Input 4: running costs
An automation has monthly costs even after it is built:
- Software plans for the automation tool, if one is used.
- AI model usage, which for most small business builds we see runs about $30 to $150 a month, paid directly to the provider.
- Hosting, if the workflow runs on a server.
- Upkeep. Our care plan is $350 a month, and it is optional.
Our post on automation maintenance costs goes through each one.
A worked example
A front-desk coordinator spends 6 hours a week copying intake forms into the practice system. Their wage is $24 an hour.
| Step | Number |
|---|---|
| Loaded hourly cost | $24 / 0.7 = about $34 |
| Yearly cost of the task | 6 x 52 x $34 = $10,608 |
| Share removed | 70% |
| Yearly saving | $7,426 |
| Running cost | $50/mo model and tools = $600/yr |
| Net yearly saving | $6,826 |
| Build price | $2,500 |
| Payback | $2,500 / ($6,826 / 12) = about 4.4 months |
These are sample numbers to show the method. Your hours, wage and build price decide your answer.
Count the money the task loses, too
Time saved is the easy half. Some tasks also lose revenue when they are slow or missed: a missed call from a new patient, a lead that waits an hour for a reply, a detention charge nobody disputed. When a task touches revenue, add a second line for recovered income, and be conservative about it.
Two of our posts work this through by industry: what missed calls cost a dental practice and what a slow lead reply costs a real estate team.
Four ways ROI estimates go wrong
- 1Counting hours that will not be freed. If the saved time is spread as five minutes across ten people, nobody gets a usable block of time back. Savings are real when they add up to a role's worth of hours, a hire you avoid or work you finally get to.
- 2Ignoring exceptions. The 10% of odd cases often take 40% of the time. Ask what happens to them after the build.
- 3Forgetting running costs. A $40 a month plan is $480 a year. It matters on small savings.
- 4Automating a broken process. If the steps are wrong today, an automation makes the wrong steps faster. Our post on the most common first automation mistake covers this.
When the numbers say no
If a task takes under an hour a week, happens a few times a year, or is about to change, the payback stretches past two years and the build is rarely worth it. A setting in a tool you already own, or a simple template, is often the better fix. We say so in the audit when that is the answer.
How this fits implementation work
Every OrchestriAI build has a fixed quote before work starts, so the build price in your ROI math does not move. Workflows start at $2,500 and integrations at $2,000; the full list is on the pricing page, and the automation cost guide shows what pushes a price up or down. Fixes in the first 30 days after launch are included.
If you want a second pair of eyes on your numbers, send the task and your volumes through the free automation audit. We reply within one business day with one specific automation idea. For workflow automation and systems integration, the audit is the normal first step.
Independence and limitations
The labor figures are national BLS averages for June 2026 and will differ by role, region and company. The case study numbers are results from specific builds and are not a forecast for yours. Treat the method as a way to test a decision, and check the inputs with your own payroll and volumes.
