Most automation ROI calculations are built to produce a yes. They count the hours saved, ignore what the system costs to run, and assume the automation removes 100% of the work. All three are wrong in the same direction.
Here is the honest version. Our ROI calculator implements exactly this if you would rather not do it by hand.
The formula
annual hours = hours per week x people doing it x 52
annual cost = annual hours x loaded hourly cost
hours saved = annual hours x % the automation actually removes
gross saving = hours saved x loaded hourly cost
net ongoing = gross saving - annual running cost
payback months = build fee / (net ongoing / 12)Nothing clever. The value is entirely in using honest inputs.
Adjustment 1: use loaded cost, not salary
The cost of an hour of someone's time is not their wage. Add payroll taxes, benefits, paid time off, equipment and software. Loaded cost is typically 1.25–1.4× the raw wage.
A $50,000 salary is roughly $31–$35 an hour loaded, not $24. Using the raw wage understates the case for automation — this adjustment works in your favour.
Adjustment 2: never assume 100% removal
This is where most vendor calculations become fiction. No automation removes an entire task. There are exceptions to handle, approvals to click, and cases the system escalates.
Well-scoped automations remove 60–85% of a task's time. Use 70% unless you have a specific reason not to. If a vendor's model assumes 90%+, ask them to show you a deployment where that held.
Adjustment 3: subtract the running costs
The line most calculators omit, and the one that decides whether your finance team believes the number.
- AI model usage — scales with volume. Typically $20–$300 a month for SMB automations.
- Platform subscriptions — Zapier, Make, or hosting for self-hosted tools.
- Maintenance — APIs change and things break. Budget for it even if it is not a line item.
- Human review time — approval gates are a feature, but someone clicks them. A few minutes per transaction adds up.
The question that decides whether any of this is real
Where do the saved hours go?
If those people now do higher-value work, the return is real but shows up as capacity and faster turnaround rather than a smaller payroll line. If they were about to hire and now do not, the saving is a genuine avoided cost. If the hours just evaporate into a slightly less busy week, the financial return is close to zero however good the automation is.
Be honest about which one applies. It changes how you should present the case internally, and it changes whether you should do it at all.
A worked example
Two people spend 6 hours a week each on manual order entry. Loaded cost $32/hour.
- Annual hours: 6 × 2 × 52 = 624 hours
- Annual cost: 624 × $32 = $19,968
- Hours saved at 70%: 437 hours
- Gross saving: 437 × $32 = $13,977
- Running costs: $1,440 a year
- Net ongoing: $12,537
- Build fee: $1,499
- Payback: 1,499 / (12,537/12) = about 1.4 months
That is a clear yes. Now halve the hours to 3 a week each and the payback roughly doubles — still good. Halve them again and it starts becoming marginal. Running the numbers is what tells you which situation you are in.
When the answer is no
If payback is beyond about two years, do not do it. Not because the automation would not work, but because something else in your business has a better claim on the money — and the process may change before the build pays for itself.
Find the task eating the most time instead. The businesses that get the most from automation start with their worst process, not their most interesting one.
Frequently asked questions
What payback period is good?
Under 6 months is clearly worth doing. Six to twelve months is a comfortable yes for most businesses. Twelve to twenty-four is borderline and usually tips over when a second process is bundled into the same build. Beyond that, wait.
Should I count error reduction?
Only if you can put a number on it. "Fewer mistakes" is real but unquantified value that makes a business case weaker, not stronger, because it invites scepticism. If a specific error costs you a specific amount at a known frequency, count it. Otherwise mention it as an unpriced benefit.
What about faster turnaround?
Same rule. If faster invoicing measurably improves cash flow, or faster quoting measurably wins more deals, put a number on it. If not, list it separately from the arithmetic.
How accurate is the calculator?
It is as accurate as your inputs, and it is deliberately conservative — 70% removal by default, running costs subtracted. It sizes the build tier from weekly hours, which is a rough proxy; real scope depends on how many systems are involved and how messy the inputs are.
Run your own numbers. The automation ROI calculator is free, runs in your browser, and will tell you when a process is not worth automating. Or message us on WhatsApp.
