AI automation ROI calculator
Estimate what automating one document workflow returns: the hours your team gets back, what they are worth, and when the costs on your quote are paid back. Every formula is on this page, and no result is a Mirage promise.
No sign-up. The figures stay in your browser unless you copy the link or download the table.
Your workflow today
Your workflow today
Mark a figure as measured once it is counted, timed or read from a quote. Everything else stays an estimate, and the summary says which is which.
One bid package measured from its drawings.
From opening the drawings to the quantity list for one bid. Time two or three recent bids.
Salary plus charges and overhead, per hour. Your figure: we publish no benchmark.
Used only to express the hours freed as people. 140 is an example.
Costs from your quote
A deployment starts with a proof of concept on your documents one week after the first meeting, then a fixed-price pilot you can stop, then an annual subscription priced by volume. Mirage's prices start at $10,000 to launch and $1,000 a month. Enter the figures from your quote, or a budget you want to test.
Workshops, sample documents, reviews of the pilot. Valued at your hourly cost.
Three scenarios
Example values, to replace with yours. The share is the part of the volume the agent takes from end to end; the rest stays manual. The review time is what a person still spends on each automated unit.
| Three scenarios | Prudent | Central | High |
|---|---|---|---|
| Share handled by the agent (%) | |||
| Review time per automated unit (h) | |||
| Months before the workflow runs | |||
| Months to reach the full share |
Results
Central scenario
- Hours freed per month, at full run
- 64.8 h
- The same hours, in people
- 0.46
- Value of that time in year 1
- Enter your hourly cost
- Payback
- Enter your hourly cost
| Results | Prudent | Central | High |
|---|---|---|---|
| Hours freed per month, at full run | 38.4 h | 64.8 h | 96 h |
| The same hours, in people | 0.27 people | 0.46 people | 0.69 people |
| Hours freed in year 1 | 250 h | 551 h | 912 h |
| Enter your hourly cost | |||
| Value of that time per year, at full run | |||
| Value of that time in year 1 | |||
| Year 1 budget that still breaks evenWhat the pilot, the integration and the first year's subscription can cost together and still be covered by year 1. | |||
| Cash you can bank in year 1 | |||
| Enter your hourly cost | |||
| Year 1 cost | |||
| Net in year 1 | |||
| Net per year from year 2 | |||
| Return in year 1 | |||
| Payback | |||
| Cash net in year 1 | |||
From estimates to measured figures
Most figures above start as estimates. Two of them cannot be known before a pilot on your own documents: the share an agent handles from end to end and the review time it still needs. Measure each figure this way, then mark it as measured.
- Volume
- Count the units in your system over the last three months and divide by three.
- Manual time
- Time ten ordinary units with a stopwatch, from opening the document to the saved record, and take the average.
- Loaded hourly cost
- Ask finance for the loaded rate of the role, not the salary.
- Share handled by the agent
- In the pilot: the units the agent completed with no correction, divided by all the units it received, on real documents over several weeks.
- Review time
- In the pilot: the minutes a person spends checking each automated unit, timed on a sample.
- Months before it runs, and ramp
- From the deployment plan agreed after the pilot.
- Costs
- From the quote, not a budget guess.
Summary for a finance sponsor
An estimate built from the inputs below, not a Mirage quote or promise.
- Workflow:
- Construction takeoff
- Currency:
- USD
Inputs
| Inputs | Value | Unit | Source |
|---|---|---|---|
| Takeoffs per month | 12 | Estimated | |
| Hours of manual takeoff per bid | 12 | h | Estimated |
| Loaded hourly cost | Not entered | Estimated | |
| Productive hours per person per month | 140 | h | Assumption |
| Share of freed time that removes a cost (%) | 0 | % | Assumption |
| One-time cost: pilot and integration | Not entered | Estimated | |
| Annual cost | Not entered | Estimated | |
| Your team's hours during the deployment | Not entered | Estimated |
Scenario assumptions
| Scenario assumptions | Prudent | Central | High | Source |
|---|---|---|---|---|
| Share handled by the agent (%) | 40 | 60 | 80 | Estimated |
| Review time per automated unit (h) | 4 | 3 | 2 | Estimated |
| Months before the workflow runs | 3 | 2 | 2 | Assumption |
| Months to reach the full share | 6 | 4 | 2 | Assumption |
Results by scenario
| Results by scenario | Prudent | Central | High |
|---|---|---|---|
| Hours freed per month, at full run | 38.4 h | 64.8 h | 96 h |
| The same hours, in people | 0.27 people | 0.46 people | 0.69 people |
| Value of that time in year 1 | |||
| Year 1 cost | |||
| Net in year 1 | |||
| Net per year from year 2 | |||
| Return in year 1 | |||
| Payback | |||
| Cash net in year 1 |
Not included
- No hourly cost entered: no amount, return or payback is computed.
- Cash: the share of freed time that removes a cost is 0%, so the value of time is capacity, not cash.
- Fewer errors, faster replies, more bids or orders won: usually the same hours seen twice, so left out.
- Taxes, financing and discounting: the figures are undiscounted and before tax.
- Price changes after year 1 and growth in volume: costs and volume are held flat.
Estimated and measured inputs
- Measured
- None
- Estimated
- Takeoffs per month; Hours of manual takeoff per bid; Loaded hourly cost; One-time cost: pilot and integration; Annual cost; Your team's hours during the deployment; Share handled by the agent (%); Review time per automated unit (h)
How do you calculate the ROI of automation?
- Hours freed per month = volume × share handled by the agent × (manual time − review time), in hours.
- People = hours freed ÷ productive hours per person per month. The same hours, not a second benefit.
- Value per month = hours freed × loaded hourly cost.
- Year 1 counts nothing during the months before the workflow runs, then a straight ramp to the full share. Year 1 value = value per month × the sum of the twelve monthly factors.
- Year 1 cost = one-time cost + your team's hours × hourly cost + annual cost. Net = value − cost. Return in year 1 = net ÷ year 1 cost.
- Payback = the first month in which cumulative value reaches cumulative cost, with the one-time cost and your team's hours at the start and the annual cost spread over the months, searched over five years.
- Cash = value × share of freed time that removes a cost. Cash net = cash − (one-time cost + annual cost).
A worked example, with invented figures
Round figures chosen to show the arithmetic, not a benchmark: 600 customer orders a month, 12 minutes to key one by hand, 2 minutes to review an automated one, 70% of orders handled by the agent, a loaded cost of 40 an hour, 140 productive hours per person, one month before the workflow runs and two months of ramp, and the published starting prices as the quote: 10,000 to launch, 12,000 a year, plus 40 hours of your team's time.
| Hours freed per month, at full run | 600 × 0.70 × (12 − 2) ÷ 60 | 70 h |
|---|---|---|
| The same hours, in people | 70 ÷ 140 | 0.5 people |
| Value of that time per month | 70 × 40 | 2,800 |
| Value per year, at full run | 2,800 × 12 | 33,600 |
| Year 1 cost | 10,000 + 40 × 40 + 12,000 | 23,600 |
| Net per year from year 2 | 33,600 − 12,000 | 21,600 |
| Payback | Cumulative value reaches cumulative cost around month 9 on these figures: nothing in month 1, half value in month 2, full value from month 3, against 11,600 up front and 1,000 a month | Month 9 |
| Cash in year 1 | Share of freed time that removes a cost left at 0% | 0: the hours are capacity until a cost is removed |
What is not counted, on purpose
Fewer keying errors, faster replies, more bids priced or more orders taken are real, but they are usually the same hours seen a second time. Add them only if you can measure them on their own, and never on top of the hours above.
Value of time is not cash
Freed hours are capacity. They become cash only when they remove a cost: overtime, temporary staff, outsourcing, or a hire you do not make. Until you decide which, the honest cash figure is zero, which is why that share starts at 0%.
Published results, for judging a scenario
Each line is one client's result on its own documents, as its case study states it. None is a forecast for yours: use them to judge whether a scenario is plausible, not to fill it.
- John Cockerill: 80 to 90% less time on piping takeoff.
- Biodéal: About 7% of orders flagged for a person to check; a median of 48 seconds from email to order ready in SAP Business One.
- OFRET Groupe: More than 90% of standard quote requests handled automatically.
- Transwin: 70% less manual entry on freight documents.
When automation does not pay
- A handful of units a month: the setup is not repaid.
- Work that is different every time: the share an agent can take stays low.
- A review that takes nearly as long as doing the work: little time is freed.
- Freed time with nowhere to go and no cost to remove: the value stays on paper.
Check these assumptions on your documents
Tell us which workflow you entered. We come back with what an agent can take over, the review it still needs and the figures to put in this calculator. No commitment.
Prefer to talk it through? Book a call
Questions
They are yours. A loaded hourly cost depends on the country, the role and what your company counts in it, and a quote depends on your workflow and volume. The calculator shows the hours freed and their people-equivalent straight away, and every money figure once you enter them.
They are round illustrative values, not benchmarks. Replace them with your own volumes and times. The published results of named clients are listed under the calculator to help you judge whether a scenario is plausible.
Because an error avoided is mostly the time it would have taken to find and fix it, which the hours already count. Adding both would count the same benefit twice. If an error has a cost of its own that you can measure, such as a penalty or a write-off, add it separately and say so.
The value of time is what the freed hours are worth at your hourly cost. Cash is the part of it that removes a cost you actually pay: overtime, temporary staff, outsourcing or a hire you do not make. A team that uses the time to price more bids or take more orders gains capacity, not a cash saving.
Mirage answers any request within two business hours, runs a proof of concept on your documents one week after the first meeting, and puts the first workflow in production in 4 to 6 weeks. A workflow spread over several systems can take longer. The months before the workflow runs, in each scenario, are there to account for it.
Then, on your figures, the workflow does not pay in that scenario. Check the review time and the share first. If they are right, keep the work manual, or start with a workflow that has more volume.
Related
- See construction workflows
- AI takeoff and estimating
- Hire an estimator or automate takeoff? A cost-per-bid example
- TMS data entry automation
- Sales order entry automation
- Sales order exception handling: the review time the calculator asks for
- How a Mirage engagement is priced
- How to build the business case for freight document automation
- All free tools