AI ROI Calculator
Estimate the annual savings, ROI, and payback period of an AI automation — plug in team size, hours, and costs.
A planning estimate: real ROI depends on adoption, quality, and maintenance. Gross annual savings ≈ $39,000; net figures subtract build + running costs.
I help teams scope, build, and ship AI automations that pay back — from architecture to production.
Work with meWhat this tool does
The AI ROI Calculator turns a handful of planning numbers into a first-pass business case for an AI automation. You enter how many people do a task, how many hours a week each spends on it, what an hour of their time costs, how much of that time you expect AI to save, and what the automation costs to build and run. It returns the estimated year-1 net benefit, year-1 ROI, the payback period in months, and the hours saved per year.
It is meant for the early conversation: a team lead, product owner or architect deciding whether an automation is worth scoping before any detailed analysis. It answers one narrow question. If the time savings you assume actually happen, do they outweigh the build and running costs within the first year, and how long until the build cost is recovered?
How it works
The tool values the time saved at your loaded hourly cost, adds up a year of savings, and subtracts the one-time build cost and twelve months of running cost. Everything is a straight multiplication, with no discounting, growth or ramp-up. Negative entries are treated as 0, and the currency selector (USD, INR, EUR, GBP) only changes the symbol and number formatting; it does not convert amounts.
- Weekly hours saved
- weeklyHours = people × hoursPerWeek × (savingsPct / 100)
- Annual hours saved
- annualHours = weeklyHours × 52
- Gross annual savings
- annualGross = annualHours × hourlyCost
- Annual running cost
- annualRun = monthlyCost × 12
- Year-1 net benefit
- year1Net = annualGross − annualRun − buildCost
- Year-1 ROI
- roi = year1Net / (annualRun + buildCost) × 100 (shown as 0 if both costs are 0)
- Monthly net
- monthlyNet = weeklyHours × 4.333 × hourlyCost − monthlyCost
- Payback
- payback = buildCost / monthlyNet (shown as a dash if monthlyNet is 0 or less; "<1 mo" if under one month)
What it assumes
- Saved time is valued at the loaded hourly cost you enter. That is the value of the capacity freed, not necessarily cash saved, unless headcount or spend actually drops.
- Savings start on day one at full strength. There is no adoption curve, training period, or build time before benefits begin.
- Year-1 ROI charges the whole build cost to year 1, so a project can show a negative year-1 result and still pay back in year 2.
- There is no separate maintenance line; put monitoring, model updates and review of AI output into the monthly running cost.
- The calculation runs in your browser and your inputs are not sent anywhere. Clicking the consulting button records a generic click event with no inputs.
Worked example
Illustrative only: a team of 8 people each spends 6 hours a week on a repetitive document-handling task. An automation is expected to save 25% of that time. Figures in USD.
| Input | Value |
|---|---|
| People doing the task | 8 |
| Hours/week each spends on it | 6 |
| Loaded cost per hour | $60 |
| Time saved by AI | 25% |
| One-time build cost | $30,000 |
| Monthly running cost | $800 |
- weeklyHours = 8 × 6 × 0.25 = 12 hours
- annualHours = 12 × 52 = 624 hours
- annualGross = 624 × $60 = $37,440
- annualRun = $800 × 12 = $9,600
- year1Net = $37,440 − $9,600 − $30,000 = −$2,160
- roi = −2,160 / (9,600 + 30,000) × 100 = −2,160 / 39,600 × 100 = −5.45, shown as −5%
- monthlyNet = 12 × 4.333 × $60 − $800 = $3,119.76 − $800 = $2,319.76
- payback = $30,000 / $2,319.76 = 12.93, shown as 13 mo
Year-1 net benefit −$2,160, ROI (year 1) −5%, payback 13 mo, 624 hours saved per year. The first year is slightly negative because the build cost is charged to it in full, yet the build cost is recovered in about 13 months, so under these assumptions the cumulative net turns positive during month 13.
Frequently asked questions
- What should I enter for the loaded cost per hour?
- Use the full cost of an hour of the people doing the task, including salary, benefits and overhead, not just their pay rate. If you do not have that figure, ask finance for a standard loaded rate for the role. The tool treats the number as given and applies no adjustments.
- How do I pick a realistic value for time saved by AI?
- Measure rather than guess: time a handful of real tasks by hand, run them through a prototype, and compare, including the time spent checking the output. Then try the tool with a lower percentage as well, since this is usually the hardest input to know and the results scale directly with it.
- Why is year-1 ROI negative when there is still a payback figure?
- Year-1 ROI charges the full build cost to the first 12 months, while payback only asks how long monthly net savings take to recover the build cost. A payback longer than about 12 months means year 1 is negative, even if the project is positive afterwards.
- What does a dash for payback mean?
- It means monthly savings do not exceed the monthly running cost, so the build cost can never be recovered with these inputs. Raise the savings, cut the running cost, or reconsider whether the task is worth automating.
- Can I use the result in a business case?
- Use it as a starting point and show your assumptions next to it. A reviewer will want a sensitivity range, an adoption ramp and evidence for the savings percentage, none of which this tool models. The linked guides cover how to build that fuller case.
Read the guide
Building this for real?
I help teams design and ship AI systems — cost modelling, RAG, agents, and architecture.
Work with me