Free tool, runs in your browser
AI ROI Calculator
Put real numbers on an AI project before you commit. Enter your team size, the hours you expect to save and what the build will cost, and get the annual saving, payback period and 3-year net value. No signup, and nothing you enter leaves your browser.
Savings assume 46 working weeks per year, which allows for UK leave and bank holidays. Nothing you type leaves your browser.
> roi for 10 people @ 60% adoption
- Annual gross saving
- £44,160
- Annual net saving
- £38,160
- Payback period
- 3.1 months
- 3-year net value (after implementation cost)
- £104,480
What realistic values look like
The calculator is only as honest as its inputs, so start conservative. For teams with properly built AI workflows, 2 to 8 hours saved per person per week is the typical range: the low end for occasional drafting and summarising, the high end for roles heavy on writing, research or reporting. Adoption is the number most business cases fudge. Without training and visible leadership use, 40 percent is common; well-run rollouts reach 70 to 80 percent.
Gross saving is calculated as team size × hours saved × adoption rate × hourly cost × 46 working weeks. Net saving subtracts 12 months of running costs, and payback divides the one-off implementation cost by the monthly net saving. If a project only stacks up at 90 percent adoption and 10 hours a week, it does not stack up.
What is a fully loaded hourly cost?
Salary plus everything else it costs to employ someone: employer National Insurance, pension, software, equipment, office overhead. A rough shortcut is salary divided by 1,600 working hours, then multiplied by 1.3 to 1.5. For a £40,000 salary that lands around £33 to £38 per hour.How many hours per week can AI realistically save?
For teams with real AI workflows in place, 2 to 8 hours per person per week is the typical range. Drafting, summarising, research, reporting and first-pass analysis are where most of it comes from. Claims much above 10 hours a week usually mean the baseline was measured generously.What adoption rate should I assume?
Be honest rather than optimistic: 40 to 80 percent is the realistic band. Even with good tools, some people will not change how they work without training and visible leadership use. If you have no rollout plan, model 40 percent and treat anything better as upside.What should I include in the implementation cost?
Everything one-off: consultancy or internal build time, integration work, data preparation, training and the productivity dip while people learn. Monthly running cost covers subscriptions or API usage, hosting and a little ongoing maintenance.What if the calculator says the project never pays back?
Believe it, then change the inputs rather than the spreadsheet. The usual fixes are picking a higher-volume process, improving adoption with training, or cutting running costs with a cheaper model. If none of those move the number, that project is not the right first project.
Want a business case that survives scrutiny?
I scope AI projects with honest numbers, pick the workflow that pays back fastest and build it, so the savings show up in the accounts rather than the slide deck.