Where will AI pay off in your company?
Pick the processes your team spends time on. Get a ranked list of AI automation opportunities — hours and money saved per year, build and running costs, payback period and 12/24-month ROI — with every number traceable to a formula.
Start — pick your processesCoefficients and defaults are pending review by our AI team.
Which processes take your team’s time?
Pick 3–7 for a meaningful comparison.
Check the numbers for each process
Defaults are typical for your company size — replace them with your own where you know them.
Your AI automation opportunities
Planning estimate based on your inputs and published research — not a quote and not a guarantee of savings.
Capacity value: hours × your hourly cost. It turns into cash only if you avoid hiring, reduce overtime or redeploy people to revenue work. Sum of processes. Shared-platform savings when you build several use cases are not counted — the estimate stays conservative.
Top candidates
All selected processes
| Process | Solution | Value / yr | Build (prod.) | Payback | ROI 24m |
|---|
Unlock with the PDF plan
Cumulative cash flow, 24 months
Build cost first, then savings minus running costs as the solution goes live and ramps up over three months.
Pilot plan for your top candidate
Get the full plan (PDF)
All processes, 24-month cash flow, a 4–8 week pilot roadmap, architecture, risks and KPIs for your top 3.
Email me the full plan (PDF)
We’ll unlock everything on this page and send the PDF plan to your inbox.
Validate it with a 4–8 week AI agent pilot
We build the top candidate on your real data, measure it against the baseline and give you a go / no-go with real numbers.
How the estimate works
Hours saved = people × hours per week × 46 working weeks × routine share × the share of that routine AI takes over × (1 − time a person spends checking the result). The share depends on the mix of task types in the process (table below); review time is 10%, 20% or 30% depending on risk.
Value = hours saved × your hourly cost. Build cost comes from the same feature prices as our Project Cost Calculator; the pilot is half of production. High-risk AI Act use cases add 20% for compliance, the conservative scenario adds a 20% reserve.
Running cost = LLM usage (official provider prices, the same engine as our LLM Spend Calculator) + hosting and monitoring + support at 15–25% of the build cost per year.
Payback counts from the start of the project: nothing is saved until go-live (6–12 weeks), then savings ramp up to full over three months. ROI compares savings with build plus running costs over 12 and 24 months.
Automation coefficients by task type
| Task type | Conservative | Base | Optimistic |
|---|---|---|---|
| Classification & routing | 50% | 65% | 80% |
| Data extraction from documents | 45% | 60% | 75% |
| Knowledge search & answers | 15% | 35% | 50% |
| Drafting text | 25% | 40% | 55% |
| Summarising & synthesis | 30% | 45% | 60% |
| Code & tests | 10% | 20% | 35% |
| Multi-step work across systems | 15% | 30% | 45% |
| Judgement, negotiation, accountability | 0% | 5% | 10% |
Coefficients are our planning assumptions informed by the studies below, not figures taken from them.
Capacity value: hours × your hourly cost. It turns into cash only if you avoid hiring, reduce overtime or redeploy people to revenue work.
Research behind the assumptions
- McKinsey Global Institute — The economic potential of generative AI: the next productivity frontier (June 2023). Generative AI with other technologies could automate activities that absorb 60–70% of employees’ time; about 75% of the value falls in customer operations, marketing & sales, software engineering and R&D. mckinsey.com
- Brynjolfsson, Li, Raymond — Generative AI at Work, NBER Working Paper 31161 (2023). An AI assistant raised customer-support issues resolved per hour by 14% on average. nber.org
- Noy & Zhang — Experimental evidence on the productivity effects of generative artificial intelligence, Science 381 (2023). Time on professional writing tasks fell by 40%. doi.org
- Peng et al. — The Impact of AI on Developer Productivity: Evidence from GitHub Copilot, arXiv:2302.06590 (2023). Developers finished an isolated task 55.8% faster. arxiv.org
- Dell’Acqua et al. — Navigating the Jagged Technological Frontier, HBS Working Paper 24-013 (2023). Outside the frontier of what the model does well, consultants using AI were 19 percentage points less likely to reach a correct solution. papers.ssrn.com
FAQ
How do you calculate the ROI of AI automation?
We take the hours your team spends on the routine part of a process, multiply them by the share AI takes over for that type of task and subtract the time a person still spends checking the result. Hours saved × your hourly cost gives the yearly value. From it we subtract the build cost and the running costs — LLM usage, hosting and support at 15–25% of the build cost per year. Payback counts from the start of the project, including the build period and three months of ramp-up.
Where do the automation percentages come from?
They are our planning assumptions for 8 task types, informed by McKinsey Global Institute (2023), Brynjolfsson, Li & Raymond (2023), Noy & Zhang (2023), Peng et al. (2023) and Dell’Acqua et al. (2023). Each has a conservative, base and optimistic value, and you can switch between the three scenarios.
How much does it cost to build an AI agent or assistant?
In this calculator a production build ranges from $12,000 for rolling out a coding assistant to $59,000 for a multi-step agent with tools, using the feature prices of our Project Cost Calculator; a pilot is about half of that. Extra integrations, scattered data and high-risk use under the EU AI Act raise the cost.
What does it cost to run an AI solution every month?
Three parts: LLM usage at the providers’ official prices (the same engine as our LLM Spend Calculator), hosting, vector database and monitoring, and support and improvements at 15–25% of the build cost per year. The LLM bill is often the smallest line.
Are hours saved the same as money saved?
No. Hours saved × hourly cost is capacity value. It becomes cash only if you avoid hiring, reduce overtime or move people to work that earns revenue. That is why the result is a planning estimate, not a promise of savings.
Which AI use cases are high-risk under the EU AI Act?
Annex III lists, among others, recruitment and evaluation of employees, creditworthiness assessment of consumers, admission and assessment in education and remote biometric identification. The calculator flags such processes and links to our EU AI Act Risk Classifier. Dates come from our rules file and are marked indicative until legal review is complete.
Do you store the data I enter?
No. The calculation runs in your browser. Only what you send through a form reaches our server: your name, email, phone and a summary of the calculation for the email with your plan.