01What this tool estimates
The calculator produces a parametric estimate of the return from adopting AI, built from a fixed set of assumptions applied consistently to whatever gets entered: industry, company size, seven yes/partially/no answers about how operations run today, how the workforce is split by role, and how aggressively the investment is planned. Change any one of those, and every downstream number changes with it, deterministically, using the exact formulas below.
It explicitly does not estimate a few things people sometimes assume it does:
- It is not a quote. No vendor pricing, contract, or procurement data feeds into it.
- It is not a certified assessment, audit, or compliance deliverable of any kind.
- It is not financial, legal or tax advice. Decisions with real consequences should be checked with a qualified professional.
- It does not account for specific vendors, contracts, existing tooling, or team capability, only industry-level averages and the inputs given to it.
The full legal scope of these tools is covered in the Terms of Use.
02The inputs you provide
Five inputs drive every calculation:
- Industry: one of twelve sectors (Manufacturing, Technology, Professional Services, Healthcare, Financial Services, Retail & Ecommerce, Transportation & Logistics, Aviation, Hospitality, Education, Government, Construction, or Other). Each carries its own loaded hourly cost, a complexity multiplier, seven sector-specific diagnostic questions, and a default role mix.
- Company size: a headcount band (e.g. "51-200 employees"), mapped to a single representative employee count used in every formula (the 51-200 band uses 120 employees, for example).
- Seven diagnostic questions, specific to the chosen industry (e.g. for Technology: "Do more than a quarter of support tickets repeat the same issues?"). Each answer is No = 0, Partially = 0.5, or Yes = 1, and each question carries its own weight (roughly 0.7 to 1.2) reflecting how strongly it correlates with automatable work.
- Role mix: the percentage split of the workforce across the industry's roles (e.g. "Engineering 40%, Support 15%..."), pre-filled with sensible defaults that can be edited. Must total exactly 100%.
- Investment level: Starter, Growth, or Transformation, explained in full in section 9. This single choice changes the adoption rate, licensing cost, services scale, and the share of saved hours that convert to real value.
Nothing else is asked. There is no login, no company name, no financial data.
03Step-by-step: from answers to "% of workforce affected"
Each answer is multiplied by its question's weight, and the seven results are summed into a single score. That score is then converted into the share of the workforce whose work is considered "affected" by repetitive, automatable tasks, with a floor and a ceiling:
score = Σ (answerValue × questionWeight) // 7 questions, weights ≈0.7-1.2
affected% = min( baseAffected + score × affectedPerPoint, maxAffected )
= min( 0.15 + score × 0.045, 0.60 )
Answering Yes to the 4 highest-weighted questions (1.2 + 1.0 + 1.1 + 1.0) and No to the rest:
score = 1×1.2 + 1×1.0 + 1×1.1 + 1×1.0 + 0×0.9 + 0×0.8 + 0×0.7 = 4.3
affected% = min(0.15 + 4.3×0.045, 0.60) = min(0.15 + 0.1935, 0.60) = 0.3435 → 34.35%
The floor (15%) reflects that almost every organization has some baseline repetitive work, even if every answer is "No." The ceiling (60%) reflects that no realistic organization has more than about 60% of its workforce doing purely automatable work.
04Step-by-step: the role-mix factor
Not every role benefits from AI equally. Each role in an industry's default mix carries a gain value (0 to 1) representing how much of that role's work AI can realistically take on today. The role factor is the weighted average of those gains, weighted by the percentage assigned to each role:
roleFactor = Σ ( share% × role.gain ) // across all roles, shares sum to 100%
| Role | Share | Gain | Contribution |
|---|---|---|---|
| Engineering | 40% | 0.48 | 0.192 |
| Product & design | 10% | 0.50 | 0.050 |
| Support | 15% | 0.62 | 0.093 |
| Sales & CS | 15% | 0.45 | 0.0675 |
| Marketing | 8% | 0.55 | 0.044 |
| Admin & operations | 12% | 0.55 | 0.066 |
roleFactor = 0.192 + 0.050 + 0.093 + 0.0675 + 0.044 + 0.066 = 0.5125 → 51.25%
This is why two companies with identical answers to the seven questions can get very different results: a workforce weighted toward high-gain roles (support, admin, analysis) shows a larger effective automation potential than one weighted toward low-gain, hands-on roles (frontline production, drivers, clinical staff).
05Step-by-step: automatable hours per year
With the affected share and the role factor in hand, the model multiplies through headcount, a fixed weekly hours assumption, and the working year, then discounts by the investment level's adoption rate (the share of affected staff who actually use the tools day to day):
hoursYear = employees × affected% × hoursPerWeekAffected × weeksPerYear
× roleFactor × adoption%
= employees × affected% × 6 × 46 × roleFactor × adoption%
120 employees (51-200 band), 34.35% affected, roleFactor 0.5125, adoption 65%:
hoursYear = 120 × 0.3435 × 6 × 46 × 0.5125 × 0.65 = 41.22 × 6 × 46 × 0.5125 × 0.65 = 11,376.72 × 0.5125 × 0.65 = ≈3,789.9 hours/year
The 6 hours/week and 46 working weeks/year figures are fixed global assumptions (not editable per industry), representing a conservative slice of a working week spent on genuinely repetitive, automatable tasks for an "affected" employee.
06Step-by-step: turning hours into money
Hours are converted to currency using the industry's average loaded hourly cost, discounted by the investment level's productivity conversion rate: the share of a saved hour that actually shows up as usable value rather than being reabsorbed, spent re-checking output, or lost to friction.
savings = hoursYear × industry.hourly × productivityConversion
savings = 3,789.9 × €45 × 0.75 = 170,545.5 × 0.75 = ≈€127,909 / year
Industry hourly costs range from €26/h (Hospitality) to €52/h (Financial Services) in the model, reflecting typical loaded labor cost differences across sectors, not any specific salary data about a given company.
07Step-by-step: the cost side, services and licensing
Year-1 investment has two components. Services (implementation, training, integration work) scale sub-linearly with headcount, i.e. bigger companies pay more in absolute terms but less per employee, an economy of scale captured by an exponent below 1, then get multiplied by the industry's complexity factor and the investment level's services scale.
services = ( costBase + costPerEmployee × employees^costExponent )
× industry.complexity × level.servicesFactor
= ( €6,000 + €90 × employees^0.9 ) × complexity × servicesFactor
licenses = employees × affected% × level.toolCostPerUserYear // per-seat, affected staff only
cost = services + licenses
complexity 0.9, servicesFactor 1.0, toolCostPerUserYear €480:
services = (6,000 + 90 × 120^0.9) × 0.9 × 1.0 = (6,000 + 90 × 74.35) × 0.9 = (6,000 + 6,691) × 0.9 = ≈€11,422
licenses = 120 × 0.3435 × 480 = 41.22 × 480 = ≈€19,786
cost = 11,422 + 19,786 = ≈€31,208
Licensing only covers the affected share of employees, since those are the people the model assumes would actually be given AI tools, not the entire headcount.
08Step-by-step: ROI and payback
With savings and cost both computed, ROI and payback are simple ratios. Nothing new gets introduced here, just division:
roi = (savings − cost) / cost × 100
paybackMonths = cost / (savings / 12) // months to recoup the year-1 investment
fte = hoursYear / 1,720 // full-time-equivalent hours/year
roi = (127,909 − 31,208) / 31,208 × 100 = ≈310%
paybackMonths = 31,208 / (127,909/12) = 31,208 / 10,659 = ≈2.9 months
fte = 3,789.9 / 1,720 = ≈2.2 FTE capacity unlocked
The resulting ROI percentage is then bucketed into an opportunity level shown as a badge on the results: Low (ROI ≤ 60%), Medium (60% < ROI ≤ 180%), or High (ROI > 180%). These thresholds only change which of three pre-written "next steps" lists is shown; they never feed back into the numbers themselves.
09The three investment levels
This is deliberately the single biggest lever in the whole model, and it exists to make one point explicit: this is not "buy everyone a Claude or ChatGPT seat." Each level implies a genuinely different depth of implementation work, and that depth, not just a subscription price, is what the numbers above assume actually gets delivered.
| Level | Adoption | Tool cost / user / yr | Services scale | Productivity conversion |
|---|---|---|---|---|
| Starter | 45% | €300 | ×0.35 | 70% |
| Growth | 65% | €480 | ×1.0 | 75% |
| Transformation | 80% | €700 | ×1.9 | 80% |
Starter
Self-serve tools, minimal setupA team starts with off-the-shelf AI assistants and learns by doing. No consultants, no heavy integration work. The cheapest way to build AI habits and validate value before committing real budget. Expect quick wins on writing, research and support tasks.
Growth
Structured program, internal ownerA coordinated rollout led from inside: shared workflows, role-specific training and light integration with the systems already in use. Balanced cost and impact. The most common path for teams that have already validated the basics and want measurable results.
Transformation
Deep redesign, external supportProcess redesign with custom integrations, dedicated budget and external specialists. The highest investment and the highest ceiling: AI embedded into core workflows instead of assisting around them. Best suited to organizations with executive sponsorship and a clear roadmap.
Notice what moves together: a higher tier assumes more people actually adopting it (adoption), more of the saved hour landing as real value (productivity conversion), and a proportionally bigger services bill. A Transformation-level ROI is not "the same plan, more expensive." It is a different, deeper plan, priced accordingly and credited accordingly.
10The optional AI-written narrative
Everything above this line is 100% deterministic arithmetic that runs in the browser. Below the results, the calculator offers one additional, optional layer: a short narrative (a diagnosis, priorities, risks, and a "should you do this" verdict) written by a third-party AI language model reviewing the numbers that were already computed by the formulas on this page.
What actually leaves the browser for that step is a small payload of already-calculated results: industry name, company size band, chosen investment level and its parameters, and the final metrics (ROI%, savings, cost, hours, payback, affected%, FTE, top impact area). The seven raw answers, the exact role percentages, and any identifying detail are never sent, only the outputs.
That model is explicitly instructed to reason only over the figures it is given, never to invent statistics or restate numbers as if they were new findings, and it structurally cannot alter any figure shown above it on the page; it only adds commentary underneath. If that service is unavailable or times out, the page falls back to a short summary generated locally from the entered numbers, clearly labeled as such, so the page never blocks or breaks.
Full detail on what data is processed and how is in the Privacy Policy.
11Limitations and honest caveats
- This is a parametric model built on reasonable, stated assumptions, not a model fitted to external market data, benchmarking studies, or a specific company's actual financials.
- Every constant on this page (weights, factors, thresholds) is a deliberate, defensible estimate, not a number derived from a dataset. The exact figure is shown outright rather than hidden behind a vague "industry average."
- Real results vary by execution quality, change management, existing tooling, and dozens of factors this model does not and cannot see.
- AI tool pricing changes frequently; the licensing figures are indicative, not live vendor quotes.
- None of this constitutes financial, legal or tax advice. See the Terms of Use for the full legal scope.