How to Calculate ROI on AI Before You Spend a
Euro
benchmarks, and a step-by-step guide for business leaders.
Every AI vendor will tell you their solution delivers "incredible ROI." Fewer will help you calculate
that ROI before you commit a single euro.
That's a problem — because in 2026, AI investment isn't a leap of faith. It's a business decision
that should be backed by numbers, just like hiring a new team member, leasing new office space,
or investing in equipment.
Here's a practical framework for calculating the return on an AI investment before you sign
anything. No jargon, no hand-waving — just the math that matters.
Why Most ROI Calculations Go Wrong
Before we get into the framework, let's address why AI ROI is so often miscalculated:
Mistake 1: Comparing AI cost to zero. The real comparison isn't "AI vs nothing" — it's "AI vs the
current cost of doing things manually." If a manual process costs your business €80,000 per year,
the question is whether AI can do it for less.
Mistake 2: Ignoring indirect costs. The price your AI vendor quotes is just the build cost. You
also need to factor in implementation time, training, productivity dips during transition, ongoing
maintenance, and API fees.
Mistake 3: Overestimating automation rates. A vendor might promise "90% automation," but in
practice, most business processes see 50–70% automation in the first year. Expecting more leads
to disappointment and under-investment in human oversight.
Mistake 4: Only counting cost savings. AI doesn't just save money — it can generate revenue.
Faster lead response times, better customer experiences, and new capabilities all drive top-line
growth. A complete ROI calculation includes both savings and revenue gains.
The AI ROI Framework: 5 Steps
Step 1: Calculate the Current Cost of the Target Process
Start with the process you want to automate. Break it down:
- How many hours per week does your team spend on this process?
- What's the blended hourly cost (salary + benefits + overhead) of the people doing it?
- Are there any direct costs (software, outsourcing, error correction)?
Formula: Annual Process Cost = (Weekly Hours × Hourly Cost × 52) + Direct Costs
Example: A construction company's admin team spends 30 hours per week processing project
documentation. Blended hourly cost: €35. Direct costs (printing, courier): €5,000/year.
Annual Process Cost = (30 × €35 × 52) + €5,000 = €54,600 + €5,000 = €59,600
Step 2: Estimate the AI Automation Rate
Not every part of a process can be automated. Estimate what percentage of the work AI can
handle versus what still requires human involvement.
Conservative benchmarks for 2026: — Email triage and routing: 75–85% automation — Document
classification and data extraction: 60–75% automation — Lead qualification and scoring: 65–80%
automation — Report generation: 70–85% automation — Customer inquiry responses: 55–70%
automation — Contract review and compliance checking: 50–65% automation
A 2025 study by Harvard Business Review found that across 400 enterprise AI deployments, the
median automation rate in the first year was 62%. Plan conservatively.
Step 3: Calculate Projected Annual Savings
Formula: Annual Savings = Annual Process Cost × Automation Rate
Example (continued): With a 65% automation rate: Annual Savings = €59,600 × 0.65 = €38,740
But don't stop here. Also estimate revenue gains: — If faster document processing lets you take on
2 more projects per year at €30,000 average margin, that's €60,000 in additional revenue. — If
faster lead response improves your close rate by 10%, calculate the value of those additional
deals.
Step 4: Calculate Total Cost of AI Implementation
Include everything: — Build/setup cost: The vendor's quoted price (e.g., €35,000) — Data
preparation: Add 15–25% if your data needs cleaning (e.g., €7,000) — Training and change
management: Internal time cost (e.g., €3,000) — Year 1 maintenance and optimisation:
Typically 10–15% of build cost (e.g., €4,000) — API and hosting costs: Monthly fees × 12 (e.g.,
€600/month × 12 = €7,200)
Total Year 1 Cost: €35,000 + €7,000 + €3,000 + €4,000 + €7,200 = €56,200
Step 5: Calculate ROI and Payback Period
ROI Formula: ROI = ((Annual Savings + Revenue Gains - Total Year 1 Cost) / Total Year 1 Cost)
× 100
Example: ROI = ((€38,740 + €60,000 - €56,200) / €56,200) × 100 = 75.7%
Payback Period: Total Year 1 Cost / (Monthly Savings + Monthly Revenue Gains)
Monthly benefit = (€38,740 + €60,000) / 12 = €8,228 Payback Period = €56,200 / €8,228 = 6.8
months
Year 2 and Beyond
This is where AI investments really shine. In Year 2, your costs drop dramatically because the build
is done. You're only paying for maintenance, optimisation, and API fees — typically
€10,000–€15,000 per year. But the savings and revenue gains continue (and often increase as the
system improves).
Year 2 ROI using the example above: Year 2 Cost: €12,000 (maintenance + API) Year 2 Benefit:
€98,740 (savings + revenue) Year 2 ROI: 723%
This compounding effect is why companies that invest in AI early tend to pull ahead of competitors
quickly. A 2026 McKinsey analysis found that companies with 2+ years of AI deployment had 2.4x
the productivity gains of companies in their first year.
Real-World ROI Benchmarks
To ground your calculations, here are real-world benchmarks from various industries in
2025–2026:
- Professional services (document automation): Average 45% cost reduction in document
processing, 8-month payback
- Construction (project management AI): Average 30% reduction in administrative overhead,
10-month payback
- Retail and hospitality (customer service AI): Average 35% reduction in support costs,
6-month payback
- Property management (tenant communications): Average 50% reduction in response
time, 25% improvement in tenant satisfaction, 7-month payback
The Break-Even Threshold
Not every AI project makes financial sense. Here's a quick test:
If the annual cost of the manual process is less than 1.5x the estimated AI implementation cost, the
ROI may be too marginal to justify the effort and disruption.
Example: If a manual process costs €20,000/year and the AI solution costs €35,000 to build, you'd
need a very high automation rate (85%+) and significant revenue gains to justify the investment. In
that case, consider simpler automation tools first.
Conversely, if the annual manual cost is 3x or more the AI build cost, it's almost certainly a strong
investment.
Building Your Business Case
When presenting an AI investment to stakeholders, structure your case as follows:
- Current state: What the process costs today (hours, euros, errors, delays) 2. Proposed
solution: What the AI system will do (specific capabilities, not buzzwords) 3. Investment
required: Total Year 1 cost including all hidden costs 4. Expected returns: Conservative savings
+ revenue gains with clear assumptions 5. Payback period: When the investment turns net
positive 6. Year 2+ projections: The compounding benefit of reduced ongoing costs 7. Risk
mitigation: Pilot approach, phased deployment, vendor guarantees
This framework works whether you're convincing a board, a business partner, or yourself.
Your Next Step
If you want help running these numbers for your specific business, LF Labs offers a free AI strategy
call where we'll walk through this framework together. We'll identify your highest-ROI opportunity,
estimate realistic automation rates, and give you a clear picture of costs and returns — before you
spend anything.
Explore how LF Labs can help you build a data-driven AI business case.