According to the PwC Global CEO Survey 2026, 56% of executives say AI has not yet delivered revenue growth or cost savings. Yet Deloitte's research shows that two-thirds of organizations report productivity and efficiency gains from AI. The difference comes down to how companies approach the evaluation. Here are five steps that help calculate AI's return before committing to investment.
What Makes AI ROI Hard to Measure
When a company buys a new machine, the calculation is straightforward: equipment cost, output per unit of time, payback period. AI works differently — its impact is often distributed across several processes, builds up gradually, and doesn't always translate directly into financial metrics.
Standard methods for evaluating IT projects don't fully apply here. Below are five steps we use with clients to assess AI's return before implementation begins.
Step 1. Determine the Cost of the Current Process
Before calculating how much AI will save, it's worth understanding how much the company is spending now. Routine processes rarely raise questions, and their true cost often becomes visible only through deliberate analysis.
When we ran a workshop with LTH Baas, a Baltic marine engineering and shipbuilding company, participants started with a simple question: where does the time go? The analysis revealed that document searches, supplier communications, and manual coordination between departments consumed around 280 hours per month — roughly 1.5 full-time employees.
What to do: Select 2–3 processes that take the most time. Measure the hours spent per week, the number of people involved, and the cost of those hours. This is your baseline — all savings are calculated from here.
Step 2. Determine Where the Freed-Up Time Will Go
Time savings alone don't automatically equal cost savings. Real financial value appears when freed-up hours are redirected to revenue-generating activities — sales, development, client work. Or when the company can handle the same volume with fewer resources. Or when faster processing directly impacts revenue.
When we worked with Arens, Estonia's largest kitchen manufacturer, the team found that sales staff spent 1.5–2 hours daily gathering information to respond to customer inquiries. Meanwhile, 300 social media inquiries per month went unanswered — there simply weren't enough resources. The estimated savings came to €130,000 per year, driven by specific lost sales and employee overload.
What to do: For each process, define where the freed-up time will go. The more specific the answer, the more accurate the ROI calculation.
Step 3. Choose What Can Be Launched Within 90 Days
According to McKinsey, only 6% of companies generate a meaningful share of profit from AI. One common reason is starting with large-scale projects that require months of development and complex integration.
An approach that works well in practice: begin with initiatives that deliver results within 90 days and don't require heavy IT infrastructure.
In a workshop with Verston, an Estonian road construction company, the team identified 34 potential AI use cases. Instead of tackling all of them, they selected 4 with the highest impact at the lowest investment. The estimated savings from just those four — €750,000 per year. Most solutions required no major development or capital expenditure.
The faster the first result, the easier it is to justify further investment.
What to do: From your list of opportunities, select 3–5 with the shortest path to implementation. Evaluate each on three criteria: business impact, complexity, and required investment. Start where all three are in your favor.
Step 4. Account for Benefits That Are Hard to Express in Euros
Not all value from AI translates into financial metrics. According to Deloitte, 65% of organizations already recognize AI as part of their corporate strategy — and understand that not all of its benefits are financial or immediate.
Examples from our practice:
- Risk reduction. During a workshop with VKG, Estonia's largest industrial enterprise, one team built a training simulator for production machine operators. The direct financial impact will emerge over time, but every prevented production incident represents significant costs — and, more importantly, employee safety.
- Knowledge sharing. At the same workshop, another participant transformed internal incident reports into audio podcasts. Written reports are a valuable source of information, but in audio format, this knowledge reaches a much wider audience within the company.
- Faster decision-making. When thousands of rows of machine performance data are turned into a visual dashboard with filters, analysis that used to take hours now takes minutes.
What to do: When evaluating AI initiatives, add a column for non-financial benefits — risk reduction, data quality, decision speed, employee satisfaction. These factors often prove decisive when making the case for budget approval.
Step 5. Calculate ROI for a Specific Pilot, Not for AI in General
AI is not equipment with fixed output. It's closer to building a capability — an investment that delivers increasing returns over time. According to Deloitte, 15% of organizations using generative AI already see significant, measurable ROI, and another 38% expect it within a year.
Rather than trying to calculate ROI for AI across the entire company, it's more effective to evaluate a specific pilot.
The formula we use:
Pilot ROI = (current process cost × expected reduction %) — pilot cost
For example:
- Current process costs €5,000/month (salaries × hours)
- AI can reduce costs by 40%
- Savings: €2,000/month = €24,000/year
- Pilot cost: €8,000
- Payback period: 4 months
One working pilot with proven results is the strongest case for scaling.
What to do: Launch one pilot on the most straightforward process. Measure results after 30–60 days. Use that data to build the case for the next steps.
Key Takeaways
- Сalculate what the current process costs — time, salaries, missed opportunities.
- Define where freed-up time will go. Without this, savings stay on paper.
- Start with quick wins. 90 days to the first result is a good benchmark.
- Account for non-financial benefits. Risk reduction, decision quality, and speed often outweigh direct savings.
- The pilot is your strongest argument. One working case with real numbers is more convincing than any presentation.
The main barrier to AI adoption isn't cost — it's uncertainty.
In a free 60-minute consultation, we help remove it: together, we choose the first pilot, estimate the potential savings, and define the next steps.
In a free 60-minute consultation, we help remove it: together, we choose the first pilot, estimate the potential savings, and define the next steps.