In 2024, we experimented. In 2025, we implemented. Now, in 2026, we have to justify.
The euphoria of the first prompts and the fun of images generated in seconds have given way to a blunt question in management meetings: “What is the real return on investment (ROI) of our ChatGPT, Claude or Midjourney licences?”
For many companies, the answer remains vague. There is a feeling that “things move faster”, teams visibly produce more, but the direct link between the SaaS invoices and the profit line stays invisible. Worse still, the time saved is often an illusion if it is immediately swallowed up by correcting AI-generated errors or by overproducing content that nobody reads.
Yet the ROI of generative AI is not a myth. It simply hides where you least expect it: not in replacing people, but in drastically reducing their “opportunity cost”.
How do you turn abstract productivity gains into concrete financial indicators? How do you measure the impact of AI on your Time-to-Market and your operating margins? A deep dive into the Smart Impact methodology for steering your marketing by performance rather than by hype.
The 3 key indicators (KPIs) of AI productivity
To steer the profitability of AI, you have to move on from anecdotes (“We generated that visual in 10 seconds!”) and into flow measurement. At Smart Impact, we recommend tracking three major indicators that isolate the impact of artificial intelligence on your value chain.
1. The “Hours per Deliverable” (HPL) ratio
This is the basic metric of operational efficiency. It consists of measuring the total human time needed to finalise one unit of content, from design to final approval.
Take, as a typical example, a weekly B2B newsletter at a Fribourg-based SME:
- Before AI: the team spends 6 hours per issue (research, writing, corrections, layout).
- With AI: the process drops to 1.5 hours, with the human keeping the brief, the critical edit and the strategy.
- The calculation: a 75% reduction in production time, or 4.5 hours of capacity freed each week, since the newsletter comes out once a week.
2. The Marginal Cost of Variation (CMD)
One of the biggest levers AI offers is not doing one thing faster, but producing fifty variations for the same effort.
- The idea: in a conventional advertising campaign (Google Ads or Meta), creating 20 variants of copy and images was expensive in designer and copywriter hours.
- The AI effect: once the prompt and the initial concept are approved, the cost of generating the next 19 variants becomes almost nil.
- The measure: divide the total cost of the campaign by the number of variants tested. If your cost per variant falls by 80%, your capacity for A/B testing grows markedly without inflating your fixed costs.
3. Reducing “Time-to-Market” (TTM)
ROI is not measured only in money saved, but in speed of reaction. In a competitive market such as Switzerland, being the first to react to a news story or to launch a promotion is a major competitive advantage.
- The indicator: measure the time between identifying a marketing opportunity and it actually going live.
- The AI effect: AI acts as a catalyst that removes the bottlenecks (waiting for a first draft, simple image retouching). If your time to market falls from 10 days to 48 hours, your commercial agility becomes a direct driver of revenue.
The Smart Impact advice: do not measure the company's overall productivity straight away. Isolate one repetitive task (monthly reporting or writing product pages, say) and apply these 3 KPIs over a quarter to get a first solid measurement, before extending the approach.
Calculating ROI: the Smart Impact formula
Measuring productivity is good. Turning that productivity into Swiss francs is better. To obtain an ROI (return on investment) you can defend in front of your management, you need to weigh the gains generated against the real costs of implementation.
1. Identifying the components of the equation
To calculate the ROI of your move to AI, we use four key variables:
- Productivity gains (G): the monetary value of the time freed (number of hours saved × the employee's loaded hourly cost). Note: G only becomes a real gain if that time is reallocated to a measurable activity (more campaigns, requests handled faster); if it dissolves into the working day, it stays theoretical and does not reach the cash position.
- Revenue increase (R): the additional turnover generated by the capacity to produce more campaigns or respond to leads faster.
- Licence costs (L): the price of the subscriptions (ChatGPT Enterprise, Claude, Midjourney, Jasper and so on).
- Training and setup costs (F): the time spent training the teams, building prompt libraries and integrating AI into your existing workflows.
2. The AI ROI formula
Applied over a given period (quarter or year), the equation looks like this:
ROI (%) = ((G + R) − (L + F)) / (L + F) × 100
A concrete example:
If a team of 3 people saves 10 hours a week thanks to AI (value estimated at CHF 4’000.- per month) and this makes it possible to launch two additional campaigns generating CHF 10’000.- of net margin, for a total cost (licences + training) of CHF 2’000.-:
ROI = ((4’000 + 10’000) − 2’000) / 2’000 × 100 = 600%
3. Opportunity Gain: the invisible multiplier
Pure accounting often forgets the most powerful factor: what do your talented people do with the time they get back?
The return mostly comes into play when your marketing manager, once buried in writing technical product sheets (a low value-added task), now spends 15 hours a week on conversion strategy or competitor analysis. AI does not only cut costs; it reallocates your human intelligence to where it counts most.
An external order of magnitude
Serious studies call for nuance. The 2023 experiment run by researchers from Harvard Business School with the Boston Consulting Group, with 758 consultants, measured that with GPT-4 they completed 12% more tasks, 25% faster, with quality rated 40% higher, on tasks within the reach of AI. On a task deliberately chosen outside that reach, the consultants using AI were, by contrast, 19 percentage points less likely to find the right answer. For your calculation, the lesson is twofold: the gain exists on well-chosen tasks, and it can turn into a loss on the others. That is why we measure task by task rather than applying one overall percentage.
The traps that distort AI ROI calculations
Careful: a positive ROI on paper can hide misleading realities. Here are three classic mistakes to avoid in your calculation:
- Quality debt: if AI produces 10x more content but your conversion rate falls because the messages are generic and soulless, your real ROI is negative. Quantity must never cannibalise performance.
- The cost of verification (human-in-the-loop): many companies forget to count the review time. An article generated in 2 minutes that needs an hour of factual correction is still profitable, but less so than people think.
- Internal information overload: producing at scale “because it is easy” saturates your communication channels and can tire your audience. ROI is measured on the final impact, not on the volume of files produced.
A typical scenario: a two-person marketing team
Imagine a services company in Lausanne whose marketing team has two people, absorbed by day-to-day execution.
- Starting point: the team estimates it spends around twenty hours a week on blog articles, newsletters and social media posts, and tests only two ad variants a month.
- Set-up: we would begin with prompt templates adapted to its voice, a systematic review step and automation of the first draft of reporting.
- Measurement: over a quarter, the team records the hours per deliverable, the number of variants tested and the time to publication, then compares them with the starting point.
- Decision: if hours are freed up, the team decides in advance where they will go (sales funnel optimisation, ad tests) and tracks these new activities, otherwise the gain stays on paper.
The financial assessment is then done with the formula above, based on the hours actually measured and the real costs of licences and training, without extrapolating a percentage increase in conversions or leads.
AI as a multiplier, not as a replacement
The ROI of generative AI is not a magic formula for cutting jobs. It is a growth lever that lets your talent step out of repetitive execution and become strategists again.
In 2026, the question is no longer “How much does AI cost?”, but “How much is the absence of AI in our processes costing us?”. The cost of the missed opportunity (campaigns never launched, leads not handled quickly enough, tests never run) is the heaviest burden on your profitability.
Measuring AI ROI is, above all, about regaining control of your most valuable resource: your experts' time. To move from the calculation to a first use case, our support with process audits and automation helps you scope the project.
Ready to turn your marketing costs into profitable investments? Do not leave your AI strategy to chance. At Smart Impact, we analyse your internal processes to spot where AI can save you time. Contact our experts for a tailored ROI assessment.




