The ROI of generative AI
ROI is the return on investment. Where it shows up, where it disappears, and why measurement has to be born with the project.
The ROI of generative AI is not about technology. It is about business indicators. ROI is the return on investment: how much the project gives back compared with what it cost.
In my daily work, I use generative AI across the whole software development cycle, to raise the team's productivity and the quality of what gets delivered. And precisely because I am in that daily work, one thing became very clear to me: the number of prompts, tokens consumed or active users are not measures of success.
What matters is a simple question: is AI improving the indicator that matters to the business?
Four dimensions to measure
An article by Paulo França presents a model in 4 dimensions that I consider essential for any technical leader who wants to move from "AI is cool" to "AI produced this result":
- Financial: new revenue, savings, return on investment.
- Operational: productivity, process time, capacity.
- Quality and risk: errors, rework, incidents avoided.
- Adoption: real usage, frequency, growth.
Measurement is born with the project
Many companies say AI helps them innovate, but few can show the impact on results. The difference is almost always in the measurement structure: it has to be designed from the start of the project, not afterwards.
In practice, this means choosing, before building, which indicator the project needs to move and how it will be measured.
Originally published in Portuguese on LinkedIn, where the conversation continues in the comments.
