Skip to content
Subscribe
Rafael Cordeiro, colunista

AI ROI, counting tokens is easy and proving value is another story

Rafael Cordeiro · 2 min read
Share

Part 1 of 2 on AI ROI. Part 2 covers the work that used to never happen.

Counting tokens is easy. Proving that AI created value is another story.

Recently, The Economist published an analysis of an important shift in the way companies are evaluating artificial intelligence.

Adoption metrics

At first, the main concern was to encourage adoption. That is why many companies began tracking four metrics.

  • Number of users.
  • Number of prompts.
  • Calls made.
  • Tokens consumed.

These indicators help show whether the technology is being used. But on their own, they do not show whether AI is generating results.

The question reaching executives

The question now reaching executives is harder. What return has AI generated?

Saving time, for example, does not automatically create value. The time freed up has to turn into something concrete.

  • Higher revenue.
  • Lower costs.
  • Better quality.
  • Shorter execution time.
  • Higher customer satisfaction.
  • More work capacity without growing the team.

Returns have a learning curve

It is also important to remember that the return may not show up right away.

There is a learning curve. People have to learn to use the tools, processes have to be redesigned and new bottlenecks may appear. In some cases, productivity even drops at first before improving.

That is why measuring usage volume alone can lead to the wrong conclusions.

What happened because of AI

The most important question is what happened because of AI that would not have happened without it.

To answer it, companies need to track at least three dimensions.

  1. Technology consumption and cost.
  2. Operational and financial results, without ignoring quality.
  3. The organization's ability to learn and adapt its processes.

The more AI evolves from support tools into systems able to execute tasks, the more important it will be to connect each action to a verifiable result.

Without that connection, companies will have dashboards full of numbers, but little clarity about the value actually created.

That Economist article helps explain why the next phase of AI will not be only about adoption, but about return, productivity and transforming the way we work.

In short

  • Counting tokens is easy. Proving that AI created value is another story.
  • The question is what happened because of AI that would not have happened without it.
  • Each AI action needs to be connected to a verifiable result.

Further reading

Originally published on LinkedIn on September 8, 2026. See the original post. This version has been revised and expanded with context and sources. The views are those of the author, a co-founder of OperOS AI.

Read next