Your CEO just handed you the mandate they received from the board: transform how this company works using AI. You've rolled out the agents and done the change management. Now, you’re wondering how to find proof that you’re making the transformation that you were asked for. 

Right now, when leaders are asked to prove ROI on their software and AI investments they are reaching for the most visible metric available: token usage. Tokens are measurable, they're trackable, and they at least confirm that something is happening.

The problem is that most enterprises are measuring AI the same way they measured software adoption a decade ago, and AI is a fundamentally different beast. High usage doesn't guarantee high value. And without understanding how AI is actually being used inside your workforce, you have no lever to pull to improve it.

Wait, hold on. What are tokens? Hint: it’s not crypto.

Tokens are how teams and vendors can measure AI usage. Every time an employee interacts with an AI agent, whether asking a question, generating content, or working through a task, that interaction is measured in tokens. Think of them as the currency of AI activity. The more your teams use AI, the more tokens are consumed.

It makes sense that this became the go-to metric. For a long time, it’s been one of the only numbers available. But token volume tells you that activity happened. It doesn't tell you whether employees are using AI for the right things, or whether any of it is connected to work that actually matters. You can have strong token numbers, and still have an AI investment that's underdelivering. 

Eat or be eaten, enterprise edition 

The enterprises that actually figure out how to transform a huge, legacy workforce with AI are treating AI adoption as something that needs to be actively managed and updated. 

That starts with seeing how work actually happens across AI workflows. If they're logging in, using them for the right tasks, getting useful outputs, and building them into how they actually work. 

In most enterprises right now, adoption looks fine on the surface but is quietly falling apart underneath. Employees hit dead ends, find workarounds, and stop reaching for the tool entirely. And nobody knows because nobody can see it. 

AI is actually more like Human + SaaS + AI

Employees don't live inside AI tools yet. They move between AI agents and SaaS applications constantly, and real work happens across both. Looking at AI usage in isolation gives you a partial view of a workflow that is anything but partial.

Understanding productivity and adoption properly means seeing the whole picture: 

  • Where is AI genuinely embedded in how work gets done? 
  • Where are SaaS tools still carrying most of the weight? 
  • Is that balance actually shifting over time as we invest more in AI?

That last question matters more than most organizations realize. The financial case for AI has always included the assumption that it would eventually reduce reliance on a bloated SaaS stack. More AI capability, less redundant tooling, a leaner and more efficient technology spend. 

But most companies are increasing AI budgets while SaaS costs stay flat, with no real visibility into whether the tradeoff is materializing. 

So now what?

IT leaders have a real opportunity here to change how their organizations think about AI: as an adoption problem that requires ongoing attention and change. The most valuable thing IT can do right now is make sure the tools that have been bought and deployed are actually working the way they were intended to, and keep working that way as workflows evolve.

That requires a platform that gives you analytics across your full employee technology stack, SaaS and AI agents together, so you can see where engagement is strong, where friction is building, and where employees are struggling in ways that nobody has flagged yet. 

It means being able to act on that in real time, pushing guidance directly into workflows so employees get help exactly when and where they need it, without waiting on a training program or a change management initiative. And it means walking into board conversations with something more useful than token counts: adoption curves, friction reductions, SaaS consolidation, productivity shifts. And real recommendations on what to do next. 

Want to see what this looks like in practice? Explore how Pendo for IT helps enterprise teams go from measuring AI to driving real adoption across their workforce.