ChatGPT
Perplexity AITL;DR
- Reach and engagement measure different things. Reach shows how widely a tool spread. Engagement shows how deeply it gets used once it arrives.
- You can get reach from licenses, logins, and billing records. Engagement requires measuring real interaction on the device itself.
- Most SAM tools report total usage, meaning session time and whether an application was opened. That counts the hours an app sat idle in a background tab.
- Xensam measures active usage down to the minute, at application and module level, which is what makes engagement figures possible in the first place.
Reach is the easy number
In our H1 2026 Enterprise AI Adoption Report, reach measures the share of organizations where more than 5% of users tried a given application at least once during the month. It answers a simple question: has this tool arrived?
You can approximate that answer from several places you already have. License records tell you what was bought. SSO logs tell you who signed in. Expense reports tell you what someone put on a card. None of it is perfect, but reach is the number most organizations can produce today with the systems already in place.
Which is exactly why it’s the number most vendors lead with. Reach is the easiest thing in the world to grow and the easiest thing in the world to report.
Engagement is the one that takes work
Engagement in the report measures the percentage of adopting users who used an application for at least one hour during the same month. That’s a different question entirely: once this tool arrived, did anyone keep opening it?
To answer that, something has to be watching real interaction. Not a license record, not a login event. The data behind the report comes from the Xensam Inventory Agent running on corporate endpoints, capturing when someone was working inside an application rather than when the application happened to be running.
That distinction is the entire reason the report exists. Without it, Google’s reach growing from 4.8% to 18.0% this year looks like a strong result. With it, you also see engagement falling to 4.7%, and the story changes completely.
What most SAM tools show you instead
Most SAM platforms can tell you what’s installed. The better ones go further and track total usage: whether an application was opened, and for how long.
Here’s the problem with that. A browser tab left open across a working day logs as eight hours of usage. A specialist tool someone opens for four focused minutes logs as four minutes. On a total usage report, the abandoned tab wins. If you’re deciding which licenses to renew, downgrade, or reclaim, you’re making that call on numbers that reward the wrong behavior.
Active usage tracking measures real interaction down to the minute. It’s a market-leading capability, and it goes deeper than the application itself. Active usage is tracked for individual modules and add-ons, so you can see whether someone used the specific features that justify an advanced license tier, rather than just whether they opened the product at all.
That’s the gap between knowing software exists and knowing it’s used. It’s also, in practice, the gap between an estimated license position and an accurate one.
The tools you’d never see otherwise
There’s a second thing this kind of measurement picks up: software that entered the organization without asking anyone.
The report tracked three small vendors worth watching for exactly this reason. Anysphere’s Cursor 3 went from no measurable reach in March to 4 in April and May, then 5 in June, the pattern you’d expect from an AI coding tool spreading through engineering teams first. Perplexity rose from 1 in January to 3 from March onward and held there, which reads like a research tool finding a niche rather than a broad rollout. AI Sweden’s Svea GPT appeared for the first time in June.
None of these would show up in a procurement record, because in most organizations they haven’t reached one yet. They show up on endpoints, because that’s where the people using them are. Xensam recognizes 135,000+ SaaS applications and 25,000 AI tools, which is what makes small early signals visible before they turn into a governance problem.
What this means for your software estate
The reach and engagement split isn’t only a way to rank AI vendors. It’s the same question you face every renewal cycle, on every line of your software budget.
Take Microsoft. In the report, Microsoft 365 Copilot Desktop finished the period at 98.0% engagement. Microsoft Copilot [Web] finished at 32.0%. Same company, same brand, two very different cases for what you’re paying for. A tool with high reach and low engagement is a renewal you should be questioning. A tool with lower reach and high engagement might be one you should be expanding.
You can’t tell those apart from a license report. You can tell them apart from usage data that measures what people did, minute by minute, rather than how long something was left running.
That’s the capability behind every number in the report. It’s also the capability behind the recommendations Xensam customers see in their own environments: which licenses to cancel, which to downgrade, which to reassign, and what each of those actions is worth.
See the full reach and engagement breakdown in the Xensam Insights Enterprise AI Adoption Report, H1 2026 below.