AI ADOPTION · 5 MIN READ

Please Stop Measuring AI Adoption by Counting Logins.

Opening an AI tool is not a business outcome. Neither is writing a prompt. Ten thousand active users can still add up to remarkably little measurable value.

By Gabby White, AI Adoption Specialist at INTAI · Published 9 June 2026

Start Reading

A Login Is an Event, Not an Outcome

Gallup’s Q4 2025 workplace panel found that 12% of US employees now use AI daily, up from 10% the previous quarter, with frequent use, a few times a week or more, reaching 26%. Those are genuinely useful numbers for tracking whether people are picking the tool up at all.

What they cannot tell you is whether any of that activity is changing how the work actually gets done. And that is precisely the number most AI adoption dashboards are missing, because it is the number vendors are least equipped to hand you, logins, active users and prompts issued are easy to instrument from inside a product. Time saved, error rates and workflow redesign are not, because they require someone on the customer side to actually go and measure them.

The Question That Replaced “Are We Using AI?”

By most accounts, “are we using AI” stopped being a useful board-level question around 2024. The 2026 version of the question is considerably harder to dodge: have we redesigned a workflow around AI, and can we prove it?

That distinction matters because the two questions can produce completely different answers inside the same organisation. It is entirely possible to have ten thousand active licensed users, healthy weekly login figures, and remarkably little to show for it in terms of hours saved, quality improved or cost reduced, because using a tool and changing a workflow around a tool are not the same behaviour, and only one of them shows up on the balance sheet.

Central Insight

A useful mental test for any AI adoption metric on your dashboard: if this number went up by 50% next quarter, would anyone outside the AI programme team actually notice a difference in how the business runs? If the honest answer is no, it is an engagement metric, not an impact metric, worth tracking, but not worth reporting to the board as evidence of value.

What to Track Instead

Recent CIO guidance on this groups useful metrics into four categories worth borrowing directly: activity (weekly active users, adoption velocity), depth (engagement depth, feature use, the small group of power users who account for a disproportionate share of value), breadth (which teams and functions are actually using it, and which are not), and newer AI-specific patterns such as the ratio of agentic to assistant-style use, and the shadow AI rate, how much AI activity is happening outside sanctioned tools altogether.

None of those four categories is a login count. All four require someone to define, ahead of time, what a redesigned workflow would actually look like for a specific team, and then check, deliberately, whether that redesign happened. It is more work than pulling a usage report. It is also the only version of the report that tells you anything about value.

If this raised more questions than it answered, that is usually the sign of a conversation worth having. Book a Consultation

Sources

Larridin, Your AI Vendor Shows You Logins. Here’s What You Actually Need to See

Worklytics, AI Adoption Benchmarks 2025: Employee Usage Statistics