How to Measure AI Adoption Across a Team

Measure AI adoption in four layers: access, use, depth and trust. Access asks whether people work on company accounts. Use asks how many use AI weekly. Depth asks how many defined workflows run on it each week. Trust asks how much AI output is used with light edits. Log-in counts alone tell you almost nothing.

What does AI adoption actually mean?

AI adoption means people use AI as part of how their work is done, on company accounts, in defined workflows, with output they trust enough to use. A licence is access, not adoption. A team has adopted AI when its recurring jobs run on it every week and the results are good enough to keep.

It helps to think in layers, because each one can stall on its own:

Most teams measure only the second layer, and only roughly. However, the layers above and below it decide whether AI pays. A team with high use and low trust is busy, not productive.

Why are log-ins and licence counts misleading?

Log-ins measure curiosity, and licence counts measure purchasing. Neither shows whether AI is doing useful work. Worse, official numbers often miss most real use, because staff use personal tools the company never sees. Adoption measured only through company systems can look low while AI use is everywhere.

The gap between official and real use is large. According to MIT NANDA's research, as reported by THE Journal, only 40% of companies said they had bought an official LLM subscription. Yet workers from over 90% of companies reported regular use of personal AI tools for work. Those are company-level shares, not shares of workers, but the direction is clear.

Microsoft and LinkedIn found the same pattern from the other side: 78% of AI users bring their own AI tools to work. So the first adoption metric is not how many people use AI. It is how many use it where the business can see it and protect its data. Moving people onto business plans comes before measuring anything else.

The guide to rolling out AI to employees covers why licences alone stall.

Which AI adoption metrics should you track?

Track one or two metrics per layer: the share of staff on company accounts, the share using AI weekly, the number of workflows running weekly per team, and the share of AI drafts used with light edits. Add a quarterly confidence question. Five numbers, reviewed monthly, are enough for most teams of 10 to 200.

Workflows per team is the metric that best predicts return, because a workflow has an owner, a baseline and an output. The light-edit rate is its partner. If drafts keep being rewritten, the problem is usually missing context rather than reluctant people. A related measure, time to first draft, shows whether those workflows are getting faster.

How do you collect adoption data without watching people?

Collect it at team level, from systems that already record it, and tell people what is measured and why. Admin consoles give usage counts. Workflow logs show what runs. A three-question pulse survey covers confidence and personal tools. Never publish individual rankings: they reward performance for the dashboard and drive real use underground.

Three rules keep the data honest:

1. Aggregate by team. Report teams of five or more, so no individual is singled out.
2. Ask about personal tools without blame. People answer honestly when the purpose is moving them onto safer plans.
3. Measure work, not keystrokes. Prompts per day is noise. Workflows running and drafts used are signal.

The survey question about personal tools matters most. If people fear trouble, they will under-report, and the business loses sight of where its data is going.

What does uneven adoption tell you?

Uneven adoption shows where to act. A team with low use usually needs training on its own tasks. Teams with high use and low trust need better context. High trust with no defined workflows means enthusiasts are carrying the load. Compare teams, roles and seniority, not individuals.

Frontline and specialist roles often lag. According to BCG's 2025 AI at Work survey, regular use among frontline employees stalled at 51%. Only about a quarter of frontline employees said they got leadership support.

Training and leadership both move the numbers. BCG found regular use was far higher among people with at least five hours of training. It also found the share of employees who feel positive about generative AI rises from 15% to 55% with strong leadership support. So a lagging team is rarely a lazy team. More often, nobody showed it how AI fits its own work.

How often should you measure AI adoption?

Measure access, use and trust monthly, because they move quickly and point to fixes. Review workflows per team and confidence quarterly, alongside business results. Each review should end in an action for any team that stalled: training, better context, a new workflow or a move off personal tools.

Adoption is a leading measure, not the return itself. It tells you whether the conditions for a return exist. The return shows up later, in pipeline, cost per lead and time to approved work. The guide to measuring AI ROI connects the two.

Skein moves people using personal AI accounts with company data onto business plans first, then tracks adoption alongside measures such as pipeline and time to first draft; the outcomes page lists what is measured. That suits a leadership team that has bought seats and wants to know whether they are being used well. A team of three using AI informally can simply ask each other once a month.

Frequently asked questions

What is AI adoption in a business?

AI adoption is AI built into how work gets done: people on company accounts, using AI weekly, with recurring jobs running on it and output trusted enough to use with light edits. Buying licences or running a training day is the start of adoption, not evidence of it.

How do you measure AI adoption in a company?

Track a few metrics across four layers: the share of staff on company accounts, the share using AI weekly, the number of workflows running weekly per team, and the share of AI drafts used with light edits. Add a quarterly confidence question and review the numbers monthly by team.

What is a good AI adoption rate?

There is no reliable standard, because surveys define use differently. A better test is internal: is every team on company accounts, is weekly use rising, does each team have at least one workflow running, and are most drafts used with light edits? Track your own trend rather than chasing a benchmark.

How do you increase AI adoption across a team?

Train people on their own tasks rather than in general, give AI the business context it needs so drafts are usable, build one defined workflow per team, and make sure leaders visibly use and support it. Then measure again and focus on the teams that are still lagging.

Should you track individual employees' AI usage?

Track by team, not by person. Individual rankings reward gaming the numbers and push real use onto personal tools nobody can see. Aggregate usage from admin consoles, report teams of five or more, and tell staff exactly what is measured and why.

Related reading