The first number anyone reports after a Claude rollout is the share of licences in use. It is easy to get, easy to chart, and tells you almost nothing about whether the rollout worked.
Active-user percentage measures who opened the tool. A team at thirty percent usage with two redesigned workflows is further along than one at eighty percent where everyone drafts emails faster. If the measure cannot tell those two apart, it is the wrong measure.
What not to lead with
Seat utilisation, message volume, prompts per user and satisfaction scores from training sessions all have a place as health signals. None of them answers the question a CFO will ask, which is what the investment returned. Leading with them invites the honest answer "people say it helps", which does not fund a second phase.
The measures that matter
Workflows changed, and what each is worth
Count the workflows that were actually redesigned around Claude — not tasks individuals speed up privately, but processes that now run differently. For each, estimate the value: how often it runs, how much time it took before, how much it takes now, and what a mistake costs.
Value against a baseline
None of this works without a before. Measure the workflow before you change it — cycle time, volume, error or rework rate — so the after is a comparison rather than an impression. A single defensible before-and-after does more to unlock the next phase than any strategy document.
Oversight health
For workflows in production: accuracy against the evaluation set, exception queue depth, override rate (how often a person disagrees with the proposal), and cost against budget. These tell you whether a workflow is still working, not just whether it once did.
For engineering teams
How many shared skills and subagents exist and how often they run, whether hooks enforce the standard, review queue depth and rework on agent-assisted changes, token spend per team, and time to first useful commit for a new joiner. The last one is the most honest test of whether a team standard exists at all.
A simple rule
If a metric would look the same whether or not any workflow changed, it is a usage metric, not an adoption metric.
Set the baseline before rollout
The most common measurement mistake is timing. Teams decide what to measure after the rollout, by which point the before has been lost. Pick the two or three workflows you expect to change first, and measure them now.
Why Claude adoption stalls after training
The full Claude Enterprise rollout guide
Where Focus20 fits
Our adoption consulting starts by setting the baseline, then ranks workflows by what they would return, so every build that follows has a number against it from the start.