An enterprise rolled an AI copilot out to 18,000 seats and watched an earlier 4,000-seat pilot collapse from 71 percent to 22 percent active use within eight weeks. Stratenity rebuilt the rollout around incentive and feedback design rather than another communication campaign. Adoption was tied to how work was actually measured, managers were held to a team adoption number, and a weekly feedback loop turned user friction into shipped fixes. Active use held above 60 percent at full scale six months in, and the copilot moved from a launch event into the operating rhythm of the business.
A pilot that peaked and then fell
The client was a global enterprise of roughly 40,000 employees that had licensed an AI copilot for 18,000 knowledge-work seats. The business case rested on a productivity uplift that only materialized if people used the tool as part of their daily work, not once a week out of curiosity. The signal that brought Stratenity in was a completed pilot. A 4,000-seat pilot had launched to enthusiasm, hit 71 percent weekly active use in its first fortnight, and then decayed to 22 percent within eight weeks. The tool was capable, the licenses were paid for, and the capacity to help was sitting idle. The tool was capable. The rollout was not, and repeating it at four times the scale would have burned the budget and the credibility of the program in one quarter.
The instinct in the room was to spend more on the thing that had already failed. The pilot had been supported by a communication campaign, launch emails, town halls, a branded intranet page, and a library of tip videos. The proposed remedy for the 18,000-seat rollout was a bigger version of the same campaign. The evidence said the opposite. Adoption had not fallen because people were unaware of the copilot. It had fallen because using it competed with how their work was measured and because early friction went nowhere. Two prior awareness pushes had already peaked and faded, so a third would have been the definition of doing the same thing and expecting a different result. Stratenity was scoped to produce the rollout operating model that would hold adoption, not another awareness push.
Design the incentives and the feedback loop, not the campaign
The first two weeks instrumented the pilot data by role and by manager rather than in aggregate. Two patterns emerged. Adoption held only in teams where a manager personally used the copilot and referenced it in how work got reviewed, and it collapsed wherever using the tool added visible time to a task that was measured on speed, because no employee volunteers to look slower on the metric that determines their review. The redesign attacked both. It tied copilot use to the metrics people were already judged on, made managers accountable for a team adoption number, and built a weekly loop that turned friction into fixes fast enough that users saw their complaints answered rather than absorbed.
| Lever | Old rollout | Redesigned rollout | Owner | Signal watched |
|---|---|---|---|---|
| Incentive | Encouragement and tips | Copilot use embedded in existing role metrics | Function leads | Weekly active use by role |
| Manager role | Optional champion | Team adoption number in manager scorecard | People operations | Manager-level adoption spread |
| Feedback | Suggestion inbox, no reply | Weekly triage to a shipped fix or a clear no | Product owner | Time from friction to resolution |
| Enablement | Generic tip videos | Role-specific use cases proven in pilot teams | Function leads | Depth of use per active user |
| Governance | Launch event, then silence | Monthly cadence retiring dead use cases | Program owner | Use cases active versus registered |
The unpopular choice was putting a team adoption number in every manager's scorecard. Managers argued it was unfair to measure them on a tool they did not choose. The program held the line, because the pilot data was unambiguous that manager behavior was the single strongest predictor of whether a team's adoption survived. A single accountable program owner held decision rights over the rollout cadence, so the incentive design could not be quietly softened function by function when it met resistance. That concentration of authority was the difference between a policy and a poster.
What the redesigned rollout delivered
- Weekly active use held above 60 percent across all 18,000 seats at the six-month mark, against a pilot that had fallen to 22 percent in eight weeks.
- The gap between the best and worst adopting functions narrowed sharply once manager scorecards carried a team adoption number, because the laggard managers finally had a reason to engage.
- Median time from a user reporting friction to a shipped fix or a clear decision fell to under two weeks, which visibly changed whether people believed reporting friction was worth the effort.
- Depth of use rose alongside breadth. Active users moved from a single novelty use case to an average of four role-specific ones, which is where the productivity case actually lived rather than in a headline active-user count.
- The productivity uplift in the original business case became measurable against a protected baseline, giving the board a defensible number rather than an anecdote.
What the engagement learned
- The pilot did not fail on awareness, so spending more on awareness would have failed again at four times the scale. The diagnosis mattered more than the effort.
- Manager behavior was the strongest lever available, and the only way to move it was to measure it. A voluntary champion program is a hope, not a mechanism.
- A feedback loop only changes behavior if users see it produce fixes. An unanswered suggestion inbox teaches people that reporting friction is pointless.
- Tying use to existing role metrics beat inventing new adoption metrics, because people optimize for the numbers they are already judged on.
- Several launch-era use cases were retired within the first quarter when usage data showed nobody used them. The monthly cadence was built to retire as well as to promote.
How to run this pattern
- Instrument any pilot by role and by manager before scaling. Aggregate adoption numbers hide the pattern you most need to see.
- Diagnose why adoption fell before spending on the fix. If the failure was not awareness, a communication campaign will not repair it.
- Make managers accountable for a team adoption number. Manager behavior is usually the strongest predictor of whether adoption survives.
- Tie tool use to the metrics people are already judged on, rather than asking them to optimize for a new one.
- Run a fast feedback loop that visibly converts friction into fixes, and retire dead use cases on a regular cadence so the program stays honest.