Microsoft 365 Copilot study shows 68% of champions lose enthusiasm
After 8 weeks of use, initial AI champions became less optimistic in a state DOT pilot.
A longitudinal study published on arXiv examined Microsoft 365 Copilot adoption at a state Department of Transportation over an 8-week pilot. Researchers surveyed 124 employees before and after use, tracking changes in perceived usefulness, ease of use, behavioral intention, and trust. They identified three baseline user personas via k-means clustering: Skeptics, Cautiously Positive users, and Champions. The key finding: perceived usefulness dropped significantly after actual use, indicating expectation recalibration.
Persona migration was substantial. While aggregate persona counts shifted modestly, individual movement was dramatic: 40% of initial Skeptics migrated upward to Cautiously Positive, but 68% of Champions moved to less enthusiastic personas (the latter driven by declines in perceived usefulness and trust). Use cases also evolved—tasks like data analysis and chart creation declined, while communication and summarization held steady. Privacy and accuracy concerns decreased, but worries about job security and skills gaps increased. The authors recommend persona-specific training, verification routines, and trust-calibration safeguards for enterprise AI rollouts.
- Perceived usefulness declined significantly after 8 weeks of hands-on Copilot use.
- 40% of initial Skeptics migrated to the Cautiously Positive persona, while 68% of Champions moved down.
- Use cases for data, chart, and presentation tasks dropped; concerns about job security increased.
Why It Matters
Initial AI excitement often fades post-adoption; agencies need dynamic support and expectation management for long-term success.