AI at Work · Source published 10 July 2025
Research signal
METR’s 2025 developer experiment provides a bounded example of why a productivity claim needs measurement and context.
Source: METR — experienced developer productivity study · 10 July 2025.
Event Systematic perspective
A useful pilot scorecard should be understandable to the manager who owns the process. Include completion time, first-pass acceptance, correction effort and user experience. Add a clear description of the tasks and people involved so the results are not presented as a universal business forecast.
For an HR reporting pilot, count the time spent preparing the report and the time another person spends correcting it. Note whether the result answers the original question. If one stage becomes faster while another becomes slower, redesign the handoff before extending the tool to the whole department.
Put it into practice
- Agree the definition of completed work before collecting results.
- Compare similar tasks and record reasons for unusual cases rather than silently removing them.
- Make the expansion decision conditional on an acceptable combination of quality, effort and usability.
Explore employer services, career development and the AI Academy. Discuss your workforce priorities with Event Systematic →
