FDE Times: Eric Ries’s The Lean Startup — write the pilot’s success and kill thresholds before any data arrives
A 6 November FDE Times books review reframes AI pilots as Build-Measure-Learn experiments: progress is validated learning, the first build only needs to measure one assumption, and pivot/persevere/scale/kill only work when thresholds are emailed to the customer before results appear.
FDE Times’s 6 November review of Eric Ries’s The Lean Startup (2011) maps the Build-Measure-Learn loop onto AI pilots at customer sites. The piece argues that lists of shipped integrations are not progress; a validated-learning report is — for example, that an agent correctly handled 120 of 500 support tickets without human intervention.
The practical habit is to treat the first build as the smallest thing that can measure the riskiest end-user assumption (often “will staff trust the output enough to stop re-checking it?”), not a scaled-down final product. Every loop then ends in a decision: scale, pivot, persevere, or kill.
Drawing on Tristan Kromer’s four-outcome framing, the guide’s worked example sets a 40% self-handle threshold to expand and a 20% floor to stop; a mid-range 24% result forces a documented pivot (narrow ticket types) instead of quietly moving the bar. For FDEs, the CV-ready skill is stating those numbers and the time window in writing before anyone sees the data.
Source: fdetimes.net