Learn and improve
Near-miss reporting + a Lessons Learned Center
Log the close calls, not just the failures
Capture the times AI almost caused a bad outcome, not only the times it did, so the team learns from close calls.
Borrowed from the near-miss reporting system and the Lessons Learned Center used by wildland firefighters in the United States. Read the original source.
Changes a rule or role · Minutes to start · Anyone on the team can start it
Try this first
Create one shared, blame-free place to write down what almost happened and what caught it.
Use it when
Only actual failures get surfaced, so warnings are missed.
Skip it when
It needs a no-blame culture. Where reporting a near-miss makes someone look careless, nobody will report one. Fix that first.
How to introduce it
Set up a simple way to record and share near-misses: the times the model almost caused a bad outcome and someone caught it. Firefighting services document near-misses as carefully as accidents. Most teams only learn from failures that already happened, which means paying full price for every lesson.
How to show up
Make reporting fast, easy, and blame-free, because people only admit the times they nearly slipped when it is safe. Then close the loop. Lessons that never circulate kill the reporting.
How long it takes
Little setup for the log. Ongoing effort to keep it used and to share the lessons.
What makes it hard
Nothing went wrong, so reporting feels like confessing for no reason. Make the value of the lesson visible. And a log nobody reads is dead weight, so the work is turning entries into changed practice.
What it looks like when it's working
People log the close calls, patterns emerge, and practice changes before a near-miss becomes a failure. An empty or unread log means either it is not safe to report or the lessons are not circulating.
How long until it sticks
Expect a few weeks of use before reporting near-misses feels safe and routine.
How you know it stuck
People log a near-miss without being asked and the team treats it as a free lesson.
The idea behind it
A near-miss is a failure that has not happened yet, and it is the cheapest lesson you will ever get.
Where it comes from
Wildland firefighting. The near-miss log, aimed at "the AI almost caused a bad outcome." The community documents and shares near-misses, not just fatalities, to learn.
Also fits: Warning signs get explained away
Home problem: Failures get buried or blamed