‹ Culture 3.14 · What we ask ourselves
Question · Operations
How do you design an exception that no rule had anticipated?
The unforeseen cannot be enumerated, but you can decide in advance what happens to it.
Why it matters
Every system in operation meets cases nobody modelled. If there is no planned route for the unplanned, the process stops silently or improvises, and both are worse than stopping on purpose.
In systems with agents, improvising is especially easy, because the model always finds something reasonable to do.
What we know so far
Treating the exception as a first-class state, with an owner, a context and a deadline, works better than treating it as an error. What reaches a person has to arrive with what was needed to decide, not with a dump.
Every resolved exception is material for the next version of the rules: that is where you really learn how the process works.
What remains open
It is still open how many exceptions a team can absorb before the system stops saving work and starts creating it.
And how to tell a new exception from a recurring one that nobody has yet turned into a rule.