Daily brief · 22 September 2026
The automation that created work
A system that appears to succeed while someone downstream fixes its output every morning is the most expensive failure mode there is.
What happened
A client’s invoice-matching automation reported a 94% success rate. Their finance team had quietly added a daily hour to correct it.
Why it matters
The automation measured what it processed, not what it got right. Nobody was measuring the correction work because it happened in a different team.
An automation that appears to succeed while a human repairs its output is worse than no automation, because now nobody is looking.
What to learn
Four numbers from day one, or you cannot tell working from theatre:
- How many ran clean
- How many were handed back to a human
- How many failed outright
- How long each took
What to try
Ask the team downstream of your most successful automation what they do with its output. Ask them directly, not their manager.
Worth stealing
Design the hand-back path before the happy path. A system that stops cleanly beats one that guesses confidently, and it is cheaper to build.