tech
Why Some AI Workflows Stick—And Others Don’t
Technology brought knowledge work into this world. Can AI take it out?
TL;DR
- Many AI workflows are abandoned because they are built for an idealized user or solve non-existent problems.
- Successful AI workflows provide immediate rewards and solve real, existing problems for the user.
- Habit formation for AI workflows takes longer than the commonly cited 21 days, with medians around two months.
- Symbolic self-completion explains the tendency to adopt tools that represent an aspirational identity rather than current reality.
- Workflows that require excessive maintenance or fail to solve recurring problems are often discarded.
- The author developed a system (Agent Ops) to audit AI systems, categorizing failures and setting rules for automation adoption.
- Key criteria for keeping an AI workflow include automatic running, minimal output review time, and addressing recurring problems.