Replace chaos from “what can we do?” with structure: Cultivate → Evaluate → Select → Learn, running at org, team, and individual levels.
Discipline novelty: interesting, valuable, and marketable are not the same. Treat “this is really interesting” as a yellow flag.
Strategy lens: magic swords (customer-visible AI) vs using magic to make excellent ordinary swords (invisible operational AI). Know your primary differentiation.
Funnel participation borrows from Toyota: respect for people; continual improvement needs contributions from people closest to the work.
Feasibility is the guardrail; speed to meaningful failure beats slow failure. Measure how fast teams recognize and act on red flags.
Close the loop: document post-mortems and feed lessons into the next Cultivate/Evaluate cycle.
Principles from the chapter
“Interesting is only loosely correlated with valuable, and valuable is only moderately correlated with marketable.”
Continual improvement requires full participation and “respect for people” and some of the greatest opportunities are invisible to management.
Your underlying purpose is to create and capture value through improved decisions and user experiences.
Feasibility is the guardrail that keeps your ambitious AI initiatives grounded in reality.
In AI projects, failing slowly is costly, but failing quickly can become your team’s secret superpower and distinct competitive advantage.
Read the chapter for…
The executive-room opening, detailed evaluation dimensions, contribution mechanics, and full post-mortem checklist for the Learn phase.