AI capability management
Learn once.
Every team in a company is introducing, adopting and using AI. Each one runs into the same scarce things: specialist time, compute and vendor budget. When should the company learn a lesson once and share it, and when should each team learn its own?
Share what recurs. Keep local what differs. Make telling the truth pay.
The idea in motion
Watch it in 92 seconds
Film · 92 seconds
Sharing wins from three teams, not from two.
The whole argument with Lesson 1’s numbers: 40 for a team alone, 10 to adapt shared work, 80 to build the shared route.
The idea in one picture
Now play with it
The shared route costs something up front: building it, coordinating, waiting in its queue. Each team that joins saves what it would have spent alone, minus the work of adapting the shared result. Move the sliders and watch the break-even move.
Research · Lesson 1
Sharing wins once enough teams need the same thing, and the rework left to each team is small.
teams
units
Shown at Teams that need the same thing: 4 teams; Rework each team still does on its own: 10 units. Turn on JavaScript to move the values.
Above 2.7 teams, sharing wins: 4 teams cost 120 units together against 160 separately, a saving of 40.
- Cost alone
- 160
- Cost shared
- 120
- Break-even
- 2.7 teams
Invented inputs, the same as Exhibit 1 and Lesson 1: 40 units for a team alone, 80 units of overhead for the shared route. The cost engine is the one tested against the Lean result in Lesson 1.
Where this stands
The models are checked. Their effect on a real company is untested.
Lean, a program that checks every step of a proof, confirms that each conclusion follows from its assumptions. It cannot show that the assumptions describe your company, and no study has yet measured what this design does in one. The company in the examples is fictional, and every target is a proposal.
Three ways in, one argument
Read the proof, the translation or the judgment.
01 · Research · 10 lessons
Why it works, when it does
Ten short lessons: when sharing wins, when it loses, and why teams behave as they do. Every model result is checked in Lean.
02 · Business · one article
Stop Paying Twice for the Same AI Lesson
The same results for people who run companies: exhibits, public examples and a list for Monday. No new assumptions.
03 · Management · a book in 8 chapters
Learn Once
The author’s judgment on building an AI capability that compounds, without building a bureaucracy. Clearly marked as opinion.