Seoul, March 2016, AlphaGo's second game against Lee Sedol. On the thirty-seventh move, the machine places a stone on the fifth line, where centuries of Go tradition advised against it. Commentators first assumed a mistake. AlphaGo estimated the probability of a human playing that move at roughly one in ten thousand. Move 37 proved decisive, and it stands as the moment the machine's capability stopped being a promise.
AlphaGo's capability is in Move 37. The lasting lesson for an organization is in the move that follows, Move 38: the one that belongs to the human, who decides what to do with that capability, frames it and answers for it. That is the move the cabinet took as the name of its evaluation method, because that, not access to the model, is where the advantage is decided.
In short: Move 38 is identifiable's evaluation method, and its name comes from the AlphaGo versus Lee Sedol match in 2016. Move 37 was the machine's move, the demonstration of capability; Move 38 is the next move, the human's, who decides, frames and answers. Now that AI capability can be bought, the advantage is in that human move: define a system, measure it against the AI id framework, and make it defendable. The move is the same for a small business and a multinational; only the scope changes, around three movements: inventory, frame, designate.
Where the name Move 38 comes from
The match was played over five games, and AlphaGo won four to one. Move 37, the machine's move, is what gets remembered. What gets forgotten is that every machine move opens a question for the human sitting across the board: what do I do with it now? Lee Sedol answered once memorably, in the fourth game, with a move players nicknamed the "hand of God," the only game AlphaGo lost. The human response matters as much as the move that prompted it.
Move 38 names that response. In the match, the move that followed belonged to the human player; in an organization, it belongs to whoever deploys the AI, in full awareness. Capability was Move 37. Responsibility is Move 38.
Why Move 38 matters now
AI capability has become a commodity. The same model, through the same interface, is available to your organization and your competitor alike, often at the same price. Generic is the default state, and it creates no gap. What creates the gap sits after the machine's move: the human's decision, and the ability to answer for it.
That response is measured by three questions, always the same, that a board, a client or a regulator asks: who validates this use, where does the data go, does the result replay? An organization that can answer with evidence has played Move 38. An organization that defers to the model's output stopped at Move 37, and it finds out at the worst moment, when the question arrives.
Move 38 for a small, a mid-sized and a large business
The move does not change with the size of the organization; its scope does. The same approach scales from a team of five to a group of several thousand:
| Size | The same move, at its scale | Starting point |
|---|---|---|
| Small (5 to 50) | A handful of named uses, a master prompt, a one-page policy, a designated owner. | The AI Index diagnostic |
| Mid-sized (50 to 500) | A system inventory, the Index and the six-property profile, a review committee. | A scoped Move 38 evaluation |
| Large (500 and up) | Portfolio governance, a profile per system, board-level reporting, the designation. | Move 38 by domain, periodic re-evaluation |
A small organization has fewer systems to frame, so a faster first reading. A large one has more, but also the resources to govern them as a portfolio. None is too small to play the move, and none is large enough to skip it.
The three movements of the move
Move 38 is played in three movements, and each produces a deliverable the next one uses:
- Inventory. Map every AI system in use: intent, data touched, dependencies, oversight in place. Most organizations discover here uses no one had named.
- Frame. Evaluate each system against the six properties of the AI id framework, on evidence. Each score ties to a level of evidence, from absent to measured and improved.
- Designate. Compute the Index, produce the profile and the roadmap. At the threshold, an Index of at least 70 and no property below 60, grant the Responsible AI Organization designation. Below it, target the gaps and plan the re-evaluation.
Capability was Move 37. Responsibility is Move 38.