Before evaluating any AI initiative, a board must approve a framework with three definitions: the company's posture toward AI, accountability over systems, and the red lines of delegation. At Equios, that framework is called the AI Adoption Charter and is built in a half-day working session with the board.
Short answer
Before evaluating any AI initiative, a board must approve a framework with three definitions: the company's posture toward AI, accountability over systems, and the red lines of delegation. At Equios, that framework is called the AI Adoption Charter and is built in a half-day working session with the board.
When the CEO or CTO presents an AI proposal, the board conversation focuses on a familiar question: is this initiative good? Does it have ROI, is it feasible, what risks does it bring? It is a competent discussion — and still leaves unanswered the prior question: under what rules is any AI proposal acceptable in this company? The first is an operational question. The second is constitutional: it defines the space within which management can propose, and outside of which no proposal should enter discussion, no matter how profitable it appears. A board that approves five AI projects without approving a single AI policy is giving five blank signatures.
1. Posture. AI-first, AI-enabled with guardrails, or AI-cautious: all three positions are legitimate. What is illegitimate is not having decided, because then each proposal drags in a different implicit posture. The posture includes non-negotiable values: what is not sacrificed for efficiency, even when a proposal is profitable.
2. Accountability. When a system decides, someone signs. When it fails, someone answers. The question is whether those roles exist before they are needed: who signs, who owns the risk, and — the most forgotten one — who has the authority and the information to stop the system when indicators start to move.
3. Red lines. Decisions an algorithm never makes alone: strategic positioning, decisions that irreversibly affect people, crisis management. They are not delegated because they define who the company is, not what it does.
In November 2021, Zillow shut down its home-buying division (iBuying) after losses above USD 400 million caused by its pricing algorithm. The easy reading is that the model failed. The useful reading is different: the missing piece was the prior mechanism that defined who, with what information, and with what authority could stop the system when indicators began to deviate. It was not a problem of model sophistication; it was a governance problem.
Once the Charter is approved, the board conversation changes in nature: the question is no longer whether a proposal is interesting; it is whether it is coherent with the framework. A proposal that crosses a red line is rejected or sent back — even if its ROI is attractive. That is the difference between evaluating and governing. In the Equios Model, this framework is Stage 0: the half-day governance session with the board, whose deliverable is the AI Adoption Charter.
It is the framework a board approves before evaluating any AI initiative: it defines the company's posture, accountability over systems, and the red lines of delegation. It is the deliverable of Stage 0 of the Equios Model.
Because defining what risks the company accepts and under what principles it delegates authority is a fiduciary function: without that prior act, management cannot operate and risks have no owner.