decision sovereignty starts with the decision itself. The AI urgency is a strategic fault. Decide before you integrate.
Christophe Picou - Founder, VEIA.AI - 2026
This text is written from the ground. From the rooms where the decisions that commit organisations for five, ten, sometimes twenty years are made. From the executive committees where I watch, every week, the same sequence unfold: a structural decision has already been made elsewhere , by a technical team, by a supplier, by a market dynamic, before the governance body had time to formulate it.
By the time the file reaches the room, the dependencies are committed. The perimeter has already been drawn by others. AI governance is an act of leadership, not a technical question. And clarity precedes power.
Does the board still decide? Or does it endorse decisions manufactured elsewhere?
Artificial intelligence in the enterprise is not one software among others. It is an outsourced decision architecture: a layer that sits between the real and human judgement, and shapes that judgement probabilistically, opaquely, gradually.
Every enterprise AI system rests on a chain : a model processes information, access grants allow the model to interact with processes, data feeds and orients it, orchestration automates flows, and contracts bind the whole to one or several suppliers. Every link commits the organisation, in its data, in its decisions, in its dependencies.
This is not a speed problem. It is a sequence problem.
Decide before you integrate requires that four questions receive an explicit answer, formulated and validated by the competent body, before any irreversible commitment.
Before deployment, not after. Algorithmic optimisation without a formalised intent maximises indicators no one has verified as translating the organisation's reason for being. AI without purpose produces performance. Rarely value. Never direction.
The value of AI for an organisation resides in its proprietary data, its distinctive processes, its non-exported business logic. The algorithm itself is being commoditised fast. What feeds it, and the decisions it serves, is not.
A system whose mechanics escape the governance body's understanding is a commitment made without visibility, a decision architecture outsourced to a black box. Explainability precedes authorisation. Auditability conditions responsibility.
The ability to exit a system is built before the contract, not after. Irreversible dependency is not the result of poor negotiation. It is the predictable consequence of an integration without a doctrine of reversibility.
The organisation defines its ends and the zones of irreversibility it consciously accepts.
Fallback architectures, data portability and system observability are designed from the start.
Clauses, transparency, exit costs and reversibility conditions are negotiated before commitment.
The organisation preserves its ability to exercise its own judgement, to maintain its evaluation criteria, and to distinguish algorithmic performance from business truth.
Adoptable by the governance body. Revised annually. The Charter does not require completion before the first trial. It requires presence before the first irreversible commitment.
The governance body formulates the purposes for which artificial intelligence is authorised to operate in the organisation, and those from which it is explicitly barred. This formulation precedes any contractual or operational commitment.
The governance body defines the perimeter of data, processes and business logic that remain under the sovereign control of the organisation. No asset in this perimeter transits to models, infrastructures or external suppliers without prior formal validation.
Every deployment of artificial intelligence with a significant impact on operations, decisions or stakeholders designates a named human accountable, with an explicit mandate. The opacity of a system does not suspend this responsibility. It intensifies it.
The governance body sets the standards of audit, traceability and validation required before any deployment authorisation. These standards define what must be demonstrated, not what is promised. They condition access to resources, to data and to operational authorisations.
The governance body defines what can be automated, what stays under direct human decision, and the conditions under which this line can be revised. These thresholds are documented, communicated and reviewed at regular intervals.
The governance body maintains a mapping of critical suppliers and systems, with contractual reversibility clauses and an operational continuity plan in case of rupture. The continuity plan includes AI incident scenarios: unavailability, compromise, behavioural drift.
The governance body defines the procedure that applies when an AI decision must be made under time pressure: which minimum process, which guardrails, which traceability. Operational pressure is the most revealing test of a doctrine. It does not suspend it.
Seven articles, short format to sign in a board of directors or in an executive committee. Formal commitment to decision sovereignty in the face of AI. The Charter is provided after an exchange.
Request the CharterThe difference between organisations that govern this transition and those that suffer it does not lie in the technology they chose, nor in the budget they devoted to it, nor in the speed at which they acted. It rests on one thing: the question they asked before choosing.