The TEAF loop, component by component

Eight words keep coming up in TEAF's vocabulary: Intent, Capability, Knowledge, Decision, Automation, Execution, Observation, Learning. Taken separately, these are classic architecture concepts, none of them genuinely new. What changes is that they're connected in a closed loop — and it's that closure that turns eight steps into a living system rather than one more documentation pipeline.

Intent — intent traced back to its source

Everything starts from a formalized intent: a board objective, a regulatory constraint, an opportunity spotted in the field. The difference from a classic requirements document isn't in the content, but in traceability: every decision made further down the loop must be traceable back to the intent that justifies it. An intent that serves no documented decision is a signal — either it wasn't clear, or nobody actually executed on it.

Capability and Knowledge — what gets mobilized, what's already known

Intent alone produces nothing: it mobilizes a capability — business, technical or compliance — that already exists somewhere in the organization, even imperfectly. And that capability draws on available knowledge: data, past decisions, dependencies, constraints. It's the Knowledge Backbone's role to make that knowledge queryable in one place, rather than scattered across three spreadsheets and two people's memory.

Decision — the choice that becomes a Living ADR

From the knowledge available, a choice gets made. TEAF formalizes it as a Living ADR — not a document written once and archived, but a decision that stays connected to what produced it, and that can be explicitly revised when its conditions change. This is the loop's pivot point: before, information gets gathered; after, action happens on a traced basis.

Automation and Execution — acting, under supervision

The decision translates into automated execution — a pipeline, an AI agent, a deployment — supervised by the AI Control Plane, which defines what can happen without human validation and what can't. Execution itself is the change actually produced in the information system: no longer an intent or a plan, but an observed fact.

Observation and Learning — where the loop actually closes

This is where most architecture initiatives stop, for lack of discipline: intent and decision get documented, rarely what actually happened. TEAF requires measuring the gap between the result obtained and the original intent — Observation — then extracting a lesson fed back into the next intent. Without this last step, the first seven produce a history. With it, they produce a system that improves with every cycle.

• The loop connects eight already-known steps; its value comes from their closure, not their individual novelty.
• Intent traces every decision back to its original justification.
• Decision is formalized as a Living ADR, not a static document.
• Observation and Learning close the loop — the step most often neglected, and the most decisive.