The constraint on AI delivery is promotion, not code

AI can generate more change than most organisations can safely absorb. The constraint is not code. It is promotion.

This is the mental model I use to keep the conversation honest.

Think of the code value stream as a Petition moving through a system. As it moves, it collects certified changes.

A Petition is a formal request for change. It means: make the world different in this way.

To fulfil a Petition, the organisation must update a set of work products. Not just code and infrastructure, but also the things that carry intent: requirements, scenarios, models, diagrams, decisions, runbooks, control evidence.

Each work product type has its own Sequence. A Sequence is a repeatable path that turns Petition intent into a candidate change, and produces evidence about that change.

When a sequence runs, it produces an Attestation. An attestation is an evidence-backed statement of truth: this change is now true, and here is the proof.

Gates evaluate those attestations. They decide whether the Petition can advance, or must stop for fixes, escalation, or rejection.

So the Petition moves forward only when it has collected the required attestations from the relevant sequences.

In plain language: sequences qualify changes to work products, attestations prove what is true, gates decide what is allowed to advance, and the Petition is what gets promoted through the system.

Sequences can be accelerated with AI, but promotion still depends on evidence, not speed.

AI will increase the volume of change. If your system cannot certify and promote change with evidence, the missing system will reappear later as rework, review load, and audit pain.