Engagement model
Each phase ends in something you can check. We do not start the next one until the last one holds. Durations are typical, not quoted.
Answer one question: which of your understanding is hard-won and deepening, and which is merely data you happen to hold? You get a written answer, a map of where information currently routes through people, and a measured latency for 3 real decisions. We name the layers that exist only to relay.
If the honest answer is "none of it", we say so and recommend against the rest of the programme. Publishing that answer is our credibility.
We make your own work machine-readable. The artefacts you already produce — decisions, discussions, code, designs, plans, incidents — become one continuously updated representation of what is being built, what is blocked, where resources sit and what is working.
Success is measured by a manager being able to stop relaying, not by a dashboard shipping.
We separate atomic capabilities from delivery surfaces. A capability has reliability, compliance and performance targets and no interface of its own. An interface composes capabilities and owns none of them.
Most of this work is organisational, not technical. The split fails when a product team keeps a capability as private property.
We build the composer: the system that recognises a moment in a customer's reality and assembles capabilities into a response before the customer asks. Then we instrument its failures.
Every composition it attempts and cannot complete is a named gap. That list replaces your roadmap: what customers actually need decides what gets built next, not a planning cycle.
You normalise to 3 roles and retire the permanent middle layer.
Expect this phase to break before it works. We plan for that in the contract and we tell you so in the first meeting.
What we refuse
Four weeks. One question. An answer you can act on either way.