Diagnostic
A bounded read of one problem. You get a cause and a next step, including the step of doing nothing.
Diagnosis, then design.
We start with the decision that is weak. The artefact is a system your team can run, not a deck that expires on Friday.
- One question in scope
- Evidence you can point at
- Handover included
The order of the work.
Skipping the first step is how engagements become staff augmentation.
Frame
Who decides, what bad looks like, and what is out of scope.
Read
Pipelines, tests, incidents, and the documents people actually use.
Design
Layers, gates, data, and the risks left open.
Prove
A thin slice that fails for the right reason, in the path of a real change.
Leave
The team can explain the system without us in the room.
Three shapes. No public rate card.
Time and price depend on access and the question. We will not invent either here.
Architecture
A design for test strategy, pipeline signal, or an AI workflow, written so another engineer can disagree with it.
Implementation
Building with your team after the design is accepted. Not before.
Principles we will argue for.
If a request fights these, we will say so.
Evidence over vocabulary
A new tool name is not progress. A check that fails for a known reason is.
The team remains the owner
We do not become a permanent gate that only we can open.
AI assists. People decide.
Especially when the action is hard to undo.
Start with the question.
If we are the wrong shape of firm, we would rather say that in the first conversation.