You leave with
- Prioritised workflow opportunities
- Risk and data-readiness assessment
- A scoped pilot brief the team can own
Project
Practical
AI systems
A bounded project prescribed after Discovery when a workflow, data, or governance question deserves a focused working session. We begin with the business constraint, not a model or a feature list.
The constraint
When this discipline is useful
The team wants AI in the operating system but cannot yet name the workflow, data readiness, or human oversight the work must improve.
Discovery recommends a 1-3 day workshop or scoped project when AI is a useful lever and the scope can be made explicit.
What the work produces
Outputs and boundaries
This is not
- An AI-first strategy with no workflow in mind
- A model bake-off before the job is named
- Autonomy claims the team cannot supervise
How we apply it
Operating principles
Workflow first
Name the job, the inputs, and the exception path before choosing a model.
Keep a human route
Useful AI systems have permission, evidence, and a way to reverse a bad action.
Pilot what you can measure
A small, observed pilot beats a platform promise with no owner.
Further reading
A related practice note
Book a call before you add more work
Bring the constraint. We will say whether this discipline belongs in the next engagement, or whether to pause.