Engagement format
AI readiness assessment before the next investment
Assess the workflow, sources, data, baseline, unacceptable failures, and decision ownership before funding an AI model, integration, or pilot.
Describe the taskProblem framing
What must be resolved
A team is considering a model, integration, or further investment without a shared view of the workflow, sources, baseline, unacceptable failures, and decision owner.
Fit
Suitable for
For a team with an idea, requirements, or a prototype but no coherent view of data readiness and quality criteria.
Deliverables
What remains after the engagement
- Workflow and source map
- Risk register
- Evaluation plan
- First-experiment scope
Process
How the work proceeds
- 01Context
- 02Material review
- 03Working session
- 04Report and decision
Scope
From input to a decision
When it is needed
A diagnostic is useful before funding a model, integration, or pilot when a team knows the problem but lacks a shared definition of the decision, input material, and failure that must not pass. It is not a security assessment or certification.
Input material
The input is one workflow, sample admissible sources, a known unacceptable failure, and a named decision owner. The material may be limited and anonymised; a confidential corpus is not required for the first conversation.
What we do
The work separates workflow and data boundaries from a model promise. We establish a baseline, failure profile, access constraints, oversight point, and conditions for an honest experiment. The outcome is checked against a real task rather than a generic tool demonstration.
Artifact and outcome
The team receives a workflow and source map, risk register, evaluation plan, and bounded first-experiment scope. Each element names an owner, missing information, and a way to challenge the underlying assumption.
Acceptance criterion
Acceptance means the decision, sources, baseline, critical failures, and decision owner are stated clearly enough to choose the next step: a pilot, a reframed problem, or a stop.
Boundary of responsibility
This is a practical readiness assessment, not a certification audit, legal opinion, or implementation guarantee. If admissible sources or a failure criterion cannot be established, the appropriate result may be no AI recommendation.
Decision record
After the review, record one decision that should improve, the material admitted for inspection, an unacceptable failure, and the person who will own the next step. This is not an organisational maturity score; it prevents an AI conversation from becoming a general list of wishes.
Boundaries
Risks checked early
- Incomplete data access
- No decision owner
- Success criterion based on impression
Evidence path
Existing material that demonstrates the scope of work
FAQ
Practical questions
Do we need a working prototype?
No. The diagnostic can start from the workflow and available sources.
Does the diagnostic require handing over the entire data corpus?
No. The first stage can use a workflow description, representative anonymised examples, and an access-boundary register. If an honest risk assessment cannot be made without the full material, that becomes a documented condition for the next step rather than a promise of a result.
What can be an honest diagnostic outcome?
The outcome can be a bounded pilot, a reframed problem, or a stop. The diagnostic is meant to expose a missing baseline, unacceptable failure, unclear decision owner, or unavailable sources before a team funds a model or integration.
Scope and cost
Evidence before an estimate
I do not publish a fictional ‘from’. Scope and estimate follow a review of the workflow, material access, risk, expected artifact, and acceptance criterion.
Start by describing the workflow
The first response will identify whether this format fits and what is needed for an honest scope.