Three distinct offers
- Consulting: diagnose the situation, frame options, quantify trade-offs and recommend a decision. The client may build internally.
- Implementation: build an agreed solution with explicit acceptance criteria, integrations and handover.
- Managed improvement: monitor a working system, review changes, run evaluations and improve a defined workflow over time.
Northstar decision
Northstar has no clean data inventory and no agreed definition of a qualified sales lead. Selling an autonomous follow-up agent today would be irresponsible. A short assessment is the correct first offer because it can establish what exists and whether a safe pilot is worthwhile.
Scope by decision gate
Every engagement should end in a client decision. Examples: approve a paid assessment, approve a contained prototype, pause because the baseline is too weak, or select a non-AI process change. A decision gate protects both the consultant and the client from endless discovery.
Initial assessment exit criteria - One process map accepted by its owner - One baseline metric with a source - Data classification and access constraints recorded - At least one option compared with a non-AI alternative - A written go / change / stop decision
Practical mission
For Northstar, write one sentence explaining why an assessment is a better first engagement than an implementation. Then write the condition that would justify moving to a prototype.
Common failure
Do not call a discovery chat an assessment if it produces no evidence, scope or decision. A friendly conversation is useful; it is not yet a professional deliverable.