Observe what happens, not what the process document claims.
Interviews and operating artifacts reveal where people compensate for unclear work.
Approach
Diagnose the workflow before recommending automation.
The work starts with evidence. It stays useful by making assumptions, decisions, owners, and boundaries visible.
Keep observation separate from interpretation.
Each finding identifies what was seen, what it may mean, and how confident the conclusion should be.
Name the person responsible for the tradeoff.
AI can summarize competing views. The business still needs a human decision-maker.
Let unknown remain unknown.
Missing ownership, timing, or evidence becomes an open question rather than a confident invention.
Decision path
Observation → interpretation → recommendation → decision.
Collapsing those four stages makes a recommendation sound more certain than the evidence allows. Northstar preserves the path so the client can inspect and improve it.
01Observation: what changed or where did the work stall?
02Interpretation: what might explain the pattern?
03Recommendation: what action fits the evidence and constraints?
04Decision: who chose the tradeoff and when should it be reviewed?
Automation rule
Use tools to remove friction, not to avoid understanding it.
Software implementation is a separate scope because selecting and configuring a tool is not the same as diagnosing the operating problem. Northstar recommends automation only after the bottleneck, owner, and expected change are explicit.