Use this when
You have a recurring reporting, forecasting, measurement, planning, or executive-readout workflow and want AI to improve the path from evidence to decision without hiding accountability.
1 · Decision
Frame the decision
- What business decision are we improving?
- Who owns it?
- What is the cost of delay, inconsistency, or poor judgment?
- What would a better decision change?
2 · Current state
Map the workflow
- Inputs and source systems
- Analytical steps
- Context inputs
- Handoffs and approvals
- Decision moments and bottlenecks
3 · Human
Keep judgment owned
- Final recommendation and accountability
- Trade-offs, ethics, people, and brand risk
- Ambiguous calls with incomplete context
- Stakeholder framing and decision ownership
4 · Tool
Make repeatable work deterministic
- Calculations, joins, and reconciliations
- Validation and source retrieval
- Scheduled checks and rule-based alerts
- Anything that should produce the same answer every time
5 · AI
Use AI for synthesis
- Candidate narratives and executive language
- Pattern-finding and contradiction detection
- Scenario drafting and implication framing
- Translation from technical analysis into commercial context
6 · Evidence
Design the evidence layer
- Source-of-truth data and metric definitions
- Business context and caveats
- Reliability checks and sensitivity rules
- Named evidence owner
7 · Trace
Capture the reasoning
- Question, evidence, assumptions, and confidence
- Recommendation, decision owner, and final decision
- Human corrections or overrides
- Outcome after the decision
8 · Trust
Review before scaling
- Source traceability
- QA checklist and review queue
- Escalation and rollback path
- Named accountable human owner
Output
Finish with a simple allocation table: what stays human-owned, what becomes deterministic, what AI supports, what must be verified, and how the workflow learns from corrections and outcomes.