Use the assistant as a critical analyst

The workspace assistant understands the six categories, the external/internal boundary, PESTEL’s relationship to SWOT and strategic planning, and the importance of geography and horizon. Ask it to propose an initial factor set, classify a factor, challenge a category, find gaps or duplicates, explain significance, identify opportunities and threats, explore dependencies and second-order effects, challenge assumptions, suggest evidence to collect, compare interpretations, prioritize discussion, or draft response options.

Structured changes appear as a pending proposal. Inspect the counts, warnings, factor text, evidence status, judgments, and response links before selecting Apply. Applying uses the same normalizer, entitlement limits, exact-revision checks, and persistence path as manual editing. A proposal generated from an older revision is rejected rather than silently overwriting newer work. Any factor or evidence revised by AI returns to AI-draft review status; revised responses return to proposed. Changing recommendation content clears an earlier human owner/date attestation so the new candidate cannot appear adopted.

The assistant must distinguish known user-supplied information, inference, assumption, and information requiring external verification. It cannot fabricate a fact or citation, claim it performed stakeholder consultation, mark its own evidence verified, approve a strategy, or record a human decision owner or decision date. Editing an AI draft does not make it evidence.

During Review, resolve high-severity findings, inspect high-impact low-confidence factors, check category gaps, confirm jurisdiction and date, revisit contradictory evidence, and test whether the recommendation remains sensible under a plausible alternative interpretation. Human judgment remains responsible for material classifications and the final recommendation.