The Universal Research Framework: SMART-AI Problems, the Research Tree, Three Source Layers, RECAP Evaluation, the Insight Pyramid, and Packaging

SMART-AI: Specific, Measurable, AI-aligned (which AI capability fits), Relevant, Time-bound. Examples: enterprise CRM pain points from 1,000+ reviews by Feb 28; three content themes driving 25%+ engagement among millennials in tech; top five predictors of employee flight risk from two years of exit and engagement data (some data sits behind paywalls). Research tree: root = core question; branches = themes; leaves = specific answerable sub-questions; answer leaves, consolidate upward. 'How can Outskill expand into the Asian market': market analysis (size, growth, segments), competition (players, strategies, gaps), regulation (compliance, entry barriers), cultural factors (online/offline/hybrid). Finance tree 'is product X viable' with branches run in parallel.

Layers: primary (your interviews, metrics, PDFs - upload them), secondary (published reports), AI-generated synthesis across both; triangulate your sales data against industry reports and let AI find correlations. Evaluate sources with RECAP (the CRAAP test): Relevance, Currency, Authority, Accuracy, Purpose - and verify the AI represents the source accurately. Red flags: 'studies show' without citation, pre-2023 statistics in fast fields, very old research, single-source claims.

Data (10,000 tickets) -> information (30% rise in password resets) -> knowledge (correlates with an app update) -> wisdom (biometric login saves $200k a year); each level uses a different AI capability (aggregation, categorisation, pattern recognition, prediction). ACE: Aggregate by running similar queries across platforms (openrouter.ai/models shows which model is most used per use case); Connect trends confirmed in three independent sessions and attend to contradictions; Elevate from interesting to actionable with a projected impact. Packaging: executive insight slide (one statement plus data), strategic memo (bottom line up front), comparison table - matched to visual, analytical or decision-making readers. Command centre: project overview, tree visualiser, source library with quality ratings, prompt bank. Four weeks: build the template and a SMART-AI project; run one question across tools; a full cycle packaged; reflect. Pitfalls: vague questions, researching everything, blind trust, paralysis, poor packaging.