Each agent has one function, a JSON schema for its output and an explicit protocol stating what it receives and what it passes on. The Critic never optimizes and the Optimizer never criticizes, because when you wear all the hats at once the dominant one always wins.
| Hat | AITS agent |
|---|---|
| White | Analytical |
| Red | Emotional-Intuitive |
| Black | Critical-Validator |
| Yellow | Optimizer |
| Green | Creative-Generative |
| Blue | Meta-Orchestrator |
| New |
|
Color: BlueOrigin: Blue hatTier: deep
Plans the waves, submits every output to the engine and closes the decision
Reads the playbook and the pattern library, chooses the agents, presents gates to the person and proposes the synthesis, which the engine accepts only when every requirement is met.
Color: WhiteOrigin: White hatTier: standard
Facts, metrics, verifiable sources, explicit gaps
Every source has a reliability level, every unsourced claim is labeled as a hypothesis, and every gap declares its impact, the cost to fill it and the agents it blocks.
Color: RedOrigin: Red hatTier: standard
Stakeholder map with deep drivers, emotional timeline, asymmetries
Separates the surface emotion from the deep driver (identity, status, autonomy, competence, belonging, safety, fairness, legacy) and tells when the window to intervene is open or closed.
Color: BlackOrigin: Black hatTier: deep
Premortem, risk map across 6 categories, fallacies
Every risk has probability and impact from 1 to 5, severity up to 25, a guardrail with an owner and the residual severity after mitigation. Runs several times independently.
Color: YellowOrigin: Yellow hatTier: standard
Business case, value sequencing, quick wins, opportunity cost
Always compares the do-nothing option too, separates results within 30 days from structural ones and states who gains and who pays.
Color: GreenOrigin: Green hatTier: standard
Options with novelty scoring, cross-domain analogies, micro-tests
Reframes the problem, dissolves false binaries and proposes for each option a small test with cost, duration and success criterion.
Color: PurpleOrigin: NewTier: deep
Seven ethical dimensions, red lines, distributional analysis, arbitration
Assesses fairness, autonomy, transparency, accountability, non-maleficence, beneficence and dignity, maps regulatory exposure and arbitrates conflicts between Critic and Optimizer.
Color: IndigoOrigin: NewTier: deep
Scenarios, sensitivity analysis, robustness ranking
Three to five scenarios with probability and interval, observable early warning signals and a robustness ranking kept apart from expected value. Runs several times independently.
Color: CyanOrigin: NewTier: deep
Feedback loops, leverage points, cascade effects
Uses the vocabulary of stocks, flows and loops, ranks interventions by Donella Meadows' hierarchy of leverage points and flags where linear reasoning goes wrong.
Color: MagentaOrigin: NewTier: deep
Options × scenarios matrix, dominance, antifragility
Fills every cell of the matrix, drops options dominated in every scenario and recognises the ones that gain from volatility. Runs several times independently.
Color: GrayOrigin: NewTier: fast
Decision archive, recurring patterns, retrospective learning
Recalls similar past decisions by meaning, stores every closed decision with its predictions and proposes new archetypes when it finds at least three similar ones.