Thirty-six specialist agent roles work across three governed swarms to challenge the brief, interrogate the catalogue, simulate the future state and prove it can be defended, funded and operated.
3.
governed swarms
36.
specialist agent roles
122.
solution-service families
210.
pairwise overlap authorities
Six adversarial perspectives challenge the brief. Thirty-six specialist passes across three governed swarms build the catalogue, write the blueprint and simulate the dream state. The system preserves disagreement, separates hypotheses from evidence and keeps publication under human control.
06.
adversarial perspectives
36.
registered specialist passes
01.
human-controlled answer
The decision swarm works on the submitted brief. Each role has a different failure to find. The moderator is rewarded for exposing disagreement, not manufacturing consensus.
Strips away unsupported assumptions and asks what would disprove the plan.
Tests staffing, alert load, handoffs, rollback and day-two ownership.
Challenges data boundaries, control failure, auditability and blast radius.
Exposes licence overlap, renewal timing, proof conditions and exit cost.
Maps architecture, dependencies, integrations and delivery sequence.
Forces disagreement into the open and converts it into one decision path.
Every number on this page is read from the registry of passes in the product. A role is a schema-bound specialist pass with its own inputs, its own output schema and its own failure to find. Remove a pass and the number moves.
Builds and polices the capability catalogue. Never reads a customer transcript.
08.
Pins the company, aliases and official hosts before research begins.
Turns prose into one falsifiable claim bound to an exact passage.
Translates buyer language into deterministic routing signals.
Freshly re-checks the source without seeing the extractor's rationale.
Retains stale, renamed, acquired and contradictory evidence as visible risk.
Compares every family with the roster and records what flips the decision.
Builds positive, negative, abstention, overlap and injection cases.
Blocks release when mandatory proof is missing.
Works only on the verified transcript and the approved catalogue. Cannot publish.
13.
Pins one horizon year and one biography for every later pass.
Reads the conversation and writes the opening in the leader's own terms.
Names the distance between today and the stated dream state.
Drafts the moves that can start before anything is bought.
Writes what a good morning looks like once the work is done.
Separates people, processes and tools instead of collapsing them.
States the interpretation and the agreement boundary.
Collects the verbatim evidence cards with their locators.
Writes the executive read and the scored verdict.
Lays out the sequenced implementation path and its proof.
Runs the adversarial debates and keeps the disagreement visible.
Matches catalogue specialists to workstreams with proof signals.
Records the risks and the counter move for each.
Simulates the dream state from a pinned snapshot. Cannot change the blueprint.
15.
Records the tools already owned, each bound to a transcript locator.
Models people, processes and capability entities as archetypes.
Stages the moments and declares the capability need behind each step.
Rehearses what each system shows at every milestone, never as telemetry.
Derives the arc across the leader's own milestones.
Runs the branch where nothing changes, without inventing statistics.
Stages earned defences per archetype using the leader's own metaphors.
Renders the defensible combination as a working system with control points.
Tests what holds and what gives, leaving the proof question open.
Fails the snapshot closed on any unevidenced or brand-leaking claim.
Connects document sections to the simulation they belong to.
Composes the document pages and the interactive lab surface.
Derives every in-frame prop from the guest's own sector, transcript and estate.
Renders every scene and fails closed on a fallback frame, empty region or budget breach.
Repairs a blocked simulation in bounded deterministic rounds before it can fail.
A messy transcript becomes an owned action plan, relevant capability categories, named vendor hypotheses, sharp proof questions and a navigable PDF. Uncertainty stays visible. Evidence and commercial approval remain separate.
Transcript, brief, pressure and constraints.
Six roles challenge the decision from different failure modes.
The real workstreams emerge without requiring vendor names.
Only approved vendor hypotheses and researched capabilities enter.
Owners, risks, proof questions, milestones and PDF.
The advantage is division of labour, independent challenge and explicit gates. Vendor names remain hypotheses. Catalogue claims require retained evidence. Research agents cannot publish. A human still owns the decision. Here, an agent role means a schema-bound specialist pass or an adversarial decision perspective - never an extra vote for a weak claim.
Upload the transcript or write the brief as you would say it to a trusted peer. No tidy slideware required.
The decision roles challenge the frame while the governed catalogue constrains capability and vendor hypotheses.
Take away the workstreams, people, processes, risks, next moves, proof questions and a board-ready PDF.
