Product · prismcouncil

PrismCouncil

PrismCouncil is the strategic decision layer for Agent governance: assisting human decision-makers in evaluating whether to build, how to execute, and how much to invest through multi-perspective advisor deliberation that synthesizes into a structured Decision Package. Once approved by humans, a Program is created in PathPilot when connected. This service does not call LLMs directly; inference runs on NodalOS.

PrismCouncil

Strategic Decision CouncilStrategic decision council for Agent governance

What to build, how to execute, and what to invest—multi-perspective deliberation with human arbitration, deployable as a PathPilot Program upon sign-off.

8+CDO: Perspective Advisors + Data Officer
Closed Loop: Deploys as a Program upon approval
Traceable: Audit-ready advisor stances & approvals

Typical pains

High-stakes strategic decisions lack structured multi-perspective vetting

Single-track Agent planning overlooks compliance, financial, and security trade-offs

Deliberation lacks traceable records and cannot convert directly into executable projects

Capabilities

Multi-Perspective Advisor Deliberation8 built-in advisors (Risk / Tech / Finance / Marketing + Strategy / Operations / Org & Talent / Compliance). Built-in advisors can be disabled but not deleted; custom advisors can be added

Councilor OrchestrationThe Councilor chooses between group and solo modes based on issue complexity; in group mode, it sets the roster and depth (quick / standard / deep). Humans can also specify the mode when initiating

Resident Chief Data Officer (CDO)The CDO provides supporting data and charts; operates as a data role rather than one of the 8 perspective advisors, and can be disabled but not deleted

Adversarial TensionBuilt-in adversarial dynamics maintain productive tension across advisor roles to serve substantive business simulation rather than theatrical debate

Configurable Deliberation DepthGroup deliberation supports quick, standard, and deep modes

Nested Sub-TopicsUp to two child levels; omitted roster and depth inherit from the parent, and parent background is written into the child. Closed or rejected topics cannot open new consults; topics already handed to a PathPilot Program can still deliberate

Advisor Stance History and OccupancyHistorical advisor stances are queryable; the advisor list shows currently occupied topics. This is not a name/role résumé, and busy advisors are not auto-replaced

Built-in Issue Templates7 out-of-the-box issue templates with suggested advisor rosters and default depths (customizable; custom templates supported)

Decision PackageStructured output covering project name, budget, work groups, milestones, and execution plans for handoff to PathPilot

Decision → Program Closed LoopOnce humans approve the Decision Package, it deploys to PathPilot as a Program and AULO records the created Program (requires PathPilot integration)

Decision TraceabilityHistorical advisor stances are logged in-service; human approval events feed into audit chains when connected to OwlAudit

Human Final ArbitrationDeployment requires explicit human approval; rejections mark issues as rejected. The system never supersedes human authority on business decisions

Does / does not

Does

Multi-perspective deliberation across 8 built-in advisors
Councilor dynamically switches between group and solo modes based on issue weight
Nested sub-topics (up to two child levels)
Decision Packages and post-approval Program deployment
Human arbitration with traceable advisor stances
Dedicated Chief Data Officer provides data analysis and charts

Does not

Does not replace final human business decisions
Is not an off-the-shelf debate toy
Does not call LLMs directly from this service

Integrations

NodalOS runtimeAULO council panelOwlAudit approval events (optional integration)PathPilot Program creation (requires integration enabled)Standalone delivery manages council SKILLs independently without LarkScout / Harness sync chains

Related products

AULOOne workspace for human–Agent work: inbox, conversations, council, and governance actions PathPilotAI Agent task execution control plane OwlAuditAI Agent compliance audit and human-in-the-loop (HITL) governance LarkScoutEnterprise knowledge platform for AI Agents

Next steps

See the full product loop Compare alternatives Integration path Book a demo