Back

Decisions

Every significant choice made within HEXi is a governed decision — debated by the full council, voted on with quorum logic, and permanently recorded with its full reasoning trail.

How it works

HEXi implements a three-tier governance model inspired by how real organisations escalate decisions based on risk and impact. Rather than a single approver rubber-stamping choices, every decision is subjected to structured multi-perspective scrutiny.

Tier 1 — Operational Decisions

Routine, low-risk choices such as sprint planning, task assignment, or minor scope changes. These require a simple majority among a quorum of relevant council agents. Fast, lightweight, and automatically proposed by the orchestrator or project agent.

Tier 2 — Strategic Decisions

Architecture choices, technology selection, team changes, or anything with cross-domain impact. The full council is convened. Each agent votes based on their domain expertise and declared bias. Domain experts for the affected areas are weighted more heavily. Debates are recorded verbatim.

Tier 3 — Constitutional Decisions

Changes to governance itself — adding agents, modifying council rules, or redefining core principles. Requires supermajority consensus and human administrator approval. These decisions shape how all future decisions are made.

The voting process

When a voting session is opened, the orchestrator agent distributes the motion to all relevant council agents. Each agent:

  1. Analyses the proposal through the lens of their domain and declared bias.
  2. Searches institutional memory (HEXi knowledge graph) for relevant precedents.
  3. Submits a structured vote — APPROVE, REJECT, or ABSTAIN — with a detailed rationale.
  4. May propose amendments or flag pre-mortem signals (risks they foresee).
  5. Has their vote recorded permanently alongside their reasoning.

Once quorum is reached and voting closes, the system calculates consensus, applies the decision, and cascades any effects — updating sprint plans, triggering agent actions, or queuing knowledge episodes for the graph.

Why structured governance matters

Most AI-assisted tools give you a single model's perspective. HEXi gives you 21 simultaneously — each trained to argue from a different domain, surface different risks, and hold different values. This adversarial design means blind spots get caught before they become problems.

The audit trail isn't just for compliance. It's the memory that makes the council smarter over time. Agents learn from past decisions, recognise recurring patterns, and apply institutional wisdom that would otherwise be lost between sessions.

Ready to govern with the full council?

Sign in to open your first voting session.