Controlled early access

ETOS Agent Assurance

Governed decision assurance for autonomous and enterprise AI agents. Connect your agents to ETOS GateKeeper before consequential decisions and actions. ETOS evaluates available evidence, governance requirements, authority, conditions and escalation requirements to determine whether a decision can progress.

Your agent asks. ETOS assures. Humans retain authority where required.

Assurance pathGovernance intact

ENTERPRISE AGENT
      |
      v
ETOS GATEKEEPER
      |
      v
GOVERNED ETOS CAPABILITY
      |
      |-- Evidence sufficiency
      |-- Provenance
      |-- Authority
      |-- Controls
      |-- Conditions
      `-- Escalation
      |
      v
DECISION ASSURANCE
      |
      |-- GO
      |-- CONDITIONAL GO
      |-- INSUFFICIENT EVIDENCE
      |-- ESCALATE
       |-- HOLD
      `-- NO-GO

What ETOS assesses

Evidence sufficiency

Is there enough, and is it good enough?

Provenance

Where did each supporting source come from?

Authority

Has the required mandate actually been established?

Controls

Are the applicable policy and control requirements met?

Conditions

What must hold true before action is permitted?

Escalation

When must a human take the decision?

Decision assurance states

Assurance satisfied. Any action remains subject to the applicable human or organisational authority.

CONDITIONAL GO

Assurance depends on the stated conditions being met; this state does not itself authorise action.

INSUFFICIENT EVIDENCE

Required evidence, provenance or control evidence is missing.

ESCALATE

Human authority is required before the action can proceed.

HOLD

Pause progression until the stated assurance requirements or unresolved conditions are addressed.

NO-GO

The proposed action cannot be supported on the record presented.

Consumption

One determination. Machine and human consumers.

The assurance response is designed to be acted on by downstream workflows and reviewed by accountable people — from the same determination, not two separate accounts of it.

Machine-consumable determination

A structured assurance record your orchestration layer can branch on: assurance state, distinct authority status, open conditions and permitted next action.

Human-review artefact

The same determination as an interactive human review document, with evidence references, reviewer notes and accountable sign-off.

Retained assurance record

Every determination is referenceable after the fact, so the route to the action can be re-examined when it is challenged.

Example determinationMachine-readable

{
  "decision": "CONDITIONAL_GO",
  "authority_verified": true,
  "human_authority_required": true,
  "open_conditions": 2,
  "agent_may_execute": false,
  "permitted_next_action": "escalate",
  "assurance_record": "ETOS-..."
}

What your agents receive

Agents are granted access to ETOS governed decision-assurance capabilities — not to the underlying ETOS repository. The repository remains protected behind GateKeeper. What crosses the boundary is the governed determination and its assurance record.

An agent-to-agent connection has already demonstrated that an external agent can consume the governed ETOS decision path. This programme is the controlled commercial stage of that capability.

Position

ETOS does not sell favourable decisions. Access purchases governed assurance. The evidence determines the outcome.

Access

Request Agent Assurance Access

Agent assurance is released under controlled early access. Credentials are not issued automatically — organisations are approved individually, as with GateKeeper platform access.

Approved organisations are onboarded with a scoped assurance allowance.