AI AGENT INTELLIGENCE · AI GOVERNANCE
AI governance tools document your policy. They do not prove your AI agents follow it.
Most AI governance platforms are registries: they inventory models, store risk assessments and track approvals. None of them listens to what your AI agent actually said to a customer this morning.
WHAT THIS ACTUALLY MEANS
AI governance tools are software for cataloguing AI systems, recording risk classifications and managing internal approval workflows.
Documentation, not behaviourThey produce documentation. What they do not produce is behavioural evidence — proof that a specific AI agent, running in production today, disclosed it was an AI, refused out-of-scope requests, escalated correctly and did not invent an answer under pressure.
One layer below the registryLexic Compass sits one layer below the governance platform: it independently audits AI agents already in production.
A score and a signed verdictIt returns a Trust Score from 0 to 100 across four pillars — Integrity & Safety, Regulatory Trust, Operational Reliability and Experience Trust — with a signed verdict.
Trust Score 0-100ClearedCleared with ConditionsNot Cleared
It is an independent audit, not a certification: Lexic is not a notified body under the EU AI Act.
Governance platform vs. independent agent audit
| Capability | Who covers it |
|---|---|
| Model and system inventory | AI governance platform — the registry is its core job |
| Risk classification and approval workflow | AI governance platform — internal process, self-declared inputs |
| Policy templates and control mapping | AI governance platform |
| Adversarial testing of a live agent | Independent audit — Compass attacks the agent to find where it breaks |
| Evidence from real production conversations | Independent audit — findings are tied to transcripts, not questionnaires |
| Article 50 disclosure verified in the wild | Independent audit — checked in the channel the customer actually uses |
| Third-party verdict for the Board or a regulator | Independent audit — the auditor is not the operator |
“A governance registry tells you which AI systems you own. It cannot tell you which of them is quietly lying to your customers.”
Executives of these companies are already decoding conversations
Frequently asked questions
AI governance tools are platforms that inventory an organisation’s AI systems, classify their risk level, store documentation and route approvals. They are process software: their output is a defensible paper trail. They do not test whether a deployed AI agent behaves as documented, which is why an independent audit layer is usually needed alongside them.