Governance
How Truo agents are governed
Truo agents do the operational work of a trade. They do not make the decisions a company is accountable for. These four rules hold for every agent, whichever party it works for.
1. Bounded autonomy
Your team writes the policy: what each agent may do on its own, and what it may only propose.
Routine work, such as collecting open issues, ranking them, checking evidence against requirements and drafting documents, can run on its own. Anything with consequences outside the workspace becomes a proposal in the approvals queue.
2. Sign-off where it counts
Supplier and specification changes, declarations, anything sent to a counterparty, and dispatch release wait for the accountable person.
An agent can prepare all of these. It cannot complete them. When applicability is unknown, release is blocked until a qualified reviewer decides.
3. Results and decisions kept apart
A system result keeps its source, scope, time and confidence. A human decision keeps the reviewer, basis and outcome.
That separation means anyone reviewing the trade later, including a buyer, a broker or an auditor, can see what the agent found and what the person decided, and never mistakes one for the other. Agent or provider results never replace qualified review, a customs outcome or dispatch release.
4. Scoped by role
Each participant in a trade sees only what their part needs, and an agent never sees more than the person or workspace it works for.
Who is responsible when an agent drafts a document?
The person who approves it. The draft and the decision are recorded separately on the trade record.
Why this matters now
The EU AI Act makes human oversight a core requirement for AI systems used in high-stakes settings, and buyers increasingly ask suppliers how automated decisions are controlled. Truo's governance is a product behaviour you can inspect on every trade record. It does not, on its own, make your use of AI compliant with any law.
See the eight agents at work · What each agent does and who decides