Frequently asked questions about Trunnion AI

This hub answers the questions buyers, technical reviewers and partners most often need before a product conversation. The answer sections below add context; the direct-answer block remains structured as FAQPage data.

What does governed agentic AI mean?

Agentic software can pursue a bounded goal through multiple steps and tools. Governed agentic AI applies identity, policy, approval and evidence controls to those actions before and after they reach external systems.

Governance is not only a dashboard or written policy. The runtime needs enforceable decisions, typed actions, parameter-bound review, deterministic failure behavior and records another authorized party can inspect.

Which products are in the portfolio?

The company site lists seven vertical platforms: Zennvue, Cortrova, Haultro, Pulse, RealtrAI, Vortiek and Gimbra. It also lists six innovations: Trunnion Core, Voice, LLM, Meet, Port and DAF. Custom software is a separate service lane, not part of the platform roster.

Live and Building describe maturity at a high level. Feature entitlement, integration, commercial terms and current availability belong to the applicable release and product offer.

How does Trunnion use models?

The public architecture is LLM-agnostic in the operational sense that model choice is intended to remain behind a governed endpoint contract. The selected model must still be evaluated for the task, data boundary, cost, performance and deployment.

Changing a provider should not erase policy or audit continuity. See why LLM agnosticism matters.

How do human approvals work?

Consequential actions should pause before execution. The reviewer sees the exact action, target, material parameters, relevant evidence and expected effect. The approval is bound to those parameters, expires and can be consumed only according to policy.

The approval-gate guide includes tiers, states, pseudocode and evaluation tests.

How are actions audited?

Audit records should connect identity, task, policy, approval, tool request, receipt, result and release context. Cryptographic linkage can make later changes to the retained order and content detectable.

The public audit-trail specification and browser verifier demonstrate the method and its limits. They do not publish Trunnion's private production schema.

Where can Trunnion software run?

Deployment planning covers cloud, on-premise connected and air-gapped environments. The complete dependency inventory matters: models, identity, policy, tools, audit, administration, monitoring, updates and support.

Review the deployment comparison before treating one label as an answer.

How should a buyer evaluate Trunnion?

Bring one representative workflow and define the action boundary, decision owner, data classes, target systems, deployment environment and acceptance criteria. Ask for release-specific evidence, negative tests and known limitations.

Use the same dimension-first process you would apply to any agentic governance platform. Public claims establish the questions. Release evidence establishes the answer.

How can someone contact Trunnion safely?

Use the contact page for a general description of the workflow and requested product lane. Do not send classified, CUI, export-controlled, procurement-sensitive or other restricted information through the public website. Authorized channels and evidence exchange are established after scope and responsibility are clear.