Products

Agent assurance built around evidence, not vibes.

DialogAssert is building a behavioural assurance layer for customer-facing AI agents, from pre-release evals to adversarial testing and pre-built regulated workflow knowledge.

Evals

Available now

Turn business journeys into evidence-grounded agent tests.

Write customer turns and expected behaviour. Run the same journey repeatedly against your agent. See which acceptance criteria failed and the transcript or tool evidence behind each finding.

Explore Evals →
Business-owned file
turns:
  - I would like to block my card.
  - the card ending with {card_number}.
  - yes, please go ahead

acceptance:
  - Agent shows the customer their cards before asking which card to block.
  - Agent blocks only the card selected by the customer.
  - Agent asks for confirmation before blocking.

Red Teaming

Coming soon

Probe regulated workflows adversarially.

Generate adversarial multi-turn journeys around sensitive data, unauthorised actions, advice boundaries and escalation pathways.

Domain Packs

Coming soon as a standalone product

Pre-built testing knowledge for regulated domains.

Start from domain-specific journeys, acceptance patterns, agent contracts, assertions and calibrated severity instead of constructing every test from scratch.

Banking Domain Pack is being proven first →

See DialogAssert on a real agent workflow.