Planned
Extend reliability from answers to AI operations.
The longer-term direction applies the same independent evaluation principles to automated work steps, agentic processes and organisation-specific standards.
Planned
Direction
Evaluate work across multiple steps
Agentic work introduces plans, intermediate results and automated handoffs that can each carry uncertainty. The intended reliability system makes those stages reviewable rather than waiting for the final answer.
- Visible intermediate states
- Requirement checks across steps
- Human review points retained
Boundaries
A direction, not a current availability claim
Organisation controls and agentic reliability remain planned. The product will distinguish experimentation, preview and availability as these capabilities become concrete.
- Lifecycle labels remain explicit
- No autonomous guarantee
- Human decision authority preserved