Promptective journal / Topic index

policy enforcement

4 journal entries tagged policy enforcement.

Two ivory blocks wait in a terracotta channel behind a lowered graphite gate while a third rests on a separate output tray.

Guide8 October 20265 min read

AI batch cancellation: reconcile partial results

Account for work that finishes after cancellation is requested.

Compare OpenAI and Anthropic batch cancellation states. Reconcile partial results by request ID, govern retries and track separate input, output and record retention.

Read guide

A brass service bell and a graphite rail with a raised end stop rest on a terracotta tile on a pale desk.

Guide7 October 20265 min read

AI agent budgets: alerts and hard limits

Identify what stops new requests and how work recovers at the limit.

Compare optional OpenAI hard spend limits with Azure and AWS budget behaviour. Define billing scope, enforcement delay, fallback authority and recovery for agents.

Read guide

Blue rectangular and orange round connector plugs lie beside separate matching metal sockets on a terracotta surface.

Guide6 October 20265 min read

AI agent delegated access: govern every API hop

Check the user, granted scope, token audience and resource permission together.

Review AI agent delegated access through Microsoft OBO. Check application scopes, user rights, token audiences, consent and Conditional Access at each API boundary.

Read guide

Three compact silver desktop displays stand on a dark surface; the front screen is cobalt blue and the two rear screens are navy.

Guide4 October 20265 min read

AI policy rollout: verify each stage

Connect the approved rule to assignment, device state and tested behaviour.

Track approved AI policy, assignment, device application and observed behaviour. Use Intune report limits and NIST change-control guidance in a bounded pilot.

Read guide

All Articles, Guides and News