MCMMCM DocsEngineering (Internal)
User StoriesMCM-47 — AI-Based Policy Creation
v1.2 is unreleased — see v1.1 for the current stable release.

MCM-47 — AI-Based Policy Creation

Overview and architecture for drafting a Governance policy through the existing chat assistant, with the Admin always creating it themselves.

MCM-47 — AI-Based Policy Creation

Governance lets an Enterprise Admin define compliance policies — either a cloud-provider rule or a host compliance check — and evaluate them against real cloud resources and hosts. Today, writing a new policy means hand-writing its definition from scratch in the existing policy creation flow. This initiative lets an Admin instead describe the policy they want in plain English to the platform's existing AI chat assistant, which drafts the definition and its classification and refines it over further conversation. The assistant only ever produces a preview — the Admin decides when a draft looks right, prefills it into the existing policy creation flow, and creates it themselves.

Each user story below is a self-contained page with its own acceptance criteria, technical design, and UI changes. This page holds only what's shared across all of them.


Personas

PersonaRolePrimary concern
Enterprise AdminOwns the organization's compliance policy configurationGetting a correct policy definition quickly, without hand-writing a cloud policy definition or a host compliance script themselves

System Context

The assistant is an extension of the platform's existing conversational chat capability, which today can already answer questions about existing policies — this initiative adds the ability to draft a new one, on top of the same conversation. The assistant never creates, updates, or deletes a policy itself; creating one is always the Admin's own action in the existing policy creation flow, with its existing validation.


User Stories

StorySummary
MCM-48 — Draft a Policy from a Natural-Language DescriptionDescribe the rule in chat, refine it through conversation, and prefill the accepted draft into the existing policy creation flow.

Out of Scope

  • The assistant creating, updating, or deleting a policy on its own — creating a policy is always the Admin's own action, through the existing policy creation flow.
  • Defaulting a drafted host policy's remediation to auto-remediation — even when the assistant proposes a remediation script, whether it runs automatically stays exactly as manual a choice as it is today.
  • Generating new compliance frameworks or groups beyond the existing built-in set — the assistant only maps a draft into a group that already exists.
  • Generating multiple policies from a single request (e.g. an entire benchmark's worth of policies in one go) — this initiative is scoped to one policy per conversation.

On this page