Skip to content

AI tools & credits

AI governance in Keel

Keel weaves AI through the whole platform as a set of concrete, optional actions, each mapped to a real job a compliance owner does, not a chat gimmick. Every AI action is credit-metered, so you only spend when you choose to run one, and you always review what it produces.

  • Control implementation guidance: For any control, get plain-English, step-by-step implementation guidance and the exact evidence to collect. Turns “where do I start” into a checklist you can act on today.
  • Evidence summary & sufficiency review: Summarize an artifact and check whether it’s likely sufficient for the control it’s attached to, flagging what’s missing. See Evidence.
  • Policy drafting: Draft a clean, framework-mapped first draft of a named policy right in the editor. See Policies.
  • Policy & document import cleanup: Rewrite a messy imported .docx into clean, framework-mapped Markdown.
  • Policy-gap analysis: Compare your policy set to a framework and flag what’s missing.
  • Readiness summary: Generate a board-ready audit-readiness summary written over your live posture.
  • Questionnaire assist: Draft answers to an inbound security questionnaire from your own controls and policies, grounded and ready to review.
  • Vendor & risk drafting: Draft a vendor risk profile from a URL, and draft scored risks to seed your register.

The exact set of tools evolves as the product grows; each one shows up in context inside the module it belongs to.

AI actions consume AI credits. Paid plans include a monthly credit allotment, and you can top up with credit packs when you need more. Because each action is a deliberate, credit-metered click, your spend stays predictable. Nothing runs in the background without you asking.

Manage and monitor credits in the app under Settings → AI credits, where you can see your allotment and usage.

Knowing what’s AI-drafted vs human-reviewed

Section titled “Knowing what’s AI-drafted vs human-reviewed”

Because AI drafts are starting points, Keel labels them clearly so you - and your auditors - always know what to trust. Every piece of AI-generated content carries a provenance badge:

  • AI-drafted — review before use: the default state for anything an AI tool produced (control guidance, policy-gap and readiness analyses, evidence and trust-centre narratives, access-review notes, questionnaire answers, vendor risk drafts, AI Insights). It’s a prompt to check the content before you rely on it or share it.
  • Human-reviewed: shown only after a member explicitly marks the content reviewed. The badge records who signed off and when. Keel never shows content as human-reviewed unless a person actually approved it.

For questionnaire answers, the badge sits alongside the grounded answering engine’s own signals - an explainable confidence level and a confirm internally flag when the engine couldn’t fully ground an answer - so you can prioritise what to check first.

AI is optional. Each workspace has a per-workspace AI on/off switch, so an organization that prefers not to use AI features (or needs them off for a particular tenant) can disable them entirely. When AI is off, every other part of Keel works exactly as before.

  • Policies: where drafting, import cleanup, and gap analysis live.
  • Evidence: the AI sufficiency check in context.
  • API & MCP: connect your own AI agents to Keel.