AI features
AI risk scoring
Every risk can be scored by hand. The AI Suggest button proposes values you can accept, adjust, or override. Knowing when to do each is the skill.
How it works
When you click AI Suggest on a risk, Radius360 sends the risk title, description, category, and tenant context (industry, active frameworks, size signals) to the scoring model. The model returns suggested impact and likelihood values on a 1 to 4 scale with a short rationale for each.
The two axes use different vocabularies (matching standard 4×4 risk matrices):
- Impact: Low (1) / Medium (2) / High (3) / Critical (4)
- Likelihood: Rare (1) / Possible (2) / Likely (3) / Almost Certain (4)
The composite score (impact multiplied by likelihood, 1 to 16) maps to a severity badge: 1 to 3 Low, 4 to 7 Medium, 8 to 11 High, 12 to 16 Critical.
Auto-assessed risks
Risks created by integration syncs (M365, Workspace, Okta, AWS, Wiz, etc.) arrive with Impact and Likelihood already set per a default mapping for each finding type. Those rows are flagged auto_assessed = true and surface a violet Auto-assessed by Radius360 banner on the Risk Detail Overview tab with a one-click Confirm button.
While the flag is true, every sync run refreshes Impact/Likelihood/score against the integration's current default. Once you Confirm (or edit and save), the flag flips to false and the assessment is locked; future syncs stop overwriting it. This closes the loop: system proposes, operator ratifies, decision sticks.
Using AI Suggest
Open any risk, click AI Suggest. Within a few seconds you see:
- A suggested impact value with a one-sentence rationale
- A suggested likelihood value with a one-sentence rationale
- The resulting composite score and severity badge
- Three buttons: Accept, Adjust, Override
Accept applies the suggestion as-is. Adjust opens sliders so you can nudge either value up or down without discarding the suggestion. Override discards the suggestion and lets you pick your own values from scratch.
Accepting vs overriding
The AI is generally good at the shape of typical risks. It is not good at business context it does not know.
Accept the suggestion when:
- The risk is a common category (unencrypted storage, missing MFA, unreviewed vendor)
- You do not have strong feelings about the score
- You are scoring a batch of onboarding-wizard-seeded risks and want to move quickly
Override when:
- The risk touches your crown jewels (production database, primary revenue system, customer PII)
- You have business context the AI does not (e.g., "this system is on a deprecation path, it will not matter in 60 days")
- You have threat intelligence pointing at increased likelihood (active campaign, recent near-miss)
- You disagree with the AI's rationale
Overrides train your team
When you override, add a note explaining why. Notes are visible to other owners and auditors. Over time the pattern of overrides is a record of your organization's implicit risk appetite.Limits
AI Suggest is available on Starter (100 per month), Program (500 per month), Practice (1,500 per month), and Enterprise (unlimited). Manual scoring is always unlimited.
See GRC Assistant and AI policy generation for the other AI surfaces, and billing for plan details.