AI features
What the AI features do, how an Org admin turns them on, and what AI never does in Prodgator.
Availability
Prodgator's AI features explain pipeline failures, triage security findings, summarize the risk of a release, draft release policies, answer questions about your data and build dashboard widgets. They use an AI model that runs in our cloud provider's United States regions.
Enabling AI features
AI features are off by default. An Org admin can turn them on in Organization > AI. When you turn them on, Prodgator shows the privacy statement so you can review how your data is used.
Who can use this: Org admin only. See Roles.
Once they are on, an Org admin can turn single features off. Answers to your own requests (Ask, widgets, policy drafts) follow your language in Profile > Preferences. Shared results (failure explanations, release risk summaries, security triage) are written once, in the organization's default language, and each person whose language differs gets a translation. See Languages.
Available features
Failure explanations
When a pipeline run fails, Prodgator writes a root-cause summary and a suggested fix. The explanation is built from the failing job's logs, the commit's changed files (GitHub only), and similar past failures from your organization.
The one-line headline appears in failure notifications, and the full explanation is shown on the pipeline run page and the Failures view. Prodgator reuses an explanation for the same failure pattern within 24 hours without regenerating.
Cost: 1 credit per explanation. Details: AI failure explanations.
Security triage
On request, from the Security page: groups your open GitHub security alerts by shared cause or fix, orders the groups by priority (P1 to P4), explains each one and points out likely false positives for you to check. Nothing is dismissed for you. Written in the organization's default language, and translated for people who use another. Generating needs the Generate AI results permission: Developer, Release manager, Security, Compliance and Org admin have it.
Cost: 3 credits per triage. Details: AI security triage.
Release risk summaries
Written automatically when a deployment starts waiting for approval (the Generate release risk automatically setting in Organization > AI, on by default; with it off, the first open of the approval dialog writes one): what changed since the last successful deployment to that environment, failures of that change set (with fixes of earlier failures as mitigations), policy results and attestations, with a risk level and what to look at first. Shared by everyone in the organization and reused until the facts change. A summary for the same code and baseline is reused, never across organizations. Advisory only; it never approves or rejects.
Cost: 2 credits per summary; a reused summary costs nothing. Turn automatic generation off to save credits. Details: AI release risk summary.
Policy drafting assistant
From the policy editor: describe a release policy and get form rules or Rego with tests. Every draft must validate, compile and pass its tests before it can be loaded into the editor, always in Observe mode, and nothing is saved until you save it. Needs release policies and the permission to write them.
Cost: 5 credits per draft. Details: Draft a policy with AI.
Ask Prodgator
The Assistant, in the sidebar under Insights or as a side panel with Ctrl+I (Cmd+I on macOS), answers questions about your runs, deployments, approvals and security findings by looking them up with your own permissions, and how-to questions from these docs. Links point only to pages or records it looked up. It only reads. Conversations are private to you and kept for 30 days.
Cost: a message that only used the docs costs 1 AI credit. A message that looked up your data costs 1 credit plus 1 per data lookup, at most 3. Details: Ask Prodgator.
Dashboard widgets from prompts
Enterprise plan only (listed as AI insights on the pricing page). Describe the chart you want, and Prodgator builds a dashboard widget from the data you have access to. Anyone in the organization can generate a widget, including Viewers. Each generation spends 1 credit from the organization's shared monthly allowance, whoever asks for it.
Widgets are private to their creator by default. Release managers, Security, Compliance and Org admins can share them with the rest of the organization, where they follow each viewer's permissions.
Cost: 1 credit per widget. Details: Build a widget from a prompt.
What AI never does
AI features:
- Never automatically approve or reject anything
- Never dismiss findings or failures
- Never save settings, policies, or widgets on your behalf; a policy draft is loaded into the editor for you to review and save
- Never decide a gate: a release risk summary is advisory. Only its level reaches a policy, through the Release risk rule or , and it can only ask for more (an approval, or a failed rule), never approve
- Never modify your data without your explicit action
Disabling AI features
An Org admin can turn off AI features for the entire organization, or disable individual features while keeping others on. In an emergency, an Org admin can disable AI for a single organization through the Organization > AI page or by contacting support.
A feature the organization's plan doesn't include (for example, dashboard widgets on Business) has no toggle at all: it's left off the page rather than shown off and disabled.