AI Assistance
AI-powered writing assistance helps Product Owners draft features, acceptance criteria and supporting notes faster. It is a separate paid bolt-on available on every subscription tier. Bring your own API key (BYOK) is supported.
Setting up AI Assistance
Turn on the bolt-on
Open Workspace → Capabilities, scroll to Bolt-Ons, and enable AI Assistance.
Configure the provider
Go to Integrations, open AI Providers and set up the Planner Agent: choose your provider, paste your API key and press Apply & Test, then choose a model. OpenAI, Anthropic, GitHub Models, Azure OpenAI and a custom OpenAI-compatible endpoint are supported. For Azure OpenAI and a custom endpoint the page also asks for the endpoint (its base URL, starting with https://): enter it before you apply the key, because it is saved with the key, and the key is only ever sent to that address. The endpoint is the base URL and nothing more: an address with a user name or password in it, or with anything from a ? or a # onwards (an access code, for example), is refused and is not saved. For OpenAI, Anthropic and GitHub Models the model list is read from your provider once a key is applied, so it reflects what your account can call. For Azure OpenAI, choose a model from the list or enter your deployment name by hand. A custom endpoint has no list to choose from: pick Other (enter manually) and type the model name.
Use the inline controls
When both the bolt-on and workspace AI settings are enabled, the subtle ✨ control appears beside supported text fields. It stays hidden everywhere else.
What AI can help with
Given a feature title, generate a draft problem statement, hypothesis, and acceptance criteria.
Expand a short description into a structured Given/When/Then acceptance criteria set.
Convert technical story descriptions into polished, stakeholder-friendly release note entries.
Analyse a large story and suggest how to split it into smaller, independently deliverable items.
Given a feature description, suggest initial WSJF dimension scores with reasoning.
For a user journey stage, suggest common touchpoints and pain points based on the persona description.
Using AI in a Feature popup
On supported popups, look for the subtle ✨ button next to text fields. Clicking it opens the AI panel. Add any optional notes, run the prompt, review the suggestion, and click Accept to apply it. AI completions are never inserted automatically.
Data privacy
- Your API key is sent to our server once, when you apply it. The server checks it against your provider and stores it encrypted in the credential vault (AWS KMS). It is never stored in your browser and never sent back to it: the browser keeps only a marker that a key is saved. Test connection, the model list and every AI request are made by the server with that key.
- The text sent to the AI model is limited to the specific field context — not the entire workspace.
- AI suggestions always require explicit accept or revert.
- If the bolt-on is off, the AI controls do not render and no AI calls are made.