AI settings
Control how messages are processed and how responses are generated in EBI's AI Studio.
Outcome
By the end of this guide you will understand (and be able to configure) how EBI's AI Studio decides whether to:
- Hand over to Live chat
- Trigger a Flow
- Generate a response using AI
Prerequisites
- You can sign in to EBI's AI Studio
- You have permission to edit settings (some options are plan- or role-dependent)
Steps
1) Open AI settings
- In the left navigation, select AI settings.
2) Configure User Message Journey
Under User Message Journey, review how a message is processed.
Step 1: Check for live chat agents first
- If On, EBI's AI Studio checks whether a live chat agent is available and can hand over the conversation.
- If Off, live chat handover is skipped.
Step 2: Use NLP intent detection to run a flow
- If On, EBI's AI Studio tries to match the user message to a flow intent.
- Review Minimum NLP confidence to trigger a flow and set it to the level you want.
Step 3: Generate responses
Select how responses should be generated when the system reaches response generation:
- Use AI: the assistant generates a response using AI.
- Select flow: the assistant generates a response using a selected flow (if available).
- Off: response generation is disabled.
Step 4: Show generated responses directly to the user
- If On, generated responses are shown directly to the user.
- If Off, generated responses are not shown directly (this can be useful if you want to review or route responses differently).
Step 5: Create flows from generated responses
If enabled, EBI's AI Studio can create flows using generated responses:
- Publish: create and publish flows automatically.
- Draft: create flows as drafts for review.
- Off: do not create flows from generated responses.
Step 6: Ask user for feedback
- If On, the assistant asks the user for feedback after responses (for example “Was this helpful?”).
- If Off, feedback prompts are disabled.

3) Configure response generation instructions and options (plan-dependent)
Under Response generation instructions, you may see options such as:
- A free-text instruction box (select Save to apply changes)
- Citations (on/off)
- Allow AI to use tools
- Model selection
- File filter / metadata filtering options
If you don’t see an option, your plan or role may not include it.
4) Configure Pattern matching (optional)
Use Pattern matching to identify and extract (and optionally redact) data from user messages.
- Under Pattern matching, enter a Name and a Regular Expression (Regex) pattern.
- If you want the matched value to be redacted when storing and processing utterances, enable Redact this when storing and processing utterances.
- Select Add.
5) Redact user inputs captured by flows (optional)
If you capture user inputs using Ask a question, you can optionally redact the captured value in conversation logs:
- In a flow, add or select an Ask a question bubble.
- Enable Redact this input in conversation logs.

6) Save your changes
AI settings are saved as you update controls on the page. If you’re testing changes, refresh Try your live AI and send a new message to confirm the behaviour.
Verify
- Send a test message in Try your live AI.
- Confirm the expected behaviour happens:
- Live chat handover (if enabled and an agent is online), or
- A flow triggers, or
- A generated response is returned.
- In the left navigation, select Request viewer and confirm you can find the request.
Common issues
- A flow doesn’t trigger: lower Minimum NLP confidence to trigger a flow, add training examples, or expand intent coverage.
- Live chat never triggers: confirm Step 1: Check for live chat agents first is On and at least one agent is online in Live chat.
- You can’t see certain options: your plan or role may not include them.
- A regex match doesn’t work: test the pattern on a sample message and ensure it matches the full value you need.