Brilo AI supports widget conversation script customisation so you can control the exact words, flow, and fallback behavior the Brilo AI voice widget uses with website visitors. You can edit greetings, prompts, branch logic, and fallback prompts in the Brilo AI console; when deployed the widget runs that script while using Brilo AI voice agent capabilities like intent detection and dynamic prompts. Script edits take effect after you save and publish the voice agent configuration; handoff, confidence thresholds, and recording settings remain governed by your routing and escalation rules. If you need a formal review for sensitive content or SSML voice changes, Brilo AI may require additional approvals.

Can I edit the whole widget script? — Yes. You can edit greetings, branches, and fallbacks in the Brilo AI console and publish updates to the live voice widget.

Can the widget use my exact phrasing? — Yes. Brilo AI lets you author exact prompts and SSML (speech markup) for precise wording and prosody.

Will edits break callers mid-call? — No. Published changes apply to new sessions; active calls continue on the version they started with.

Buyers ask about widget conversation script customisation because website voice widgets represent the brand’s primary live interaction. Enterprises need to ensure scripts match legal disclosures, regulatory language, and brand tone before the widget speaks to real customers. Banks, insurers, and healthcare providers especially worry about sensitive language, error handling, and how the widget manages intent failures or requests for a human. Knowing whether the entire conversation can be controlled — and how changes are deployed — affects testing, compliance review, and go-live timing.

Brilo AI voice widget scripts are managed in the Brilo AI console as conversation flows (call flows). You author the script using Brilo AI’s flow editor, mapping prompts, branches, and fallback routes. The Brilo AI voice agent uses intent detection (natural language understanding) to choose the next prompt or branch at runtime, and it injects contextual data from the session when available.

Conversation flow is the ordered set of prompts, branches, and actions that the voice widget will execute during a session. Widget script is the editable text and SSML attached to each node in a conversation flow.

For details about how Brilo AI detects caller intent and routes within flows, see the Brilo AI article on how the AI understands caller intent: How Brilo AI understands what the caller wants.

Brilo AI enforces guardrails so widget scripts do not create unsafe or noncompliant behavior. Common boundaries include refusing to collect or retain sensitive data unless explicitly enabled, escalating when confidence is low, and stopping scripted actions that may trigger regulatory obligations. You can configure confidence thresholds and keyword triggers so the Brilo AI voice agent hands off when the detected intent is uncertain or when a caller asks for a human.

A handoff trigger is a configured condition (for example, low confidence or specific keywords) that causes the voice widget to transfer or queue the session to a human. To review voice and prosody limits or request advanced SSML changes, consult Brilo AI’s guidance on naturalness and voice configuration: Does the AI sound natural or robotic?

Healthcare example: A hospital places a Brilo AI voice widget on its patient portal. The script includes a secure greeting, prompts to authenticate by patient ID, and a fallback that routes complex clinical questions to a nurse triage queue. The Brilo AI voice agent escalates to a human if authentication fails twice or if the caller explicitly requests a clinician.

Banking example: A retail bank uses the Brilo AI widget to handle balance inquiries and simple payments. The widget script contains precise disclosure language and a fallback that queues calls to live support when the voice agent detects transaction disputes or low confidence in intent detection.

Insurance example: An insurer configures the Brilo AI voice widget to collect claim intake details with structured prompts, then offers a warm transfer to an agent for high-severity claims or potential fraud indicators.

Brilo AI voice widget workflows can hand off to a human agent or another workflow when configured. Handoffs can be triggered by specific keywords, repeated failures to resolve intent, elapsed duration, or an explicit caller request. During transfer, the Brilo AI voice agent passes session context (transcript, detected intent, extracted entities, and recent prompts) so the receiving human agent can continue without asking the caller to repeat information. You can configure warm transfers (with context) or cold transfers depending on your telephony and routing setup.

For guidance on handling long or complex conversations and related configuration, see: Can the AI handle long conversations?

Customising the widget conversation script with Brilo AI reduces miscommunication and rework by ensuring the first-line voice interactions reflect approved language and operational routing. Clear scripts reduce unnecessary escalations, shorten time-to-resolution for routine requests, and improve agent readiness when handoffs occur. For regulated organisations, script control simplifies compliance review and auditability because every deployed prompt is versioned and associated with the Brilo AI voice agent configuration.

Can I upload my own legal disclosure language into the widget script?

Yes. You can place required disclosure language into scripted nodes. Have your legal and compliance teams review wording and keep those nodes in a locked or review-only part of the flow if necessary.

Will changing the script immediately affect callers in-progress?

No. Changes apply to new sessions after you save and publish the agent. Active calls continue using the version that started their session.

Can the widget use SSML for prosody and pauses?

Yes. Brilo AI supports SSML in script nodes for advanced voice control; however, changes that require voice cloning or other high-risk voice modifications may need a Support request and legal consent.

How does Brilo AI handle sensitive information captured by the widget?

Brilo AI surfaces session transcripts and metadata according to your account recording and data handling settings. Configure retention and export rules in your Brilo AI settings and ensure your internal policies cover any sensitive data capture.

Can I require human approval before a script change goes live?

Yes. Implement a change control process outside Brilo AI or use role-based permissions in the Brilo AI console so only authorized users can publish changes.