Direct Answer (TL;DR)
Brilo AI Out Of Scope Control lets the Brilo AI voice agent detect requests that fall outside approved topics and decline them using conservative fallback responses, confidence-based escalation, and optional human handoff. When the system detects low confidence in intent or recognizes a sensitive or prohibited topic, Brilo AI either asks a short clarification, politely refuses to answer, or routes the caller to a human agent according to configured escalation rules. These behaviors are configurable so customers can control when the agent refuses, retries, or transfers the call. Out Of Scope Control is designed for predictable, auditable refusals rather than improvisational answers.
How does Brilo AI decline requests that aren’t allowed?
The agent uses intent detection and confidence thresholds to refuse or escalate; it will give a short refusal and offer to connect to a person when needed.
How does Brilo AI stop itself from guessing on sensitive topics?
Brilo AI triggers a refusal when a request matches out-of-scope rules or fails verification checks, and logs the event for review.
What happens when the agent is unsure about an account-specific question?
The agent asks one clarification; if the system still lacks grounding (KB or CRM), it declines and routes to a human.
Why This Question Comes Up (problem context)
Enterprises need predictable refusals when an automated voice agent encounters requests that could cause harm, regulatory exposure, or incorrect actions. Buyers ask about Out Of Scope Control because unsupervised agents can fabricate answers, attempt prohibited transactions, or provide partial guidance on regulated topics in healthcare, banking, or insurance. Brilo AI customers want transparency about when the agent will say “I can’t help with that” and how those refusals are logged, audited, and routed.
How It Works (High-Level)
Brilo AI Out Of Scope Control operates by matching caller intent against an approved scope, applying confidence thresholds, and following configured decline rules. In typical workflows, the Brilo AI voice agent listens, classifies intent, checks scope and grounding (for example, a knowledge base or CRM), evaluates confidence, and then chooses one of: answer, ask a clarification question, decline with a fallback response, or escalate to a human.
Out-of-scope control is configured as a routing and response layer that sits between intent detection and action execution. A fallback response is a safe, pre-approved message the agent uses when refusing or deferring a request. For more on uncertainty and escalation behavior, see the Brilo AI article on what happens when the AI is unsure: Brilo AI: What happens when the AI is unsure?
An out-of-scope request is any caller intent or phrasing that your organization has marked as prohibited, sensitive, or unsupported by configured grounding sources.
Guardrails & Boundaries
Require grounding before answering account- or health-specific questions; without grounding, the agent must decline.
Use confidence thresholds for both speech recognition and intent classification; below threshold, decline or escalate.
Limit clarification loops (for example, max clarifying questions) to avoid caller frustration.
Prevent the agent from initiating high-risk actions (payments, policy changes, medical recommendations) unless a human has authorized the action.
Log every decline event with the reason and conversation snippet for review.
Confidence threshold is the numeric or rule-based cutoff that determines whether the agent proceeds, asks for clarification, or declines. For details on preventing fabricated answers and enforcing conservative fallbacks, see: Brilo AI: How do you prevent wrong or made-up answers?
Applied Examples
Healthcare: A patient asks the Brilo AI voice agent for a diagnosis. The agent recognizes the request as clinical and out-of-scope for automated advice, issues a refusal, offers to connect to nursing triage or schedule an appointment, and logs the interaction for audit and clinician review.
Banking: A caller asks to reverse a wire transfer. The Brilo AI voice agent checks for required verification and, seeing the request as a high-risk transaction not allowed for automated execution, declines and routes to a fraud specialist with a secure handoff token.
Insurance: A customer asks the agent for legal advice about claim denial. The Brilo AI voice agent identifies the topic as legal and out-of-scope, returns a conservative refusal, and offers to transfer to a claims representative.
Human Handoff & Escalation
When configured, Brilo AI hands off to humans using explicit escalation rules:
Escalate immediately for matched high-priority keywords (for example, “reporting harm,” “fraud,” or “legal” when set as triggers).
Escalate after a set number of failed clarifications or when confidence remains below threshold.
Optionally capture a brief summary and caller context (intent, last N turns, tags) to pass to the human agent for faster resolution.
Handoffs can be configured to transfer to a queued agent, open an incident ticket in your CRM, or call a supervisor number. The Brilo AI voice agent tags and logs the escalation reason for auditability.
Setup Requirements
Define which topics and phrases are allowed, which are sensitive, and which must be declined (approved scope).
Provide your knowledge base (KB), CRM access, or grounding sources the agent may use to answer account-specific queries.
Configure confidence thresholds and maximum clarification attempts in the routing rules.
Upload pre-approved fallback responses and escalation messages for different out-of-scope categories.
Integrate your webhook endpoint or CRM for handoff routing and context passing.
Test scripted call scenarios that include out-of-scope prompts to confirm declines and handoffs behave as expected.
For guidance on session limits and capacity planning when enabling these controls, see: Brilo AI: Can the AI handle long conversations? and Brilo AI: How does performance scale with high call volume?
Business Outcomes
Out Of Scope Control reduces regulatory and reputational risk by preventing the Brilo AI voice agent from providing unsupported or sensitive advice. It improves auditability by tagging and recording every refusal and escalation. Operationally, it helps prioritize human effort toward interactions that require judgment, lowering the frequency of costly error remediation and supporting predictable automation coverage.
FAQs
How does Brilo AI decide what is out of scope?
You define the approved topics and sensitive categories. Brilo AI matches detected intents and keywords against that configuration and applies confidence checks and grounding requirements before answering.
Will customers see a record when the agent refuses?
Yes. Brilo AI logs decline events with a reason code and the conversation context you configure for audit and quality review.
Can the agent ask follow-up questions before declining?
Yes. You control the number and type of clarification questions. If the agent still cannot meet grounding or confidence rules, it will decline and follow the configured escalation path.
Does declining impact caller experience negatively?
When configured with concise, empathetic fallback responses and an immediate handoff option, declines minimize frustration and increase trust—especially in regulated contexts like patient or financial inquiries.
Can declines be customized per business unit or phone number?
Yes. Out Of Scope Control is configurable per routing profile so different business units can have different scopes, fallback responses, and escalation targets.
Next Step
In Brilo AI, fallback response is a pre-approved, conservative message the agent uses to refuse or defer a request to a human.
In Brilo AI, clarification limit is the configured maximum number of follow-up questions allowed before escalation or decline.
In Brilo AI, decline event is a logged record containing the refusal reason, conversation excerpt, and escalation action for audit and quality review.