Guides
Build an AI voice agent that calls every inbound insurance quote request in under 60 seconds, pre-qualifies shoppers by coverage type and risk profile, and routes warm prospects to licensed producers.
Published
Insurance leads are expensive. Whether they come from EverQuote, MediaAlpha, your own quote forms, or a referral partner, every inbound shopper represents ad spend, aggregator fees, or producer effort. Yet most insurance teams only follow up with the top 10% of leads — the ones who pick up on the first call and seem ready to bind. The other 90% go cold before a producer ever dials them.
A Thoughtly AI voice agentVoice agentA voice agent is a conversational software system that listens, responds, follows a defined workflow, and can take approved actions during a phone conversation. can call every inbound quote requestQuote requestA quote request is an inbound inquiry asking for pricing, coverage, or a cost estimate for a product or service. in under 60 seconds, pre-qualify the shopper's intent and fit, capture their risk profile, and route only the warm prospects to a licensed producer — or book a callback for later. This guide walks through building that agent step by step, using real Thoughtly product mechanics.
Before opening the Agent Builder, write out the questions your best producers ask on a first call. Insurance qualification is not BANTBANT qualificationBANT qualification evaluates a prospect’s budget, authority, need, and timeline before a sales handoff or appointment. — it's fit screening. You're checking whether this shopper can actually get a policy and whether they're worth a producer's time.
A typical auto insurance qualification flow might look like this:
For life insurance, swap in: age, health status, smoking status, coverage amount, term vs. whole life. For commercial: business type, number of employees, annual revenue, years in operation.
The Start node is your agent's opening line — spoken exactly as written every time. For insurance, this is where TCPATCPAThe TCPA is a US federal law governing certain calls and text messages, including restrictions tied to technology, consent, revocation, identification, and do-not-call protections. disclosure and consent captureConsent captureConsent capture records the permission a person gave for a specific type of contact, together with evidence of when, where, and how that permission was obtained. live.
In the Agent Builder, open the Start node and write a compliance-aware opener:
Hi, this is Alex with [Agency Name]. I'm calling about the quote request you submitted online. Before we begin, this call may be recorded for quality and training purposes. Is now a good time to talk, or would you prefer a callback?Key considerations for the Start node:
Each qualification question becomes a Speak node. Inside each Speak node, you'll add a Variable that extracts the caller's answer and an Outcome that routes to the next question.
Add variables from the right panel of a Speak node. Open the node, scroll to Variables, and click Add variable. Variables extract immediately after the caller's reply and before outcome evaluation, so your branching logic always sees fresh values.
For a ZIP code question, configure the variable like this:
Name: zip_code
Source: Current speak node
Extraction instructions: Extract the 5-digit ZIP code from the caller's response. If they give a 9-digit ZIP, use the first 5 digits. If they give a city name instead, leave empty.
Format: TextFor a coverage type question, use a prompt-based speak node that asks naturally, then extract:
Name: coverage_type
Source: Current speak node
Extraction instructions: Extract the type of insurance the caller is shopping for. Normalize to: "auto", "home", "bundle", "renters", "life", "commercial", or "other". If unclear, leave empty.
Format: TextFor a driver age or number-of-drivers question, use a Number variable:
Name: driver_count
Source: Current speak node
Extraction instructions: Extract the number of drivers the caller mentions. If they say "just me", return 1. If they say "me and my wife", return 2. If not provided, leave empty.
Format: NumberRepeat this pattern for each question in your qualification flow. Each Speak node asks one question, extracts one or two variables, and uses outcomes to route to the next step.
Outcomes tell your agent where to go next after a caller responds. You have two types: prompt-based (AI decides) and rule-based (deterministic checks). For insurance qualification, you'll use both.
Use prompt-based outcomes for open-ended questions where phrasing varies — like 'Are you interested in a callback?' or 'Would you like to compare quotes?' The AI interprets intent and selects the best matching outcome.
Use rule-based outcomes for structured checks where exactness matters — like ZIP code validation, eligibilityEligibilityEligibility is the set of conditions that determines whether a prospect can move to a service, quote, appointment, application, or specialist. gates, or compliance flags. Rule-based outcomes are 100% deterministic and recommended for critical paths.
After the ZIP code question, add rule-based outcomes to check service areaService areaA service area is the geography in which a business can provide a particular service, product, appointment, or licensed activity.:
Outcome A (checked first):
Rule: zip_code matches /^9[0-5]/ → Routes to: "Outside service area" End node
Outcome B (checked second):
Rule: zip_code is not empty → Routes to: "Ask coverage type" Speak node
Outcome C (Else/Default):
Rule: Always matches → Routes to: "Re-ask ZIP" (self-loop, max 2 retries)After all qualification questions are answered, add a final set of rule-based outcomes to classify the lead:
Outcome A: coverage_type is not empty AND driver_count >= 1 AND zip_code is not empty
→ Routes to: "Qualified — transfer or book"
Outcome B: caller_said_stop == true
→ Routes to: End node (conversation terminates)
Outcome C (Else):
→ Routes to: "Ask remaining questions" Speak nodeBefore transferring to a producer, your agent should check whether this lead already exists in your CRMCRMA CRM is the system used to manage leads, contacts, accounts, opportunities, activity, ownership, and follow-up. or AMSAMSAn AMS, or agency management system, is the operational system insurance agencies use for client, policy, carrier, document, activity, and servicing records.. This prevents duplicate records and lets the agent reference existing context — like a prior quote or a lapsed policy.
From a Speak node, open Actions and click Add new action. Choose your CRM integrationCRM integrationA CRM integration connects an agent or communications platform with a CRM so the systems can exchange records, events, and actions. (HubSpot, Salesforce, Pipedrive, or a webhookWebhookA webhook sends event data from one system to another through an HTTP request when a defined event occurs. for AMS systems like Applied Epic or Vertafore). Configure the lookup using the caller's phone number or email:
Action: HubSpot - Lookup Contact
Lookup field: phone
Value: {{system.contact.phone_number}}
Variables to set from result: existing_customer (boolean), lifecycle_stage (text), last_contacted (text)After the action runs, use rule-based outcomes to branch on the result. If the lookup finds an existing customer, the agent can say 'I see you've been with us before — I'll connect you with [producer name] who helped you last time.' If no match, proceed with fresh qualification.
Action outputs set or update variables like existing_customer and lifecycle_stage. These values are immediately available to rule-based outcomes in the same node.
When a lead passes qualification, two things can happen: a live warm transferWarm transferA warm transfer connects a live caller to the right person while preserving the context already collected during the conversation. to an available producer, or an automated appointment booking for a callback.
For live transfer, add a Transfer node. Thoughtly supports two transfer types:
For booking, add a scheduling action (Cal.com or Calendly) inside the Speak node where the caller agrees to a callback:
Action: Cal.com - Book Meeting
Event type: Insurance consultation (15 min)
Invitee name: {{caller_name}}
Invitee email: {{email}}
Output: booking_link, booking_timeAfter the booking action, the agent confirms the time with the caller and ends the conversation. The booking link and time are captured as variables and available for the post-call automation.
After every call, you need the conversation data — qualification answers, disposition, booking status — written back to your CRM or AMS. Thoughtly handles this with Automations triggered on On Call Completed.
In the Automations builder, create a new automation with the On Call Completed triggerTriggerA trigger is an event or condition that starts, resumes, changes, or stops an automated workflow.. Scope it to your insurance qualification agent. Add steps to:
Use Attributes for persistent facts (qualification_status, coverage_type, zip_code) and Metadata for per-call context (campaign, call variant). The automation can copy Metadata keys into Attributes after the call.
Asking too many questions before establishing value. If your agent's first 30 seconds are a questionnaire, callers hang up. Open with the reason for the call ('You requested a quote online — I can get you pricing in two minutes') before asking for details.
Using prompt-based outcomes for eligibility gates. If a rule must be exact — ZIP code in service area, driver count above zero, consent captured — use rule-based outcomes. Prompt-based outcomes are flexible but not deterministic. Compliance-critical paths require rules.
Not handling empty variables. If a caller doesn't answer a question, the variable stays empty. Without an Else/default outcome, the agent gets stuck. Add validation: if the variable is empty, loop back to re-ask (max two retries), then offer a callback.
Forgetting DNC and suppression. Insurance outbound is heavily regulated. Import your DNC list into Thoughtly's suppression management, set quiet hoursQuiet hoursQuiet hours are time windows when an outbound call or message should not be sent under applicable law, customer preference, or business policy. for your state's time zone, and ensure the Start node includes any required recording-consent script.
Transferring every qualified lead live. Not every producer is available. If you don't have a phone queue, route to booking instead of transfer. A scheduled callback converts better than a voicemail.
Once your agent is live, track these metrics in Thoughtly Analytics:
Review these weekly for the first month. If qualification rate is below 20%, tighten your variable extraction instructions. If transfer-to-bind is below your human baseline, your qualification criteriaQualification criteriaQualification criteria are the required signals an agent must collect before a lead can be booked, transferred, quoted, or routed. are too loose — add more gates.
Yes. Build a single agent with a prompt-based outcome after the first question ('What coverage are you shopping for?') that routes to different qualification branches — one for auto, one for home, one for life. Each branch has its own Speak nodes and variables. This keeps the conversation natural while still capturing line-specific data.
Thoughtly's insurance solution page references state-specific recording-consent disclosures, FCC-compliant time windows, and internal and external DNC scrubbingDNC scrubbingDNC scrubbing checks eligible outbound calling records against applicable national, state, and company-specific do-not-call lists before contact.. Configure quiet hours in your agent settings to enforce time-of-day windows per state. Import your DNC list into Thoughtly's suppression management. Use a verbatim Message speak node for the Start node to ensure the consent script is spoken exactly.
Thoughtly integrates with Applied Epic, Vertafore, EZLynx, and HawkSoft via webhook or Zapier. After the call, an On Call Completed automation sends the call summary and qualification data to your AMS webhook endpoint. For HubSpot or Salesforce, use the native CRM steps in the automation builder.
Add a prompt-based outcome like 'Already insured — comparison shopping' that routes to a comparison branch. The agent can ask about their current carrierCarrierA carrier is a telecommunications provider that transports voice calls or messages between phone numbers and networks., premium, and renewal date, then transfer to a producer who can quote against the existing policy. This is one of the highest-converting lead types in insurance.
You can. Inbound calls (callers dialing your published number) and outbound calls (automations dialing leads) can both route to the same agent. The qualification flow works identically. The difference is the trigger: inbound uses a phone number connection, outbound uses a Call Contact or Call Phone Number automation step with metadata like campaign source.