Quote request intake
A quote request often arrives before the caller knows which details matter. They may describe the outcome they want but omit the service location, timeline, quantity, dimensions, or constraints needed to estimate the work. If the call reaches voicemail, the sales or estimating team inherits a name and number instead of a quote-ready brief.
An AI phone agent can turn that first conversation into structured intake. It asks the business's approved questions, confirms the answers, and sends the request down a defined path. It does not calculate a price unless an attached, business-controlled system exposes an approved operation for that purpose. More often, its job is to prepare a complete record for an estimator, schedule a site visit, or explain which detail is still missing. The exact questions vary across home services, equipment rental, agencies, and other fields; the industry guides cover that context without changing the core intake pattern.
A concrete quote request call
Imagine a property manager calling a commercial cleaning company.
Agent: Thank you for calling Clearline Services. I am the company's AI phone assistant. I can collect the details needed for a quote. What type of space needs service?
Caller: A small office building. We need cleaning during the week.
Agent: What is the service address, approximate area, number of restrooms, preferred service days, and desired start date?
Caller: The building is in Oakland. I do not know the exact area, but it has two floors and four restrooms.
Agent: That is fine. I will mark the area as unknown rather than estimate it. Would you like me to check times for a site visit so the team can assess the building?
The agent collects the caller's name, company, phone number, email address, access constraints, and requested schedule. It repeats the address and contact details back. The missing area triggers the business's site-visit branch, so the agent uses Cal.com or Google Calendar to offer approved times. If the caller instead describes work outside the service area, the agent follows the business's out-of-scope language and records the disposition without inventing an exception.
This is lead qualification in a narrow form: apply explicit routing rules to a request while leaving exceptions and commercial judgment with people.
How to configure the workflow on ThunderPhone
Build the prompt around the inputs an estimator actually uses. A useful outline includes:
- a disclosure and a concise explanation that the agent is collecting quote details;
- the service or product categories callers may request;
- required fields shared by every request;
- conditional questions for location, dimensions, quantity, timing, access, current setup, or supporting context;
- explicit handling for unknown answers, so the agent records “unknown” instead of guessing;
- rules for a site visit, sales appointment, routine follow-up, or human transfer; and
- prohibited promises, including unapproved prices, availability, completion dates, and scope commitments.
Attach approved service descriptions, coverage areas, preparation instructions, and policy documents to the agent's knowledge base. ThunderPhone supports text, Markdown, CSV, PDF, and DOCX uploads and can search selected documents during a call. Use that material to answer stable questions, not to convert an old brochure into a binding quote.
For testing, assign an inbound demo U.S. number. Demo numbers are inbound-only and not intended for production. For live traffic, bring a production number through a supported direct VoIP connection or manual SIP configuration. Verified imported numbers support inbound and outbound calls. If the workflow requires an attended transfer to an estimator, use an eligible production number with outbound SIP capability.
Connect the systems that own the next step. HubSpot or Salesforce can support the sales record. Google Sheets can support a simpler intake queue. Cal.com or Google Calendar can support site visits and discovery calls. Slack can notify the responsible team, while REST or webhooks can deliver a completed call record to an existing estimating workflow. If the business exposes approved tools through Zapier MCP or another remote MCP server, ThunderPhone can add the server by URL, synchronize tool discovery, and attach those tools to the agent. Streamable HTTP is required.
Test more than the happy path. Run simulations where the caller lacks a measurement, changes the service address, asks for an immediate price, requests unsupported work, or reaches a failed scheduling or record-creation step. ThunderPhone supports reusable scenarios, graded call logs, regression suites, and minimum pass-rate gates. Simulations are billable real calls, with the charge shown before a run.
What to measure
The main question is whether a human can act on the record without repeating the call. Track:
- quote requests with all required fields or an explicit “unknown” value;
- time from the call to estimator review;
- site visits or discovery calls successfully scheduled;
- duplicate, misrouted, and out-of-area requests;
- records returned for clarification; and
- transfer and integration outcomes.
Review those measures by request type. A high clarification rate for one service usually points to a missing conditional question, while a long intake across every service may mean the prompt asks for details too early. Related workflows include lead qualification calls and appointment scheduling by phone.
Where a human remains necessary
Estimators should own pricing, feasibility, discounts, contractual terms, and unusual scope. A caller may also be unable to describe the site accurately, or the connected system may not expose enough current information to complete the request. In those cases, the agent should summarize what it knows, identify what is missing, and arrange a human next step rather than fill gaps with assumptions. The receiving CRM integration should remain a workflow aid, not the authority for a promise made to a caller.