Voice AI Intake Agent Developer — Retell / Vapi — Phase 1 Standalone Test

Job Type:
Project-based / paid test first

Project Overview:

I am looking for a Voice AI developer to build a **standalone Phase 1 intake-only voice agent** for a mobile mechanic / field service business.

This is separate from my n8n / FieldPulse / CallTrackingMetrics automation build. Another developer is currently handling the first n8n automation proof test.

For this role, I want to start with a small paid Voice AI test that proves the agent can answer calls, collect basic service request information, classify the call, and produce structured output that can later connect to n8n, CTM, and FieldPulse.

Preferred platforms: **Retell AI or Vapi**. I am open to either, but I want the developer to explain which platform is better for this Phase 1 intake-only use case.

## Important Scope Boundary

This first Voice AI agent is **intake only**.

The agent must not:

* diagnose mechanical problems,
* estimate repair cost,
* quote prices,
* schedule appointments,
* dispatch technicians,
* assign technicians,
* promise arrival times,
* promise same-day service,
* create live FieldPulse jobs,
* make commitments on behalf of the business.

The agent’s job is to collect information and route the call for human review/follow-up.

## What the Agent Should Do

The Voice AI agent should:

* Answer incoming calls professionally.
* Explain that it will collect information so someone can follow up.
* Collect required intake details.
* Classify the call type.
* Flag incomplete calls for human review.
* Produce a transcript, summary, classification, missing fields, and structured JSON output.
* Send or simulate an email recap and SMS alert during the test.
* Be designed so it can later connect to n8n, CallTrackingMetrics, and FieldPulse.

## Required Intake Fields

The agent should attempt to collect:

* Caller name
* Callback phone number
* Whether the callback number can receive texts
Suggested wording: “Is this a mobile number we can text if needed?”
* City / ZIP where the vehicle or equipment is located
* Vehicle or equipment description
* Brief description of the issue
* Urgency
* Roadside / safety status
* Existing customer vs. new caller if clear

## Call Classifications

The agent should classify calls into categories such as:

* Qualified service request
* Incomplete intake / needs human review
* Roadside or safety review
* Existing customer follow-up
* Non-service call / sales / spam
* Out-of-area or maybe-out-of-area

For incomplete calls, the agent should list the missing fields and mark the call as:

**Needs Human Review — Incomplete Intake**

## Structured Output

The test should produce structured output that can later be used by n8n.

Example fields should include:

* caller_name
* callback_number
* can_text_callback_number: true / false / null
* city
* zip
* vehicle_or_equipment
* issue_summary
* urgency
* roadside_or_safety_status
* classification
* incomplete_intake
* missing_required_fields
* human_review_required
* transcript
* summary
* recording_or_call_link if available

## First Paid Test Deliverables

For the first paid test, I expect:

1. Voice AI agent configured in my account.
2. Intake-only prompt / rules / behavior.
3. Guardrails to prevent diagnosis, pricing, scheduling, dispatching, or response-time promises.
4. 8–10 sample test calls.
5. Transcript and summary output.
6. Classification and missing-field handling.
7. Structured JSON output.
8. Email recap and SMS alert, or clearly mocked/sample alert output if full alert setup is not included.
9. Final prompt/configuration.
10. Setup notes.
11. Loom walkthrough explaining how the agent works.
12. Explanation of what I can safely edit later myself, such as script wording, approved responses, alert recipients, and minor rule changes.

## Future Expansion

Later phases may include:

* Connection to n8n
* CallTrackingMetrics tracking number integration
* FieldPulse-ready customer/job payloads
* Better alert routing
* More advanced review workflow
* Miro flowchart update for the Voice AI portion

These are not required for the first test unless specifically agreed.

## Ideal Candidate

The ideal candidate has experience with:

* Retell AI and/or Vapi
* Voice AI prompt design
* Intake call agents
* Call classification
* Webhooks
* JSON outputs
* n8n or Make
* Twilio or SMS alert workflows
* CRM or field service workflow integrations
* Human-review routing
* Clear documentation and Loom walkthroughs

Direct experience with FieldPulse or CallTrackingMetrics is helpful but not required.

## Please Answer These Questions

Start your reply with:

**I can build the Voice AI intake agent.**

Then answer:

1. Have you built voice agents using Retell, Vapi, or another voice AI platform?
2. Which platform do you recommend for this Phase 1 intake-only agent, and why?
3. How would you prevent the agent from diagnosing, estimating, scheduling, dispatching, assigning technicians, or promising response times?
4. How would you handle incomplete calls and missing required fields?
5. Can you produce structured JSON output for future n8n integration?
6. Can you include the field `can_text_callback_number: true / false / null`?
7. Can you send or simulate email recap and SMS alert output during the test?
8. What accounts/access would you need fro ---------- before starting?
9. Can the agent be built inside my account so I own the configuration?
10. What simple things would I be able to edit later myself?
11. Can you provide a Loom walkthrough at the end?
12. What fixed price and timeline would you quote for the first paid test?
13. Are you willing to send a short video reply before I make a final decision?

## Budget

Please quote a fixed price for the first paid test.

I am not looking for a full production system yet. I want a small controlled proof test first. If the test goes well, there may be additional work.