Skip to content
All case studies

Sun Speed CRM, Solar and home energy

A voice agent that qualifies solar leads and books the visit

A voice and CRM workflow for solar-sales enquiries, qualification, booking, and staff handoff.

Built for: A residential solar sales company handling inbound and follow-up calls.

agent tools for booking and transfer
4
agent tools for booking and transfer
booking steps: check, then confirm
2
booking steps: check, then confirm
OpenAI voice bridge
Realtime
OpenAI voice bridge
Get the AI Operations AssistantSee the AI Operations Assistant

First agent live in one week. A short call. Your first agent, a fixed quote, and the timeline. No obligation.

Sun Speed CRM case study overview: what we built, key numbers, and AI used
The situation

Sound familiar?

  • “Callers ask about financing, leases, and incentives our script can't answer.”
  • “Appointments get booked when the caller never actually agreed to a time.”
  • “Lead details from the call never make it into the CRM.”
What we built

How the system works.

  1. Step 1

    Real-time voice agent

    Twilio connects calls to an OpenAI Realtime voice bridge that qualifies on homeowner status, electricity bills, credit questions, and roof context.

  2. Step 2

    Four operational tools

    The agent can check availability, book a confirmed appointment, search the solar knowledge base, and transfer the caller to a person.

  3. Step 3

    Booking only on explicit yes

    Availability and booking are separate actions, and the agent will not book after a refusal such as "no" or "that doesn't work."

  4. Step 4

    Solar knowledge base

    PDF ingestion and vector search answer detailed questions on financing, PPAs, leases, installation, savings, and incentives.

  5. Step 5

    CRM and follow-up

    Leads and call history land in the CRM, with SMS templates for welcome, missed calls, confirmations, and reminders, plus email sequences.

Results

What changed.

  • 4 defined tools behind every call: availability, booking, knowledge search, and human transfer.
  • Bookings happen only after the caller explicitly confirms a time.
  • Every call is matched to the right lead record with its full history.
  • Tested across voice services, lead handling, scheduling, SMS, and email sequences.
AI used

The AI inside.

OpenAI Realtime

Holds the live voice conversation and decides when to call each tool.

Vector search

Retrieves answers from ingested solar product and financing documents.

Technology

The full stack.

Frontend

  • Next.js

Backend

  • Node.js
  • Express 5
  • WebSockets
  • Queues and schedulers

Data

  • Supabase
  • PostgreSQL

AI and automation

  • OpenAI Realtime
  • PDF ingestion
  • Vector search

Integrations

  • Twilio
  • Google Calendar API
  • Resend
Is this you?

This fits you if...

  • Your sales calls need real answers to detailed product and financing questions.
  • You want calendar bookings only when the customer clearly agrees.
  • Call outcomes should flow into your CRM and follow-up automatically.

Start with a free Agent Audit. First agent live in one week

The AI Operations Assistant brings this to your team.

An AI team member that handles the admin work across your email, documents, CRM, and spreadsheets, and asks before anything important goes out.

Get my free Agent AuditLearn more
Full project write-upRead it

Project overview

A solar-sales agent that can check the calendar and hand over the call

Services: AI Receptionists ; AI Agents ; RAG Knowledge Systems ; AI Workflow Automation ; AI ERP & CRM
Delivery: Voice calling, qualification, knowledge retrieval, appointment tools, and CRM follow-up.

Project Scope

Four defined agent tools:

  1. Check appointment availability.
  2. Book an explicitly confirmed appointment.
  3. Search the solar knowledge base.
  4. Transfer the caller to a human.

The surrounding workflow includes leads, call analysis, SMS, email sequences, and reminders.

The Problem

A solar conversation can move between qualification, financing questions, installation details, objections, and booking. A fixed script cannot supply every answer or determine whether a calendar slot is available.

The agent also needs to respect a refusal. Mentioning a time in conversation must not be treated as permission to book it.

What We Delivered

Twilio connects the phone call to an OpenAI Realtime voice bridge. The backend prepares lead context and exposes the four operational tools.

Qualification instructions cover homeowner status, electricity bills, credit-related questions, and roof context. Knowledge search supports questions about financing, PPAs, leases, installation, savings, and incentives.

The booking tool requires explicit confirmation of the appointment time and accepts time-zone information. A transfer tool provides a route to staff when a caller requests a person.

The CRM stores the lead and call history. SMS templates cover welcome messages, missed-call follow-up, confirmations, and reminders. Email templates and sequences support continued contact.

Engineering Decisions

Make booking an explicit action. Availability lookup and booking are separate tools. The booking instructions reject action after phrases such as "no" or "that doesn't work."

Retrieve detailed product information. PDF ingestion and vector-search infrastructure supply the knowledge tool.

Preserve lead identity. Phone-matching and context services connect the conversation to the right CRM record.

Test the supporting workflow. The repository contains tests for voice services, lead handling, context, scheduling, SMS, email sequences, and script configuration. Prompt tests check instructions; they do not replace live conversation tests.

Results

The delivered voice workflow connects a call to product knowledge, calendar actions, human transfer, and subsequent communication.

A seller can review the lead and call rather than receiving an isolated booking notification. The available records do not establish a booking-conversion increase or production call-capacity benchmark.

Technology stack

Next.js; Node.js, Express 5; Supabase/PostgreSQL; Twilio, WebSockets, OpenAI Realtime; PDF ingestion and vector search; Google Calendar APIs; Resend; queues and schedulers.

More work

Related case studies.

Connexions Recovery

Every admissions call becomes a record staff can act on

AI: OpenAI Realtime, OpenAI Whisper

Closeum

One system from the sales call to the commission

AI: OpenAI, Recall.ai

Planroom

Bid packages turned into answers with the source page attached

AI: OpenAI Responses API, LangGraph, Pinecone

Want this for your team?

Book a short audit call and show us the workflow. You leave with the first agent we would build, a fixed quote, and the timeline.

Get my free Agent Audit