A capital funding company generates leads from paid ads but was losing them to slow follow-up. So we built Alex, an AI voice agent that calls every new lead, confirms they got their product, qualifies them with two simple questions, and books the right ones straight into a strategist's calendar. After each call, Alex sends a fixed set of 24 structured fields to GoHighLevel, which decides the next step automatically. No one has to listen to a single recording.
The problem: leads go cold in hours
The client is a capital funding company. It helps business owners get funding and business credit. The leads come from Facebook and Instagram ads: people request or buy a digital product, a funding guide, a playbook, or a low-cost funding bundle.
The gap was speed. A buyer fills out the form at 9pm, and by the time anyone could follow up, the moment had passed. Many never opened the product email. Nobody was reaching out fast enough to find out who actually wanted funding and to get them on a strategist's calendar while they were still interested.
This is the quiet leak in almost every ad-driven business: the lead is most valuable in the first few minutes, and that is exactly when no human is available to call.
Meet Alex
Alex is an AI voice agent that phones every new lead automatically. In one short, natural call it does four things:
- Confirms the product arrived. Did they actually get what they signed up for? If not, Alex sorts it out on the spot.
- Separates real prospects from browsers. Who is actively looking for funding, and who is just researching?
- Books qualified owners into a strategist's calendar. Real time slots, booked live, during the call.
- Hands the full results to the CRM. So the right follow-up happens automatically, with context already loaded.
What a call actually sounds like
The whole conversation is designed to feel like a helpful human, not a phone tree.
- Opener. "Hi, this is Alex, an AI assistant calling on behalf of the company. Am I speaking with the right person?" Alex is open about being an AI and notes that the call is recorded.
- Delivery check. Did the product arrive? If not, Alex confirms which bundle and which email, then gets it resent.
- Interest check. Actively looking, just researching, interested later, already funded, or "take me off your list." Each answer is handled differently.
- Two qualifying questions, and only two. "When are you hoping to have the funding in place?" and "What is your credit score?" That is enough to route the lead correctly without an interrogation.
Smart routing, without playing judge
Alex never tells a caller "you qualify" or "you don't." It simply routes to the right next step:
- Strong fit (good score, funding needed soon): offered a strategist call.
- Interested but not a clear fit: still offered a strategist call. A human makes the judgment, not the bot.
- Preparation path (score needs work, or a longer timeline): offered a Strategic Action Plan or a preparation call, so they are not dropped, just pointed at the right first step.
- Not looking, or a hard no: a polite close. Do-not-call requests are respected immediately.
When a caller qualifies, Alex checks the strategist calendar, offers two real open time slots, and books the one the caller picks. Then it captures a little context, business type, use of funds, amount, and credit utilization, so the strategist walks into the meeting already knowing who they are talking to.
The part that matters: the call triggers the whole workflow
This is where a real agent separates from a voice toy. A toy answers the phone and the conversation dies there. An agent turns the conversation into structured data your systems can act on.
The moment an Alex call ends, it sends an end-of-call report to a GoHighLevel workflow with a fixed set of fields, the same 24 every single time:
- Qualified, booked, booking time, bundle resend requested
- Funding interest, timeline, credit score, routing outcome
- Appointment status, call notes, and more
GoHighLevel reads those fields and decides the next move on its own: send the booking confirmation, write the strategist a prep note, drop the lead into the right nurture sequence, or schedule a callback. One phone conversation sets off booking, CRM updates, and the correct follow-up, with no one touching a keyboard.
Most voice-agent demos sound great on the call, then the data goes nowhere. The value is not the voice. It is that every call produces clean, identical, structured data, so your CRM can run the rest of the business automatically. The call becomes the trigger for the entire workflow.
Built to be trusted
A voice agent that talks to your leads has to behave well every single time. These were deliberate design choices:
- Honest by design. Alex says it is an AI, notes the call is recorded, never claims information it does not have, and never promises funding.
- Graceful fallback. If the calendar is slow or a resend hiccups, Alex says so naturally ("our calendar is not loading on my end, so I will have a strategist reach out") and keeps the conversation moving. It never goes silent.
- Natural pacing. Alex waits a beat when callers pause mid-sentence or read out numbers, so it does not talk over them.
- Voicemail handling. If it reaches voicemail, Alex leaves a short, friendly message.
- Consistent data. The same structured fields come out of every call, so the CRM can automate without anyone reading transcripts.
The stack underneath
| Part | What it does |
|---|---|
| Vapi | The voice agent platform: speech-to-text, the AI model, and the voice |
| GPT-4.1 Mini | Alex's reasoning, the brain that decides what to say next |
| Twilio | The phone number for outbound and inbound calls |
| GoHighLevel | CRM, strategist calendar, contacts, and the workflows that start calls and act on results |
| Knowledge base | Approved answers to objections and product questions |
This works for more than funding
The shape is not specific to capital strategy. Any business with new leads that go cold can run the same play: call fast, confirm, qualify, book, and hand clean structured results to the CRM. Home service companies use the same pattern to catch missed calls and book jobs, which is exactly what our AI receptionist for home service businesses does.
If you have ever said "we lose leads because nobody follows up fast enough," this is the fix. It works the first night it catches a lead you would have lost to a slow callback.
Building the agent is the visible part. The real craft is the conversation design and the structured handoff to your systems, which is the kind of work we do in our Automation Builds engagements.