Voice AI

AI voice agents for real estate, dental and inbound calls

AI voice agents that answer and make real phone calls. They call a new property lead back while the enquiry is still fresh, answer a dental practice’s phones when the desk cannot, book into the calendar you already use, and hand hard calls to a person with the context attached.

All services

Typically built with

  • Vapi / Twilio
  • Deepgram speech-to-text
  • Llama on Groq / OpenAI
  • Google Calendar / Cal.com
  • HubSpot / Salesforce

Scope

What can an AI voice agent do for a business?

An AI voice agent answers and makes phone calls on its own: it understands the caller, books into a real calendar, and passes the call to a person when it should. THWorks builds them for jobs where speed on the phone decides the outcome, such as calling back a property lead before they go cold or catching a dental patient who rings after hours.

The voice is the part people notice, but most of the build is the plumbing behind it. The agent has to check live availability before offering a slot, write the booking where the team will see it, log every call to the CRM, and know exactly which requests it must not handle. We have built three of these for clients: an outbound agent that phones new property leads and books viewings, an after-hours concierge for a dental practice that books by procedure, and an inbound assistant that transfers to a human without the caller repeating themselves. Each is written up below without the client’s name.

  • Outbound calls to new leads

    An agent that phones a new enquiry while it is still fresh, answers questions about the listing or service, and books the next step.

  • Inbound call answering

    Picks up when the front desk cannot, at night, at lunch or when every line is busy, takes the caller’s details and books them in.

  • Booking against live calendars

    Appointments go straight into the calendar the business already uses, such as Google Calendar, Cal.com or a practice system like Dentrix, by procedure or by agent.

  • Handoff to a person

    When a call needs a human, the agent transfers it with the conversation summarised, so the caller never repeats themselves.

  • CRM logging

    Every call is transcribed, summarised and written to the CRM, HubSpot or Salesforce among them, with the next action attached.

How the work runs

How a voice agent goes from first call to full line

We start from real calls: recordings or notes of the conversations the agent will have, and the systems it has to write to. The agent is built around those, tested against scripted and live calls, and launched on a slice of traffic before it takes the whole line.

  1. Map the calls: who rings, what they ask, what a good outcome is

  2. Connect the calendar, CRM or practice system the agent books into

  3. Test against real call scenarios, including the awkward ones

  4. Launch on part of the traffic, review transcripts, then widen

Before you ask

Questions we get.

  • What is an AI voice agent?

    An AI voice agent is software that holds a phone conversation on its own. It listens to the caller, turns their speech into text, decides what to say or do with a language model, and answers in a natural voice, fast enough that the exchange feels like a call rather than a menu. Unlike an old phone tree, it understands questions asked in the caller’s own words. THWorks builds voice agents that do a specific job end to end: calling back a new property lead, answering a dental practice’s phones after hours, or booking appointments and handing hard calls to a person. The agent works inside your systems, so a booking lands in your real calendar and the call is logged in your CRM.

  • How much does an AI voice agent cost?

    Two kinds of cost, and it helps to separate them. The build is priced by THWorks as a fixed price for the agreed scope, settled before work starts; what moves it is how many systems the agent connects to, such as a calendar, a CRM or practice software, how many call types it handles, and how much testing the conversations need. If a new call type or integration comes up midway, it is priced and agreed with you before it is built. Running costs are separate and usage-based: the telephony provider and the speech and language models bill per minute of conversation. We estimate those with you at scoping from your real call volume, so the monthly figure is known before launch.

  • Can an AI voice agent book appointments into our calendar?

    Yes, and booking is usually the point. The agent checks live availability in the calendar the business already uses, offers the caller real slots, and writes the confirmed appointment straight in, so nobody re-keys anything. We have built this against Google Calendar, Cal.com and a dental practice management system, Dentrix, where booking by procedure matters because a cleaning and a root canal need different chair time. The agent can also send a confirmation, take the intake details the practice needs before the visit, and reschedule when a caller asks.

  • Is an AI voice agent a good fit for a real estate agency?

    Often, because property leads go cold quickly. Someone who fills in an enquiry form is usually comparing several listings at once, and the agent who calls first has the advantage. A voice agent can phone a new lead while they are still on the website, answer questions about the listing from the details you provide, qualify their budget and timing, and book a viewing against the agent’s live calendar. The conversation is logged to the CRM, and anything it cannot answer goes to a person. It works best for agencies with steady online enquiries and too few people to call each one back straight away.

  • What happens when the agent cannot handle a call?

    It hands the call to a person rather than guessing. The agent is built with clear limits on what it handles, and when a caller asks for something outside them, is unhappy, or simply wants a human, it transfers the call. The person picking up gets a summary of the conversation so far, so the caller does not have to start again. If nobody is available, the agent takes a message with the details and books a callback. Every call is transcribed and logged, so the team can review what the agent handled and adjust its limits over time.

  • Will callers know they are talking to an AI?

    They should, and we build agents that say so near the start of the call. Being upfront tends to go better than callers working it out halfway through, and in a growing number of places disclosure is expected or required, so your legal team should confirm the rules where you operate. The voice and pacing can still be natural and on-brand. What matters to most callers is that the agent understands them, gets the booking right, and passes them to a person when they ask.

  • Do we need a voice agent, or would a booking link do?

    If most of your customers are happy to book online, a booking link and a reminder email are cheaper, and we will say so. A voice agent earns its cost when people prefer to call, when calls arrive while nobody can answer, or when speed matters, as it does with a fresh property lead. A useful test is how many calls go unanswered or how many enquiries wait more than an hour for a reply. If that number is small, start with the simpler fix.

Tell us what you are building.