In-house build

/ Case study / AI calling agent

An AI calling agent that rings every new lead within seconds

The AI calling agent is a phone system Safock built that makes and answers business calls in a natural, human-sounding voice. When a new request lands in a Google Sheet, form or CRM, it calls the customer within 2 to 3 seconds, books the slot or qualifies the lead, and writes the result back. Calls that need a person go to a live team member.

Book a 30-Min CallSend us a brief
Project
AI calling agent
Industry
Phone automation for clinics and service businesses
Stack
MakeVapiElevenLabsTwilioGoogle Sheets

What it includes

  • Calls new leads in seconds
  • Books, confirms and reminds
  • Human-sounding voice
  • Live transfer to staff
  • Outcomes logged to sheet or CRM
  • Many calls at once

/ How it works

From a booking form to a booked appointment, with no one on the phone

brightsmiledental.com

Book a visit

NameSara A.Phone(512) 555-0147I'd likeA cleaning
Request a booking

Step 1 of 5: Sara asks for an appointment. She fills in the clinic's booking form at 9:47 PM, long after the front desk has gone home.

An illustrative example, slowed down so you can follow it. In real use the phone rings 2 to 3 seconds after the form is sent.

/ At a glance

  • 012–3 secFrom new request to ringing phoneSteps one to seven of the workflow
  • 0224/7Calls answered and placedEvenings and weekends included
  • 03Four platforms, one job eachMake, Vapi, ElevenLabs, Twilio
  • 04Eight steps from request to logged outcome

Timings come from the system's workflow. The dental clinic in this case study is an illustrative example, not a named client, and no customer results are reported.

On this page
  1. 01Why small businesses lose customers on the phone
  2. 02Four platforms, one clear job each
  3. 03Speed to lead: a ringing phone 2 to 3 seconds after the request
  4. 04A dental clinic, before and after
  5. 05One system, many kinds of calls
  6. 06Live in five steps, with the rules built in
  7. 07What the agent changes for a small business

01The challenge

Why small businesses lose customers on the phone

Dental clinics, salons, agencies and service companies run on the phone, and rarely have enough people to answer it well. The receptionist is checking in a patient when the phone rings. A web form arrives at 9 PM and waits until morning. By the time someone calls back, the customer has often booked somewhere else.

Five problems come up again and again:

  1. 01

    Missed callsStaff are busy with customers, calls go to voicemail, and many callers never try again.

  2. 02

    Slow follow-upA lead fills in a web form and nobody calls for hours or days, by which time they have chosen a competitor.

  3. 03

    High staffing costExtra receptionists or call center agents are expensive, and call volume rises and falls through the day.

  4. 04

    Repetitive workConfirmations, reminders and the same questions over and over take up most of the team's time.

  5. 05

    No after-hours coverEvenings and weekends are when customers are free to call, and exactly when the business is closed.

02The solution

Four platforms, one clear job each

This is how we build AI voice agents for business: not a phone system from scratch, but four proven platforms that a small business can start on quickly and pay for by usage as it grows, with a Google Sheet or CRM as the record of every request and every result:

  1. 01

    Make, the brainIt watches for new requests, passes data between every tool and decides what happens next, from triggering the call to logging the result.

  2. 02

    Vapi, the agent managerIt runs the AI agents, for example one for booking, one for sales and one for support, manages each live conversation and transfers to a person when needed.

  3. 03

    ElevenLabs, the voiceIt gives the agents a warm, clear, human-sounding voice instead of a robotic one.

  4. 04

    Twilio, the phone lineIt provides the business phone numbers and dials and connects calls, inbound and outbound.

  5. 05

    Google Sheets or a CRM, the notebookRequests come in here, and every outcome is saved here: booked, interested, call back later or not interested.

03How it works

Speed to lead: a ringing phone 2 to 3 seconds after the request

Speed to lead is the time between a customer asking for something and the business first calling them back. The AI calling agent keeps it to about 2 to 3 seconds. Every call follows the same eight steps, and steps one to seven happen in that window, so the phone rings while the customer's interest is at its highest. The replay at the top of this page runs the same sequence, slowed down:

  1. 01

    TriggerA new request lands in a Google Sheet, form, CRM or booking system with the customer's name, number and need.

  2. 02

    Make picks it upThe Make scenario detects the new row and reads the details.

  3. 03

    The agent is preparedMake calls the Vapi API to set up the right agent with its script, business details, customer context and ElevenLabs voice.

  4. 04

    Back to MakeVapi returns the agent and call details, and Make checks them and prepares the call.

  5. 05

    Twilio dialsMake tells Twilio to call the customer from the business's own phone number.

  6. 06

    Make connectsThe live call is linked to the AI agent, so the conversation starts the moment the customer answers.

  7. 07

    The conversationThe agent answers questions, books the slot or qualifies the lead in a natural voice.

  8. 08

    Results and handoffMake logs the outcome and sends confirmations, and Vapi transfers complex calls to a person on the team.

04How it works

A dental clinic, before and after

Take a three-dentist practice with one receptionist and a steady stream of online booking requests. Before the agent, the receptionist answers the phone while checking in patients, calls go unanswered at busy times, online requests get a call back the next day, and reminders happen by hand when there is time, so no-shows are common.

Working as an AI appointment booking agent, the calling agent changes the clinic's day in four ways:

  1. 01

    Every web request gets a call within seconds, even late at night.

  2. 02

    The agent confirms every appointment the day before.

  3. 03

    Cancelled slots are offered to the waiting list straight away.

  4. 04

    Emergencies and insurance questions go to the receptionist, with a summary of the call.

In the example call, a patient named Sara submits the clinic's web form at 9:47 PM, and her phone rings three seconds later. The agent offers Thursday at 10 AM or Friday at 4 PM. She picks Friday, and the agent books her with the dentist and tells her a confirmation text is on its way. The sheet row is marked Booked, and the receptionist never touched the phone.

05How it works

One system, many kinds of calls

The same flow serves any business that books, follows up or answers the same questions every day, from clinics and other healthcare practices to sales teams. Only the script and the connected tools change, so one AI phone agent covers all of these:

  1. 01

    Sales teamsEvery new lead from ads, forms or landing pages gets a call within seconds. The agent qualifies budget, need and timing, then books a meeting or passes hot leads to a salesperson.

  2. 02

    Customer supportOrder status, opening hours, pricing and simple troubleshooting are answered day and night, and complex issues reach a person with full context.

  3. 03

    Call centersThe agent takes first-level calls and runs outbound campaigns for reminders, renewals, surveys and payment follow-ups.

  4. 04

    Appointment businessesClinics, salons, physiotherapists, law firms, repair services and property viewings all use the same book, confirm and remind flow.

06How it works

Live in five steps, with the rules built in

Each rollout starts with one use case, usually reminders or lead follow-up, and grows from there:

  1. 01

    Map the callsList the call types the business handles most: bookings, reminders, lead follow-up and common questions.

  2. 02

    Write the scriptSet the greeting, the questions, the business information and when to transfer to a person.

  3. 03

    Connect the toolsLink the Google Sheet or CRM with Make, Vapi, ElevenLabs and Twilio.

  4. 04

    Test with real scenariosRun test calls and tune the voice, wording and timing.

  5. 05

    Go live and growStart with one use case, review the results weekly, then expand.

The agent tells callers they are speaking with an AI assistant where local rules require it. Call recording and data storage follow privacy law, including healthcare rules for clinics, and outbound calls respect consent and do-not-call rules.

07The outcome

What the agent changes for a small business

This case study describes the system we built and how it runs. It reports no measured customer results.

The design gives a business specific things. A new request gets a call in 2 to 3 seconds instead of hours. The phone is answered around the clock, and many calls run at once instead of one. Reminders and instantly refilled cancellations are there to cut no-shows, every outcome is logged, and a new location or campaign means a script change rather than a new hire.

Staff stop spending the day on routine calls and spend it on the customers in front of them.

/ Services used

What this project drew on

/ FAQ

About this project

What is speed to lead, and how fast is the AI calling agent?

Speed to lead is how quickly a business first contacts a new lead after they ask for something. The AI calling agent calls about 2 to 3 seconds after the request lands: Make detects the new row, Vapi prepares the agent with its script and voice, and Twilio dials from the business's own number, so the phone rings while the customer is still thinking about the enquiry.

Will callers know they are talking to an AI?

Where local rules require it, the agent says so at the start of the call. We also set up call recording, data storage and outbound calling to follow privacy, consent and do-not-call rules, including healthcare rules for clinics.

What happens when a call needs a person?

Vapi transfers the call to a live team member. A clinic's agent, for example, passes emergencies and insurance questions to the receptionist with a summary of the call, so the caller does not have to repeat themselves.

Can the AI calling agent work with our CRM instead of Google Sheets?

Yes. Make connects to most CRMs, booking systems and forms, so requests can come from where they already arrive and outcomes are written back to the same place. We map your call types and systems during discovery and give a fixed quote for the setup.

How is an AI calling agent paid for?

By usage. The platforms charge for the calls and automations the agent actually runs, so many calls can happen at once without adding salaries. We estimate running costs from your call volume during discovery.

/ Keep reading

Other systems we have shipped

All case studies
usebollard.com
The Bollard website at usebollard.com, presenting procurement and inventory software for multi-site operators
Next case study / Our product

Bollard

Our own procurement and inventory software for multi-site operators: seven modules, controls enforced in the database, and an append-only audit trail.

Read the case study
  1. EduHubWeb3 developer tooling
  2. DyorAICrypto trading and market research
  3. TalknWeb3 video streaming
  4. AI readiness scorecardAI strategy for businesses of any size

/ Next step

Want every new lead called back in seconds?

Tell us what you are working on. You will speak with a founder, and you will leave the call with a clear next step, even if that step is not us.

Book a 30-Min CallWhatsApp us

Prefer writing? Send us a brief and a founder will reply.

WhatsApp usBook a call