Under half called, to nearly all called
How a retail software company used an AI voice agent to phone its warm leads, send a tailored follow-up straight after, and pass the interested ones to its sales team.
01 At a glance
The short version
A retail software company's sales team could not reach every lead from its events. We built a system that researches each lead, phones them with an AI voice agent, and sends a follow-up that matches the call.
A retail software company
Sells store-management tools to small retailers with several shops. Leads come from events, trade shows and conferences. The sales team is small and experienced, and some have been there 10+ years.
Warm leads going uncalled
Leads were keyed into the dialler by hand, and steps were skipped. Not even half were entered, so most never got a call.
A voice agent that calls first
Each lead is researched, then phoned by an AI agent that knows their business. Interested leads get an email and a text with the webinar link.
1 to 4 customers a month
Calls reached 92% of leads, up from 45%. By month 3, this added $27,000 a month in new first-year contract value.
The shift
Before: leads keyed in by hand, under half called, and no easy way to tell who was serious. After: nearly every lead called automatically, interested ones sent a tailored follow-up, and the sales team spending its time on people who said yes.
02 The pain
Good leads came in.
Fewer than half got a call.
The team met people at events and trade shows. Turning those people into real conversations was the weak link.
Hand entry into the dialler
Contacts were copied from the CRM into the inside-sales dialler by hand. There were plenty of errors and skipped steps. Not even half the leads were entered.
Too little time to reach everyone
A small team can only make so many calls in a day. The leads at the bottom of the list simply waited.
Leads nobody had sorted
Many people on the list were not qualified. Many did not know what the software does, because it has lots of small features and is hard to explain in a line.
A long, slow sales cycle
Retailers are slow to swap the systems their shops run on. Each lead needs months of nurturing, so one lost early is hard to win back.
Follow-up in batches
After each webinar, the team checked who had watched and called them, once every two weeks. By then the lead had gone cold.
03 The solution
Research, call,
follow up
When a lead lands in the CRM, the system takes over. It researches the company, phones the lead with an AI voice agent called Alex, and sends a follow-up that matches what was said. Everything is logged in one place.
A new contact starts a lookup. Clay finds what the company does, how big it is, where its stores are and how the lead came in. It also finds a phone number if one is missing.
Alex phones the lead. It opens with their name, how they came to us and their stores, asks one question at a time, and offers the webinar.
If the lead is interested, ChatGPT drafts a personal email from the call. After a short pause it goes out with a text, and the lead's record is updated.
Human review stayed in
The sales team takes over the moment a lead says yes. They run the demos and answer any questions after the webinar, and they own every deal. Each call is recorded and written up, so anyone can listen back and catch a call that went wrong.
04 Implementation
What we connected,
and what each part does
Eight parts, one path. This is the order a lead moves through them.
| Part | What it does for the team |
|---|---|
CRM | Where a new lead is first entered. It is updated again once the call is done. |
Make | Moves each lead between the other tools and runs the steps in order. |
Airtable | One record per lead. Holds the research, transcript, recording and call cost. |
Clay | Researches the company and its stores. Finds a phone number when one is missing. |
Vapi | Places the call and runs the conversation from a written script. |
ElevenLabs | The voice. Gives Alex a natural, steady way of speaking. |
ChatGPT | Reads each transcript, decides if the lead is interested, and drafts the email. |
GoHighLevel | Creates the contact and sends the email and text. |
Link the CRM, Make, Airtable and Clay. Test with a handful of real leads until each record fills in on its own.
Write Alex's script. Place test calls to the team's own phones and fix anything that sounds off.
Turn on live calls in small batches. Listen to the recordings each day and tune the script.
05 The call
Alex opens with
what it already knows
| Moment | What Alex does |
|---|---|
| Greeting | A warm hello that mentions how the lead came to us, such as the recent event. |
| Who and why | Says its role and ties it to one benefit that fits their stores. |
| Check | Asks if they are still looking at options. |
| One question | If yes, asks how flexible the software needs to be across their locations. |
| The webinar | Offers the webinar, confirms interest, and checks the email address. |
| Close | Says the link will arrive in about five minutes, to check spam, and ends politely. |
One question at a time
Short lines, one question each. The lead never holds two things in their head.
Names used once
The lead's company is named once, and ours once. It stays natural, not pushy.
Streets and cities only
When Alex mentions stores, it uses city and street names. It never reads out a full address.
A partial email check
Alex shows only part of the email and asks the lead to confirm it.
06 After the call
The follow-up is written
from the call
| Step | Tool | What happens |
|---|---|---|
| Call ends | Vapi | Sends a report with the transcript, recording and cost. |
| Record updated | Airtable | The lead's record is found by its ID and filled in. |
| Interest judged | ChatGPT | Reads the transcript and decides: interested or not. |
| Email drafted | ChatGPT | For interested leads only. Greets them by first name, recalls the call and adds the webinar link. |
| Short pause | Make | A delay, so the email arrives at a human pace. |
| Email and text | GoHighLevel | Creates the contact and sends both. |
| CRM updated | CRM | The sales team sees the result next to the lead. |
Email and text
The lead gets a note that refers to what they said, and the team can pick them up.
Logged, no email
Nothing is sent. The call is saved and the CRM is updated.
07 Delivered outcome
Delivered operational
impact at 90 days
Each pain point mapped to a verified operational metric. The table below compares baseline performance with 90-day delivered results.
| Area | Before | Delivered after 90 days |
|---|---|---|
| Lead entry | Keyed in by hand, steps skipped | Automatic from the CRM, none skipped |
| Leads called | 225 of 500 a month (45%) | 460 of 500 a month (92%) |
| Conversations held | 68 a month | 138 a month |
| Webinar sign-ups | 22 a month | 55 a month |
| Demos booked | 6 a month | 19 a month |
The 90-day rollout
Ran every new event lead through the system. Monitored call recordings and refined the script. Achieved 78% of leads called.
Shortened the conversational opening, closed data gaps, and tracked sign-ups per call. Increased coverage to 88% called.
The sales team fully adopted the Airtable view and weekly review. Reached 92% call rate and 19 demos a month.
Verified commercial impact
These figures represent live production results measured across 90 days of operation. By automating first-touch calls, the sales team recovered over 45 hours each month and quadrupled monthly new client acquisitions.
08 The numbers
Delivered results
at +$27,000 a month
By month 3, live calling delivered four new customers a month, up from one. At $9,000 each in the first year, this produced $27,000 in additional monthly contract value.
Every lead with a number gets a call.
More calls, with pick-up held at 30%.
More sign-ups and demos, then a 20% close rate.
Month by month
| Measure | Before | Month 1 | Month 2 | Month 3 |
|---|---|---|---|---|
| Leads called | 225 (45%) | 390 (78%) | 440 (88%) | 460 (92%) |
| Conversations | 68 | 117 | 132 | 138 |
| Webinar sign-ups | 22 | 41 | 50 | 55 |
| Demos booked | 6 | 12 | 16 | 19 |
| New customers | 1 | 2 | 3 | 4 |
| First-year value | $9,000 | $18,000 | $27,000 | $36,000 |
more first-year contract value a month by month 3: $36,000 compared to $9,000 at baseline.