MadRent is a South African vehicle rental company that came to CloudTalk for one thing: a dialer that worked natively with Pipedrive. What they built from there was something bigger. By connecting CloudTalk’s AI Voice Agent and outbound campaigns to a fully automated Pipedrive workflow, MadRent now runs 60,000+ outbound minutes a month — with fresh-lead conversions coming in at R31 per conversion, squarely inside what their best human agents would cost.
About MadRent
MadRent is a South African vehicle rental company operating alongside a vehicle dealership business. With multiple divisions — including sales, applications processing, servicing, and accounts — the company handles high lead volumes and document-heavy workflows.
Led operationally by Finance Director Wessel Davel, MadRent is deeply process-driven. The team relies on Pipedrive as its CRM and continuously experiments with automation to reduce manual overhead while maintaining strong customer engagement.
Problem
A team of five can only call so many people
MadRent generates 350 to 500 new leads every day. Each one needs a call: qualify interest, explain the product, move the deal forward. Wessel had a validation team of five doing exactly that. On a good day, working in parallel, they could reach around 800 people but only convert around 50. And it was enough… until it wasn’t.
The team could only call during working hours. Leads that came in overnight waited until morning. By then the moment had passed.
We used the dialer but that still meant we had to sync lists, add tags, adjusting stages. It worked, but it wasn’t scalable for our volume.
The answer wasn’t more people
Hiring more validators would solve the volume problem but not the timing problem — and it would add fixed cost that was hard to justify. There was also a quieter problem extra headcount didn’t fix: Wessel was spending time chasing his own team for status. Were all the leads called? How many times? What was the outcome on each one? He didn’t want to keep managing that loop by hand.
What Wessel needed was something that could call at any hour, handle every lead, report back cleanly without him having to ask, and do it at a cost comparable to what his team was already spending.
That pointed to AI Voice Agents.
Evaluation
Testing before committing
Wessel didn’t roll out the AI Voice Agent across everything at once. He ran it as an experiment first — alongside his existing team — and tracked every number he could find.
What was each call actually costing? What percentage were reaching real people? Was the AI qualifying accurately?
His benchmark was simple: what did a human validator cost per conversion? Between R20 and R40, depending on the lead. If the AI couldn’t match that, it wasn’t worth the switch.
The seven-ring test
The first thing he noticed was that a lot of calls weren’t reaching people at all. He called his own phone and timed it: in South Africa, voicemail kicks in around seven rings — 15–30 seconds depending on the carrier. The AI was ringing far longer than that on many calls, burning minutes on voicemail pickups it couldn’t use.
He pushed for a dial cap of 13 seconds — just inside the voicemail window. Before the cap, roughly half of all calls were short, a sign of voicemail. After, that dropped to around one in four.
Our voicemail hits at 15. I tested it on my phone, it was literally about seven rings. I want to cut it just short so we don’t hit it. Because then I’ll be more confident to call multiple times, we’re not hitting voicemail.
Old leads vs. fresh leads
He also tested targeting.
The AI was calling both fresh inquiries and older leads that hadn’t converted yet. When he ran the cost comparison, the split was stark: old leads were converting at around R90 per conversion. Fresh inquiries — people who had actively reached out within days — converted at R31. He cut the old list entirely and focused the AI on new leads only.
That is the decision that brought the cost per conversion inside the human benchmark.
Calling old leads was too expensive. Once we focused on fresh leads, the numbers made much more sense.
Almost pulled the plug
Not everyone was convinced. When early results came in slower than expected, management pushed to reverse course: move validators back into qualification and scale back the AI. Wessel held the line.
We almost had a knee-jerk reaction and thought of pulling back. But once you look at the data, you realize it’s not the AI, it’s about optimizing how you use it.
Solution
How it’s set up
CloudTalk’s AI Voice Agent connects directly to Pipedrive. Every new lead triggers an automatic call. The agent introduces MadRent, qualifies interest, and feeds the outcome back into the CRM — updating deal stages without any manual input. Leads that come in at midnight get called at midnight. That immediacy matters.
What’s running now
Outbound lead qualification
The agent handles around 1000 outbound calls on a busy day. Roughly 350 connect. Around 100 return a clear positive or negative outcome — qualified leads the sales team can act on immediately.
If we’re not hitting voicemail, the AI becomes extremely powerful — we were at about R31 per conversion, which is very good for us.
Document collection
MadRent had a backlog of a couple of thousand deals stuck at the “Missing Documents” stage — clients who had shown genuine intent but stalled on paperwork. When a deal hits that stage in Pipedrive, the AI calls automatically, tells the client exactly what’s missing, and directs them to WhatsApp to upload. The CRM updates itself.
These clients are already halfway to a deal. If we can automate that follow-up, conversion becomes much easier.
Inbound AI receptionist
MadRent is building an inbound AI receptionist to handle call routing, service bookings, arrears follow-ups, and common questions — drawing on a structured knowledge base of MadRent’s products and processes.
I want something that handles the volume. We’ll still rely on people — but the AI needs to manage scale.
Results
The numbers — and what moved them
MadRent’s human validators were costing between R20 and R40 per conversion. The AI, configured correctly and pointed at fresh leads, sits right in that range.
-
60,000+ outbound minutes per month
-
1000 outbound calls in a single day
-
1300 billable minutes processed in one day
-
~100 clear outcomes from 350 connected calls
-
R31 cost per conversion on fresh-lead campaigns — inside the R20–40 human benchmark
Two shifts matter as much as the unit economics. On a good day the validation team of five reached 800 leads and converted around 50; the AI now lands closer to 100 in the same window. The impact goes beyond the call data. With the AI handling qualification, Wessel moved his entire validation team of five into sales closing roles.
And Wessel no longer has to chase his team for status — whether every lead was called, how many times, what the outcome was — because the AI logs each call and updates Pipedrive automatically.
I’ve already moved all of my validators that used to do the qualification work. I’ve moved them all to sales now.
Next:
From here, the plan is to put voice agents into other high-volume parts of the business.
The vehicle dealership division is first — it receives around 900 leads a month, all specific inquiries about specific cars. Wessel is building an API-powered agent to handle those calls, pulling live inventory data to give callers accurate answers in real time.
From there, the focus shifts to existing customers. An accounts receivable agent will call clients automatically when their debit order (automatic payment) fails. After that, voice agents will roll out across customer care, customer service and the tracking unit.
It changes quickly. As it becomes more useful, we build it out, and it grows naturally.
