Trevo is a Polish e-commerce business selling packaged coal and wood for home heating — products so seasonal that three months of the year drive the majority of revenue, and three months see almost no activity at all. Before CloudTalk, Trevo ran an in-house call center of 50 agents, missed more than half of all inbound calls at peak, and had no way to reach lapsed customers at scale. With CloudTalk’s AI voice agent, a custom CRM integration, and outbound dialer, Trevo transformed their campaign model: now reaching 20 times more customers per campaign while cutting call time from three minutes to under 30 seconds for returning buyers.
About Trevo
Trevo is a Polish company selling packaged coal and wood for home heating — products that are deeply familiar to their customers, and deeply tied to the calendar. The business runs both an online store and a call center, because many of their customers simply prefer to pick up the phone and place an order directly.
The seasonal nature of the product shapes everything. Three months a year are peak. Three months see almost no activity at all. In between, the team runs targeted outbound campaigns to warm up demand before the rush hits.
Problem
A 50-Person Call Center That Still Couldn’t Keep Up
Before CloudTalk, Trevo had built and operated their own call center infrastructure — servers, software, and a team of 50 agents working from a central office. It was expensive to run, technically demanding, and inflexible.
Having our own solution was expensive. We had to always check everything — the technical stuff — and we still had to order numbers and pay per minute.
The deeper problem was structural. Fifty people made sense during peak season. The rest of the year, they didn’t.
During off-season there was nothing, so we just tried to figure out how to use this time with our employees. It wasn’t efficient.
During peak, even 50 agents weren’t enough. The team estimates they picked up less than 50% of inbound calls — and their system had no way to identify missed calls or create routing rules for follow-up. The result: a year where revenue hit 40 million, but 4 million was lost to operational inefficiency.
Manual Work at Every Step
Beyond the call volume problem, every order required manual effort after the call. An agent would speak to a customer, then stop to write notes, listen back to the recording, and manually create an order in their software. No automation. No integration. Just people doing what a system should.
That’s what Trevo set out to fix.
Evaluation
Looking for Automation First, Everything Else Second
When Trevo began evaluating alternatives, automation was the non-negotiable requirement. The goal was to connect call activity directly to their order management system — so a tagged call could trigger the right workflow without anyone touching it manually.
The mobile app mattered too. Moving to a distributed, remote team would let Trevo hire the right people wherever they lived, not just whoever was within commuting distance of the office.
Everybody thinks working in a call center is easy, and it is — but if you want to do it efficiently, you have to bring the right people.
CloudTalk’s API and the ability to build a custom CRM integration made the decision clear. Once they saw that call tags could flow directly into their order system, the business case was made.
Solution
A Custom Integration That Made Every Agent Smarter
The first major unlock was a custom integration between CloudTalk and Trevo’s internal software. When a returning customer called in, their full history appeared on screen before the agent said a word.
Somebody could pick up the phone and just say, ‘Dear XYZ, thank you for calling — do you want the same order as last year?’ And typically the answer was, ‘Oh wow, you know me. Yes, same product.’ Tag. Order.
The impact was immediate. A call that previously required asking for names, addresses, and order details — about three minutes — dropped to under 30 seconds for returning customers.
Smarter Campaigns, Smaller Peaks
Trevo also changed how they approached their twice-yearly outbound campaigns. Instead of blasting their entire database and creating an unmanageable demand spike, they now analyze call history, notes, and tags to build targeted lists with scheduled contact dates per customer. The power dialer handles the outbound volume. The result: the peak that used to overwhelm the team is now shaped, predictable, and manageable.
The AI Voice Agent: Reaching Who Humans Can’t
The final layer was CloudTalk’s AI voice agent — deployed on outbound cold calls to lapsed customers who hadn’t been contacted in five to ten years.
The team spent two to three months testing internally, correcting hallucinations by adding to the knowledge base and refining prompts. Customer reactions fell into three camps: short conversations where people didn’t seem to realize they were talking to AI; longer conversations — some up to ten minutes — where customers got everything they needed and didn’t care either way; and a smaller group who ended the call when they figured it out.
Someone spoke for five, ten minutes, asked many questions about products, and at the end said, ‘Am I speaking with a human or AI?’ The agent said, ‘I’m just a voice agent — would you like to speak with a colleague?’ And they said, ‘No, I got everything I needed. Thank you.’
Results
20x the Reach. Still Improving.
The AI voice agent converts at roughly 2-3% on cold calls, which is a bit less than Trevo’s best human agents, and Ariel is clear-eyed about it.
We are not at the moment where AI can read emotions and figure out what the person on the other side really needs. But I can call 20 times more people. So even if it’s less efficient, the math works.
Where humans can realistically cover 5% of the customer database during peak season, the AI voice agent can reach the full database — on schedule, without additional headcount, and without the seasonal hiring problem that used to cost millions.
-
2-3% conversion rate on AI cold calls vs. 5% human benchmark — a known gap Ariel is willing to trade for scale
-
20x more customers reached per campaign vs. human-only outbound capacity
-
Under 30 seconds per call for returning customers identified via custom CRM integration, down from ~3 minutes
-
Less than 50% inbound pickup during peak before CloudTalk — now fixed and managed with routing rules and missed-call follow-up that didn’t exist before
The inbound line stays human for now, but the outbound model is set — and ready to scale.
Takeaway
Trevo didn’t replace their team with AI. They gave their team the tools to stop doing what a machine can do, and used AI to reach the customers no human team could ever get to.
Maybe in two or three years, AI will be better than humans in complex conversations. I don’t know. But right now, I can call 20 times more people. That’s enough.
