It was a candid look at how a once spreadsheet-driven operation transformed itself into one of South Africa’s most technologically advanced automotive businesses.

WeBuyCars began in 2001 as a modest two brother operation. By 2018 it was already moving about 2 000 vehicles a month, but according to him “the entire business ran off a Google Sheet”. That was the year private equity entered the picture, along with two immediate requirements: a CFO and somebody who understood digital transformation.

He accepted the challenge, despite coming from a technology background into what he described as “a car dealership with no systems”. But the lack of legacy software became an opportunity. “We had a blank canvas,” he said. “We could colour in the technology story the way the business needed it.”

The first major decision was whether to buy an enterprise system or build one from scratch. The team chose the harder path. “We did not want to adapt the business to the software. We wanted the software to work for the business,” he explained. The second principle was even more important: “We wanted to be in control of our data.”

This meant hiring developers and, crucially, a data specialist as one of the very first appointments. “If your data person is not there from day one, your pipeline and quality will suffer,” he warned. Clean, structured data would later become the backbone of the company’s AI capability.

Today the company calls its in house system the experimentation digital business platform. Every change, whether to an internal workflow or the public website, is tested through strict A/B experiments. “We try to take the emotion out of decision making,” he said. “Let the data decide.”

He made a clear distinction between type 1 decisions, which are slow, high risk and difficult to reverse, and type 2 decisions, which are quick and cheap to test. Buying and selling vehicles falls squarely into the second category, which makes it ideal territory for experimentation at scale.

From 2018 onwards WeBuyCars moved steadily through the layers of AI maturity. Machine learning now powers the company’s pricing engine, trained on hundreds of thousands of historical transactions. The shift to Bayesian statistical models allows the team to set buy and sell prices based on dynamic probability distributions rather than static depreciation curves. “Every vehicle has a price,” he said, “but you must be sure you are paying the right one.”

Computer vision models evaluate images of cars, detect damage, and even analyse undercarriage photos made from thousands of stitched images. These models feed directly back into pricing accuracy.

Generative AI supports operations ranging from customer conversations to automated social media responses. On the website, an AI assistant named Orange can explain vehicle features, compare models and even calculate the cost of driving between cities.

He also spoke about the next frontier: agentic AI. In future, autonomous software agents could buy or sell a vehicle on a customer’s behalf through a conversational interface. “The concept of the website as we know it will fade,” he predicted. Instead customers may rely on discoverable intelligent agents that negotiate with one another.

He ended with a simple message: focus on technology that improves unit economics. “Do not build technology for its own sake,” he said. And above all, embrace AI as a partner rather than a threat. “Your role may change,” he added, “but you are not being replaced. You are becoming more effective.”

This, he emphasised, is not the future but the present. “Everything I have spoken about today is already live. None of this is theoretical.”

(Photo: CIO South Africa).