Why Cars24 Can Become the World’s First Consumer AI Company

I think Cars24 can become the world’s first consumer AI company. By that I do not mean adding an AI assistant to an app. I mean rebuilding the work behind buying and owning a car around intelligence that can understand each vehicle, remember each customer’s context and coordinate what happens next. The opportunity comes from the complexity we have spent eleven years trying to manage.

Internet software works well where products are standard and the steps can be defined in advance. With used cars, the vehicle and the customer’s circumstances can change what needs to happen.

Imagine someone in their first job opening our app at night to look at the Swift they hope to afford after a few more months of saving. They may want advice well before they are ready to buy. When they are ready, they will need help comparing cars, understanding condition, arranging finance and completing the transfer. Some buyers will also have an old car to exchange. The questions continue after purchase. For many people this is one of the largest purchases they will make, and they have good reason to worry about getting it wrong.

That journey can last for months. Much of the product is not a screen in our app. It is the conversation around the decision: a call, a WhatsApp exchange, a question about finance, a worry that appears after a test drive. A customer should not have to restart that conversation every time the channel or the person changes.

The car they are looking at is not a product in the way a phone is. Two cars of the same model and year can differ in owners, service history, accident record, tyres, battery, paperwork and wear. Each car is its own SKU with a stock of exactly one. We have to work out what that particular car is worth, who might buy it, what risks it carries and what needs repairing.

Behind that one car and that one customer sits an operation. A car has to be inspected, valued, moved, reconditioned, parked, photographed, financed, documented, delivered and serviced. Each step has its own exceptions and its own handoffs between teams. Across the cars we handle, that becomes thousands of workflows running at once. An app can help coordinate this work, but the cars still have to be repaired, moved and delivered.

What happens to the car affects everything around it. A question about its condition can change the price or the loan, and someone has to explain that to a customer who may have spent months deciding. People have to work out what the new information means and what to do next. Much of running this business comes down to those judgments.

For most of our history, more judgment meant more people. Handling an unfamiliar case often meant another step or handoff, and another place where context could be lost. I have to own my part in that. As transactions grew, I let operations take priority over engineering. By 2023 it was clear that buying and selling more cars was not enough on its own, for growth, for economics or for the experience we were delivering. We had to change how the company worked while continuing to run it every day.

Our first wave of progress had come from digitising the market. We built systems for inspection, pricing, transactions, documents, lending, transfer and ownership. Those systems helped us keep track of the cars and the work around them as the business grew. But software of that kind works when the answer can be reduced to a rule, and the hard cases here rarely can. Is the damage cosmetic or structural? Does a mismatch between two documents matter? Is this customer confused, worried, or simply not ready? Was a promise made in a phone call that never entered the system? Wherever our systems met that kind of ambiguity, a person stepped in and carried the context by hand to the next step.

What AI changes is where that boundary sits. Digitisation gave us systems to record the work and execute steps we could define in advance. Automation repeats those defined steps. It did not remove the judgments buried in photographs, video, calls, WhatsApp threads and scanned documents. An AI-native system can work with that raw material as it actually exists and decide that the right action is no action. Within the right limits it can help form a view, and it can call the transaction systems that turn that view into an action. People still own the consequential decisions and the exceptions that need accountability.

Inspection is where this is most visible so far. Between April and June 2026, AI assisted 80 percent of our inspections, and the average inspection came down from 45 minutes to 26. In the same quarter, AI assisted at major checkpoints in 65 percent of our loan disbursals. These numbers show that AI is being used across a substantial part of our work. They do not tell us whether inspections are more accurate or customers trust us more, and they do not mean the work runs on its own. We remain responsible for what we sell.

We are now seeing a second kind of proof in the customer journey. Our long-running sales agent carries context from clickstream, calls, WhatsApp and app behaviour. It decides whether to call, write, change what a customer sees or leave the customer alone. In a recent two-week period, it made about 800,000 decisions. An early internal comparison from that period reported a 230 percent improvement in conversion versus the human sales team. Among the actions it chose, the majority were decisions not to contact the customer. Those are encouraging signs, not proof of a rebuilt journey or a result we can assume will hold at scale. But they show something important: intelligence is not the same as sending more messages.

That is the shape of the company we are trying to build. For the customer, AI can stay present through the months before they are ready, answer plainly without pressure, and hold the context so that nobody explains the same exchange, loan or delivery five times as channels and people change. For the car, it can help build a fuller view of condition, price, demand, fraud and financing risk for one specific vehicle. For the operation, it can connect the images, documents and exceptions that today sit in separate places, so that a problem found in one workflow actually triggers the next and someone confirms it was closed.

For years, handling unfamiliar cases added to the cost of running the business. If we record how we resolved them and test whether those lessons apply elsewhere, that experience can help us handle the next case. The improvement depends on doing that work. It does not follow simply from using AI.

Our advantage cannot come just from having data or access to foundation models. We also need to know what happened after a decision: whether the repair solved the problem, whether the documents were accepted, whether the transfer finished. Over eleven years, we have built the transaction systems, inspection processes, lending relationships, ownership products and showrooms where that work happens. We still have to connect that experience to the AI systems and show that it improves the work.

That proof cannot be how much AI we consume. It has to show up as better pricing, more consistent inspections, fewer cars sitting idle, less repeated contact, quicker finance decisions and problems resolved before the customer discovers them. Above all it has to reduce uncertainty. People do not care what model we run. They care whether the car is what we said, whether the price makes sense, whether the finance is ready, and whether someone will take responsibility when something goes wrong. That is why I keep saying that in our category, premium is certainty.

Someone can reasonably say that ChatGPT is already a consumer AI product. It is. I mean something different: a consumer company serving an old, physical, high-stakes need, with its operating layer rebuilt around AI. “The world’s first consumer AI company” is an ambition, not a trophy. We will earn the phrase through customer and operating outcomes, or not at all.

We are not there yet. Early systems across inspection, lending and customer assistance are not a rebuilt company, and the test has to be met across the whole journey.

Someone who has been thinking about a car for months finally decides. The car in front of them is exactly what we said it was. The finance is ready, the papers move, and the transfer finishes on the day we promised. They drive away without ever having to notice the thousands of small judgments that had to go right for that to happen.