CarMax’s top tech exec shares his keys to reinventing a legacy retailer in the age of AI

Mohammad spoke with Modern Retail at the Virginia-based company’s technology office in Plano, Texas, which it opened three to four years ago to recruit for tech workers like software engineers and analysts in the region home to tech companies such as AT&T and Texas Instruments. At that office, CarMax has since hired almost 150 employees — more than initially expected — including some of Mohammad’s former colleagues from Blockbuster, which he had worked for in Texas in the early 2000s.

He explained how other legacy retailers can learn from how CarMax leveraged new technology like artificial intelligence and a startup mindset as it embraced change, becoming an omnichannel retailer where customers can buy cars in person, entirely online or through a combination of both. Many customers find a car online and test-drive and complete their purchase at the store.

“Every company, every industry is going through a lot of disruption because of technology,” Mohammad said. “It’s much better to do self-disruption: changing our own business model, challenging ourselves and going through the pain of change before we are disrupted by somebody else.”

Digitizing the dealership

Mohammad has been with CarMax for more than 12 years and had also been vp of information technology for BJ’s Wholesale Club. Since joining the auto retailer, he and his team have worked to use artificial intelligence to fully digitize the process of car buying, which is especially complex given the mountain of vehicle information and regulations dealers have to consider.

He said the company has been using AI and machine learning for at least 12-13 years to price cars, make sure the right information is online for the cars, and understand where cars need to be in the supply chain and when. That, he said, has powered the company’s website in becoming a virtual showroom that helps customers understand the vehicles, their functions and how they fit their needs. Artificial intelligence has also powered its online instant offer tool for selling cars, giving customers a fair price that doesn’t lose the company money, Mohammad said.

“Technology is enabling different types of experiences, and it’s setting new expectations, and new types of ways to shop and buy. Our industry is no different. We wanted to be that disruptor,” Mohammad said. “We want to make sure we change our business model and we bring those experiences so that we continue to remain the market leader in our industry.”

About three or four years ago, CarMax was an early adopter of ChatGPT, using it to organize data on the different features of car models and make it presentable through its digital channels. Around the same time, the company also used generative AI to comb through and summarize thousands of customer product reviews — it did what would have taken hundreds of content writers more than 10 years to do in a matter of days, he said — and keep them up to date.

As the technology has improved over the last few years, the company has adopted several new AI-powered features. One is Rhodes, a tool associates use to get support and information they need to help customers, which launched about a year ago, Mohammad said. It uses a large language model combining CarMax data with outside information like state or federal rules and regulations to help employees quickly access that data.

Anything that requires a lot of human workload and mental capacity can be automated, he said, from looking at invoices and documents to generating code for developers and engineers, saving them time to do more valuable work. Retailers like Target and Walmart have done the same by using AI chatbots as tools for employees.

“We used to spend a fortune on employee training, and employees only retained and reliably repeated a small percentage of what we trained,” said Jason Goldberg, chief commerce strategy officer for Publicis Groupe. “Increasingly, AI is letting us give way better tools to the salespeople, to train them and to support them when they’re talking to customers.”

In just the last few months, Mohammad said, CarMax has been rolling out an agentic version of a previous buying and selling assistant on its website called Skye that better understands the intent of the user — not only answering the question the customer asks directly, but also walking the customer through the entire car buying process.

“It’ll obviously answer [the customer’s question], but it will also try to understand what you’re trying to do and help you proactively through the entire process. It could be financing; it could be buying; it could be selling; it could be making an appointment; it could be just information about the car and safety,” he said.

The new Skye is more like talking to an actual human being, Mohammad said, where, in addition to answering the question, the agent can make other recommendations in a more natural conversation. For example, if someone is trying to buy a car and asks for a family car that’s safe, it will pull one from its inventory, but it may also ask if they’d like to talk to someone or even how their day is going.

“It’s guiding you through the process beyond what you initially asked. It’s building a rapport with you,” Mohammad said. “It knows you very well, it knows our business really well, and then it’s really helping you get to the right car and the right process.”

Goldberg said that while many functions of retail, from writing copy to scheduling shifts, have also been improved with AI, pushing things done by humans to AI chatbots could lead to distrust or create results that are inappropriate or offensive. “At the moment, most of the AI things are about efficiency and reducing friction,” Goldberg said. “They’re taking something you’re already doing and making it easier, which is generally appealing, but there is also the potential to dehumanize the experience.”

In testing CarMax’s new assistant, other AI agents are actually monitoring it to make sure it’s up to the company’s standards and not saying bad words, Mohammad said, adding it would be impossible for humans to look at everything the new assistant is doing.

The company doesn’t implement AI just to implement AI, Mohammad said, adding that his teams are using generative AI as a tool when needing to solve particular problems instead of being forced to use it.

“Companies don’t need an AI strategy. … They need a strategy that uses AI,” Mohammad said. “Use AI to solve customer problems.”

Working like a tech startup

In embracing change, CarMax has had to change the way it works, Mohammad said. It has created a more startup-like culture, going from cubicles to more open, collaborative office spaces where employees know what everyone else is working on.

About a decade ago, he said, the company started working with a project-based mindset, where it would deliver a new project every six to nine months — each taking about a year in total, with phases for designing and testing.

Now, the company has small, cross-functional product teams of seven to nine people, each with a mission around improving a particular area like finance, digital merchandising, SEO, logistics or supply chain — some even have fun names like “Ace” or “Top Gun.”

Teams have just two weeks to create a prototype of a feature and get it in front of customers. He said that, stacked up over time, those small new changes those teams created completely transformed the business.

“The teams are empowered, and they’re given a mission. I’m not telling them what to do. I’m giving them a goal. They figure out how,” Mohammad said. “Create a culture of experimentation, and don’t wait for things to be perfect. Create a culture where your teams are empowered. It’s OK for them to make mistakes; it’s OK for them to learn from their mistakes.”