CarMax is the largest used car retailer in the United States, with more than 240 stores across the country. In the past few years, the company has embraced generative artificial intelligence for tasks like aggregating content from user reviews and organizing descriptions of cars.
Shamim Mohammad, Chief Information and Technology Officer at CarMax, spoke with us recently about how the company sets up product-oriented teams for innovation, and its AI Center of Excellence, among other topics.
We spoke with Mohammad as part of our research report, Creating New Value in Large Organizations: What It Takes.
• • •
Could you tell me a bit about some of your recent AI projects?
At CarMax, we’ve been using data science, machine learning, and AI for a long time to optimize different parts of our business and experience for our customers. A couple of years ago, one of our teams identified that we could leverage generative AI to make car information easier for customers to review through the shopping process. They used GenAI to go through the thousands of customer reviews on our online channels and structure them in a way that’s easier for customers to read.
The second thing they did is use GenAI to organize vehicle information… so it’s easier for customers to identify them and make decisions based on that information. which also boosted our search engine ranking. We continue to refine this work to enhance customer engagement. Those two use cases were designed primarily for the customer experience, but they also helped us in search engine ranking.
The other area is a virtual assistant that we have rolled out to our customers…Traditionally a customer can just go to the store and talk to somebody who’s knowledgeable about whatever they have a question about, but we built this virtual assistant called Skye to give customers a way to get information while they’re shopping on their own.
And then there’s UVeye. Those cars that we sell through auction, we’re using AI to understand the condition of the car and making that available to the business customer to make an informed decision as to how much to bid. It’s a really interesting technology—you drive through an arch, and UVeye takes some pictures, and it can quickly assess the car based on the pictures and data it already has about the car, and comes up with a good report that can be used by B2B customers.
Most of the work we do that drives great experiences and innovation is done through product teams. Those are small teams, highly cross-functional, that focus on a business outcome goal…
What’s the process of moving from an idea to a finished product?
Over the last eight or nine years, we’ve moved to a product-based organization… most of the work we do that drives great experiences and innovation is done through product teams. Those are small teams, highly cross-functional, that focus on a business outcome goal… They have the ability to experiment, and test, and iterate with customers or end users, and figure out the best solution.
These teams have the freedom to experiment, iterate, and work directly with customers to find the best solutions… How they do this is not something we dictate. We give them the goal, and they figure out how to get it done. It allows them to be very empowered and very creative.
How do you decide which problems teams should focus on?
The business outcome goals are set at a high level in order to think about our overall strategy, and broken down into objectives… and then broken down so the teams have their own annual and quarterly objectives and key results (OKRs).
The teams [form] biweekly goals, so they look at two-week sprints, and can evaluate tests and how the customers are responding. And based on that they can iterate. Rather than going after one or two good ideas, they can go after a lot of ideas… get rid of the bad ideas quickly and focus on the good ideas.
It’s accelerating the pace of innovation, because we have empowered teams working on a lot of ideas at the same time, and the teams are very much focused on driving business outcomes. They’re not just doing technology for the sake of technology. They’re working on things that have meaningful business impact and meaningful impact to the overall goals of the company.
How do you stay updated on what the teams are working on?
Every other week, a majority of the teams are showcasing what they’re working on, how they’re progressing, what experiments they’ve done, how they track into their OKR goals, and what they’ll be doing for the next couple of weeks. It gives me and my C-Suite colleagues visibility into all the technology products and services that teams are working on, and the teams themselves can see that the leadership are very interested in the work they do… And the product [and leadership] teams get together every quarter and calibrate their priorities. We’re able to adjust and change our direction if we need to, based on external business factors or if priorities change. I think that is the biggest benefit of the way our teams work today.
The teams are between seven to nine people. We like to keep it very small, for a startup-like environment. Every team must have a product manager, a lead developer, and someone who is focusing on the experience of whatever we deliver.
For cross-functional teams, what skills do you look for?
Different products will have different cross-functional teams, but there are some things that every team must have. The teams are between seven to nine people. We like to keep it very small, for a startup-like environment. Every team must have a product manager, a lead developer, and someone who is focusing on the experience of whatever we deliver. And the product manager and lead developer will bring in different expertise depending on the product.
What’s really interesting in this model is the entire team work as one. Although they’re cross-functional, and they may have different reporting relationships, they have a common shared goal. The goal is to go through the ideas, come up with a minimum viable product you can get in front of a user, get feedback, and based on the feedback, make it better. And teams are generally durable—we don’t take one team and say, finish this product and move on to something else. We try to keep the teams together whether possible. There’s lots of camaraderie, and we want to keep that cohesiveness, so we’ll change the mission of the team, but not the team, sometimes.
How do you decide on team structures and focuses?
It all depends on the company’s mission, vision, and goals. Finance is a great example. We had one team working on a certain part of the finance experience, and later the need grew, so I think we have four finance teams now… The product managers and the product leadership teams get together and hold a [quarterly] planning session. But throughout the quarter, there’s a lot of conversation going on… because they’re agile product teams, they also do daily standups, so within the team they identify and raise anything that needs to be escalated or resolved.
How does the AI Center of Excellence fit into this structure?
Data is a very key asset for CarMax. We spent a lot of time over the last few years building good data governance, architecture, and data capability talent, and that allowed us to do a lot of AI on that strong data foundation. So, as AI was becoming more commonplace, we evolved the data governance into AI governance and built an AI Center of Excellence because we recognize that AI is a very rapidly evolving space.
The AI Center of Excellence is a very cross-functional team across the company. We have a steering committee… [and] a Use Case Advisory Team: somebody from legal, somebody from cybersecurity, somebody from our data practice, and someone from user experience. This small group is there to advise all teams—product teams, non-product teams, maybe not even technology—considering a particular use case with AI… Basically overall, AI governance is to encourage teams to use AI but do it in a very responsible, safe way.
Are there lessons you’ve learned about where AI is most useful?
AI is not new, and we’ve had good successes with AI so far. GenAI is something new… it’s evolving quickly, so that’s why having a mindset of experimenting and trying things out, validating, double and triple checking, and making sure you have good guidelines and training is important.
The key lesson… is we need to make sure we are using new technologies to improve or solve a real business problem—we need to solve for the business goal and drive competitive advantage and innovation for the company, versus just using the technology because it sounds cool and exciting. If the team did not use GenAI to organize customer reviews and vehicle information, we would [need] 200 or 300 content writers, and it would take them years to do the same thing. Now they can do the same volume of work in a matter of hours…And that’s really the beauty of using GenAI.
