There is a lack of hype about the possibility that AI will convert workplaces. A recent McKinsey Report compared the technology to the birth of the Internet and the arrival of steam engines.
But that reality is still in shape. AI adoption is inconsistent in most organizations, workers have different levels of interest, and there is often a difference between the topic of AI and its practical application.
Cargurus, an online marketplace for buying and selling cars, is one company that is trying to bridge that split. Last October, we launched AI Forward. This is a 20-person working group that brings together leaders across the division, including products, engineering, law, and sales. The group's goal is to identify the right applications for AI, assess potential tools, and encourage employee experimentation through workshops, one-on-one guidance, and pilot programs.
“If everyone has to find their own AI tools, we risk losing interest,” said Sarah Rich, senior former data scientist at Cargurus and lead coordinator at AI Forward. “We're offering cheat sheets and trying to share what's working.”
She added that once employees understand how AI can make it more efficient on a day, or offer new approaches, they tend to board. “We want to make sure that when we ask people to invest their time in AI, they'll see the reward right away.”
Rich spoke with Business Insider about how AI Forward can help employees gain confidence in exploring technology.
The following has been edited for clarity and length.
Business Insider: What was the reason for AI Forward?
Sarah Rich: There's a lot of pressure to move AI forward. And I think this is true for many businesses – if you don't keep up, there's a sense that you're leaving innovation at the table. At the same time, there is a gap between excitement about AI and understanding the meaning of each role.
We have moved AI forward to meet all our business units and capabilities. This group will work together to evaluate use cases and AI tools. This is important given how quickly AI is evolving and the constant onslaught of capabilities. The group also offers structured support to help employees learn how to use the tools.
How often did the group meet and what was your first business order?
We meet monthly as a group, in between with intensive sessions within each department.
One of the first things I did was meet individually with the leaders to identify some solid use cases that could really move the needle for the team. Some were ready to go. Others didn't know where to start. We spent a lot of time brainstorming, realising where the underlying technology was, and realising that some features still lacked technology.
However, other features include engineering coding tools and natural language-based solutions for reviewing legal contracts.
What happens next?
We carve out the time and space for people to experiment. The engineering team runs office hours and jam sessions, which are essentially open collaborations, so that they can learn coding tools such as Cursor and Windsurf. We also held AI coding week to allow everyone to start using AI tools at work.
LLM solutions are effective for labor-intensive, language-centric tasks. When teams experiment with these tools, they can see how quickly your work accelerates. Make time for the experiment. It doesn't just happen. But usually people see what impresses them, and AI starts selling themselves.
What does the group do to support employees who are open to AI?
People are in different places on the curve of adoption and enthusiasm. Some people are excited about the open-ended jam session. Others need a structure where they need to try out the tools on ticket-style tasks and assigned tasks and projects.
Our group learned that there is a need for different levels of delivery. It's important that everyone comes to some extent, but not everyone has the same level of enthusiasm. That's fine.
How do you measure the success of AI forwards?
We track some metrics: how often people use AI, the tools they use, the confidence to use it safely, and the overall sentiment about AI.
Many focus on adoption in terms of efficiency or saving time, but people tend to misjudge it. AI may not always save time, but before you're sure you've landed on the best product, you may find it helpful to create a better product as you explored six different directions to test your options. Beware of emotions as AI can feel destructive and threatening. Push AI without acknowledging that the nuances tones feel deaf.
What did you learn from AI Forward?
Patterns appear in the data in three phases. First of all, people feel enthusiastic because AI is said to be magic and solve everything. Secondly, there is this middle ground disillusionment, people interacting with AI tools in some way, but have not been able to stand the hype or endure the hype. There is a story about AI that reinforces jobs and expands them.
The ideal third stage occurs when people start using AI and don't feel threatened by it. They see it makes them better in their work. They also say that without real people, AI can't do meaningful, impactful work.
Emotions depend on where an individual or team is in their adoption efforts and how successful they have in finding the right use case. Based on internal data spanning the use of AI productivity tools across the enterprise, procurement requests for new AI products, and anecdotes from the entire team, it is clear that the majority of employees have tried AI at least in their daily work.
What advice would you give to companies that want to start a similar AI working group?
AI is novel in many ways, but it's also quite familiar, especially with how it affects people psychologically and emotionally.
We tend to get caught up in technology, but the real challenge is human. I recommend focusing on them: bring people together, make them feel safe, and give them a reason and space to pay attention. Rather than alienating you, you need to feel good and encourage.
