Executive Conversations

From AI Experiments to AI at Scale: A Conversation with State Farm’s Fawad Ahmad

“This is not a hype cycle. The real benefits will be realized by those who move from experimentation to scale implementation.”

Par Rebecca Stephens-Wells


Fawad Ahmad believes that doing right by customers and employees requires embracing new technologies. He has a unique perspective on generative AI as State Farm’s current chief strategy and transformation officer and former chief digital officer. Fawad has been leading the company’s generative AI efforts, balancing thoughtful experimentation and capability building with bold moves, speed, and agility. Rebecca Stephens-Wells, a partner in our Financial Services practice, sat down with Fawad to hear about State Farm’s journey to date. Q: How do you see generative AI shaping the future of State Farm's business operations? Ahmad: At its core, insurance is a data-driven business. Generative AI has unlocked the ability to both harness unstructured data and create new content from data, both of which help us better serve customers. For example, digital knowledge assistants can help our employees more efficiently and effectively respond to our customers when they need us. This is not a hype cycle. There’s a lot of experimentation across all industries, but the real benefits will be realized by those who move from experimentation to scale implementation, who build AI capabilities, and who also prove that they can scale at pace. Q: Given the transformative potential of generative AI, how have you approached starting State Farm’s AI journey? Ahmad: We have a two-pronged approach. We’re intentionally building our capabilities in a crawl, walk, run fashion, ensuring that they can extend across different areas of our business. At the same time, we’re setting—and achieving—bold, near-term goals, such as scaling minimum viable product solutions in the first 12 to 18 months. One of the solutions we’ve introduced is a digital knowledge assistant to help employees in our contact centers more efficiently navigate our internal knowledge bases to handle inbound questions and interactions. Overall, we’re zeroing in on our biggest opportunities and challenges and identifying how generative AI can help us address them. Q: How has that informed how you brought your initial generative AI use cases to life? Ahmad: Three principles have guided our journey. First is building an intentional path to the right use cases. We start by evaluating potential starting points based on relative value to the business, complexity, and risk. We then focus on common patterns and the capabilities required for our highest-priority use cases. Focusing on building knowledge and skills was integral to our early efforts. We started with small-scale experimentation, with a primary goal of learning and capability building and a secondary goal of proving the business benefits and ultimately scaling the impact. A second fundamental principle is partnership. We ensure that our business partners and enterprise technology teams (including our data science and machine learning experts) are in close collaboration as they develop use cases. Finally, we’re intentional about kicking off and advancing our efforts at a pace that is ambitious but also consistent with our risk appetite. Q: How have you thought about managing risk and taking a responsible approach to deploying generative AI? Ahmad: We believe in being intentional about how we control for and manage the inherent risks of generative AI. We created a cross-functional risk team, which has been closely engaged in all phases of our journey. We also started our experimentation with lower-risk use cases, such as those that are co-piloted with a human and are focused on enhancing humans’ efficiency or effectiveness (vs. replacing them). We also prioritized use cases with less sensitive data such as internal knowledge bases vs. customers’ personally identifiable information. Q: What are your biggest learnings from the journey thus far? Ahmad: We’ve had a few big reflections. First, there is value in learning from others. Taking stock of what peers in and outside our industry are doing helps us stay current on the fast-paced and evolving landscape. We have also learned that we need to be thoughtful about what we build vs. buy, recognizing where there is an advantage to owning and controlling vs. leveraging others’ capabilities. Many of our core partners in technology and across the underwriting and claims ecosystems are becoming more AI-enabled. It’s important for us to understand what’s useful and proven—and then incorporate that appropriately into our decision making. Having an integrated team that moves in lock step is also key. The more we can shorten the distance between the user and the development team, the better the outcomes we will achieve. Lastly, driving adoption is as important a piece of the equation (if not more important) than building the solutions themselves. Engaging the target users of a solution from initial discovery and design to user acceptance testing, training, and rollout is critical to driving the behavioral changes needed for these solutions to be used consistently and drive real impact. Q: What’s next for State Farm from here? Ahmad: We are making significant strides against a portfolio of initial use cases. Our focus now is on continued optimization, enhancement, and scaling. In parallel, we need to think closely about the business problem and how we deploy the capabilities we have built (and will continue to build) to meet our biggest opportunities and challenges.

Fawad Ahmad believes that doing right by customers and employees requires embracing new technologies. He has a unique perspective on generative AI as State Farm’s current chief strategy and transformation officer and former chief digital officer. Fawad has been leading the company’s generative AI efforts, balancing thoughtful experimentation and capability building with bold moves, speed, and agility.

Rebecca Stephens-Wells, a partner in our Financial Services practice, sat down with Fawad to hear about State Farm’s journey to date.

Q: How do you see generative AI shaping the future of State Farm's business operations?

Ahmad: At its core, insurance is a data-driven business. Generative AI has unlocked the ability to both harness unstructured data and create new content from data, both of which help us better serve customers. For example, digital knowledge assistants can help our employees more efficiently and effectively respond to our customers when they need us.

This is not a hype cycle. There’s a lot of experimentation across all industries, but the real benefits will be realized by those who move from experimentation to scale implementation, who build AI capabilities, and who also prove that they can scale at pace.

Q: Given the transformative potential of generative AI, how have you approached starting State Farm’s AI journey?

Ahmad: We have a two-pronged approach. We’re intentionally building our capabilities in a crawl, walk, run fashion, ensuring that they can extend across different areas of our business. At the same time, we’re setting—and achieving—bold, near-term goals, such as scaling minimum viable product solutions in the first 12 to 18 months.

One of the solutions we’ve introduced is a digital knowledge assistant to help employees in our contact centers more efficiently navigate our internal knowledge bases to handle inbound questions and interactions. Overall, we’re zeroing in on our biggest opportunities and challenges and identifying how generative AI can help us address them.

Q: How has that informed how you brought your initial generative AI use cases to life?

Ahmad: Three principles have guided our journey. First is building an intentional path to the right use cases. We start by evaluating potential starting points based on relative value to the business, complexity, and risk. We then focus on common patterns and the capabilities required for our highest-priority use cases.

Focusing on building knowledge and skills was integral to our early efforts. We started with small-scale experimentation, with a primary goal of learning and capability building and a secondary goal of proving the business benefits and ultimately scaling the impact.

A second fundamental principle is partnership. We ensure that our business partners and enterprise technology teams (including our data science and machine learning experts) are in close collaboration as they develop use cases. Finally, we’re intentional about kicking off and advancing our efforts at a pace that is ambitious but also consistent with our risk appetite.

Q: How have you thought about managing risk and taking a responsible approach to deploying generative AI?

Ahmad: We believe in being intentional about how we control for and manage the inherent risks of generative AI. We created a cross-functional risk team, which has been closely engaged in all phases of our journey. We also started our experimentation with lower-risk use cases, such as those that are co-piloted with a human and are focused on enhancing humans’ efficiency or effectiveness (vs. replacing them). We also prioritized use cases with less sensitive data such as internal knowledge bases vs. customers’ personally identifiable information.

Q: What are your biggest learnings from the journey thus far?

Ahmad: We’ve had a few big reflections. First, there is value in learning from others. Taking stock of what peers in and outside our industry are doing helps us stay current on the fast-paced and evolving landscape.

We have also learned that we need to be thoughtful about what we build vs. buy, recognizing where there is an advantage to owning and controlling vs. leveraging others’ capabilities. Many of our core partners in technology and across the underwriting and claims ecosystems are becoming more AI-enabled. It’s important for us to understand what’s useful and proven—and then incorporate that appropriately into our decision making.

Having an integrated team that moves in lock step is also key. The more we can shorten the distance between the user and the development team, the better the outcomes we will achieve.

Lastly, driving adoption is as important a piece of the equation (if not more important) than building the solutions themselves. Engaging the target users of a solution from initial discovery and design to user acceptance testing, training, and rollout is critical to driving the behavioral changes needed for these solutions to be used consistently and drive real impact.

Q: What’s next for State Farm from here?

Ahmad: We are making significant strides against a portfolio of initial use cases. Our focus now is on continued optimization, enhancement, and scaling. In parallel, we need to think closely about the business problem and how we deploy the capabilities we have built (and will continue to build) to meet our biggest opportunities and challenges.