Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More
In the race to develop innovative AI experiences, companies are investing heavily in various models and technologies. But what does it take to create an AI product that truly meets customer needs? Experts from Capital One, Pinterest, and Slack, who shared their experiences at VB Transform 2024, agree that cross-functional collaboration is key.
Deepak Agarwal, VP of Engineering at Pinterest, mentioned that building an AI product requires a team effort. He noted that it involves engineering, design, product development, data, and even legal aspects. Agarwal, with prior experience leading AI engineering at LinkedIn, emphasized that companies need to adopt an AI-first mindset. This means fostering a culture where entire teams work together to create remarkable experiences for customers.
Traditionally, software products were developed using a deterministic approach with standardized practices for development, testing, and iteration. However, with the advent of generative AI, the development process has become more complex and less predictable.
Developers now have to keep up with rapid innovation while also focusing on the quality, safety, and performance of their AI applications. They must consider various factors, including the AI model being used, the data involved, and how users interact with the system.
Jackie Rocca, VP of Product at Slack, explained that in the past, developers could use tools like Figma to get a good sense of what the user experience would be like. With AI and large language models, predicting outcomes is much more challenging, leading to a rapid prototyping environment that is highly iterative.
In this fast-paced environment, companies sometimes fail to bring together all the necessary teams to develop a consumer-facing AI product. Fahad Osmani, VP of AI/ML, Data, and Software Experience Design at Capital One, pointed out that teams often miss out on involving stakeholders responsible for risk assessment and compliance.
Osmani also noted that even when teams do collaborate, they may over-optimize within their own functions without considering the overall ecosystem.
To address these gaps, Rocca suggested that organizations should prioritize what’s important for customers while continuing to learn and iterate on their AI products. For example, instead of simply launching an AI chatbot, Slack took a different approach by addressing common user problems like information overload and difficulty in finding information. They introduced AI-powered search and summarization features to enhance user experience.
Osmani and Agarwal both advocated for cross-functional collaboration and problem discovery. They emphasized the importance of involving different roles early in the process and gathering feedback from various sources, such as A/B testing and telemetry, to understand the user’s context and needs before diving into development and deployment.
Osmani highlighted that engaging all relevant parties from the beginning—from problem definition to concept and usability testing—can lead to better outcomes. He found that having everyone in the same room early on yielded surprising and valuable insights.