Turning Generative AI Into A Product: It Looks Nothing Like You’d Expect
Infinite technology of the universe. We’re driving around in a Ford POS. It’s a line from a movie, but it could just as well apply to many of our AI-supported products today. They took an average car and added advanced functionality to it. The modified car made no sense. There were buttons everywhere. Each did something different, and everything was documented in tribal knowledge alone.
Generative AI and foundational models provide the potential for broad and advanced functionality. Unfortunately, the products we wrap around them don’t meet expectations. They are a lot like that car but often even more confusing.
Amazon’s Alexa is a good case in point. You can ask for something as mundane as ‘What time does the Costco pharmacy close today?’ and Alexa cannot help you. It’s a real-world example of a use case that should and can be supported but isn’t.
The problem is that Alexa’s a platform in search of a product. I’ll use Alexa to explain how LLMs should become products that actually generate revenue but often won’t. Amazon has struggled with the Alexa business unit, and we will see its pattern of mistakes repeated with this new wave of Generative AI products.
What’s Going Wrong? There’s No Alignment Or Product Vision.
The root cause for why so many common and easily supported use cases are left on the sidelines is a lack of alignment across a product. In my Data and AI Product Management course, I teach a framework to overcome this challenge that leverages a platform strategy.
Platforms are a way to unify functionality across multiple workflows and customer segments. Amazon’s Alexa group has been challenged to create a coherent, unified product vision. When the device or platform is capable of a massive functional range, how does the business decide what the platform should support?
For Amazon, and many others building Generative AI products today, functionality is implemented to support several different workflows and needs. However, they don’t connect with and support each other. As a result, the experience feels inconsistent or incomplete. Google often ships products like this too. There’s an innovative core, but the implementation is partial.
In this post, I will start with the pharmacy hours use case and build a product around it that lives on the Alexa platform. I will explain how the product will be monetized, deliver near-term value, and incrementally evolves to support a long-term vision.
Without the platform framework, implementations begin and end as one-off implementations. Customers don’t adopt the partial solution, and the business can’t monetize it. Without the vision, platforms like Alexa are built as a series of these one-offs and never become coherent products.
Thinking About Platforms Because The Product Is Never Done
A platform vision is powerful because it creates a single framework for products to be built around. The vision supports the big picture and more granular initiatives. Products must be built and delivered incrementally. Each initiative must follow a long-term plan for what the product will do in the next one to two years.
Products are never finished in a multi-technology environment. Products that start digital scaled into the cloud, and many now include data or analytics features. Just as the cloud didn’t replace digital altogether, data and AI will not replace earlier technologies. Data and AI are not the last technologies that products will integrate to support new features.