Platform thinking in the age of Generative AI: A practical guide for Product Managers

Imagine this: You’ve created a popular AI product. Users love it. Growth is steady. And then one day, your CEO says: “How do we make this a platform play?” Sound familiar? This is a conversation playing out in product teams everywhere today in 2024 and will be the theme for 2025. And for good reason – the most valuable businesses today aren’t products, they’re platforms. But here’s the challenge: Everybody talks about platform thinking but very few product managers have had a real framework to switch that way of thinking, particularly in the Generative AI era. Let's change that.

Do you recall when Perplexity was just another AI search engine? It's now evolving into an incredibly robust platform that allows developers to create custom search experiences powered by Perplexity's API, publishers to aggregate specialized knowledge bases, and users to curate collections and the newest addition-perplexity shopping. This evolution is a quintessential example of the modern platform evolution.

That pivot happens when you stop asking “How do we capture all the value?” and instead ask, "How can we empower others to add value?”. Much like how Figma evolved from a design tool to a thriving ecosystem of plugins and creators, the shift happens when you stop thinking about building everything and start enabling others to build with you. Read more.

Figure 1. Platform evolution pathway

As a product manager who has launched and scaled AI-powered platforms, there are three essential elements of the platform. Let’s distil these into actionable pieces you can use:

Consider this to be the atomic unit of exchange for your platform. For Perplexity, search queries, and AI-based answers. For your platform, consider: “What is the smallest unit that could be considered a value that would pass from participant to participant?”

Get it down on paper: Summarize your value unit in one sentence. If not, it isn’t clear enough.

This is where many platforms fail. You must make it incredibly easy for the two sides to create and capture value. Perplexity accomplishes this expertly: Developers receive strong search functions without having to train their own models, while users get access to highly nuanced and specific search throughout diverse areas.

This is where your AI capabilities come heavily into play. With every interaction, the platform gets better, bringing network effects creating exponential value to the overall ecosystem. Your infrastructure needs to:

The magic of Perplexity's platform is how its learning loops work in a self-reinforcing way. Every user query improves search precision, and curated collections enhance context for richer interactions. If developers make domain-specific search experiences, their expertise goes back into the core system.

Let’s break this down to be super actionable. Here’s your how-to guide, in six steps:

Start by answering these key  questions:

Create two columns:

For each group, list:

Figure X. Flow of network effects

Draw your network loops:

This is where AI gives you a unique advantage. Start with:

The classic chicken-and-egg problem takes on new dimensions in the AI era. As Lenny Rachitsky explores in his widely-read newsletter, successful platforms must solve this fundamental challenge: how to attract producers without consumers, and consumers without producers. However, in 2024, AI offers unique solutions to this age-old dilemma. Try following Perplexity’s approach for example:

And each step fed on the previous step producing more and more powerful network effects.

Tips for successfully executing

Platform thinking in 2024 is not just limited to connecting the two sides of the marketplace, it is about enabling an ecosystem for AI and customer (or platform participants) to create compounding value together. Keep in mind: Every successful platform began as a product. Solve one problem well, and then scale — indiscriminately. Leverage AI as an enabler — not just as a feature — to enable network effects to create value for all participants.

Disclaimer: The views and opinions expressed in this article are those of the authors solely and do not reflect the official policy or position of any institution, employer, or organization with which the authors may be affiliated.