At Pendomonium 2023, there was no shortage of AI best practices and real-world examples for product teams. But an overarching theme rose to the top: AI does not mean the end of the product manager (PM) role.
If anything, the rise of generative AI unlocks new opportunities for product managers to operate more efficiently and strategically. This technology is also making PMs think about how they can build AI capabilities into their products—so that their users reap the benefits, too.
Before we explore why the PM is critical to the success of any AI initiative, let’s go over a framework that we use at Pendo when thinking about AI features.
The four levels of AI
In Level 1, there is no involvement from AI and everything is manual. This was product management before AI ever came about.
In Level 2, the process is still manual and driven by the human—but AI is there to assist, provide insights, and give recommendations. Amazon’s new AI-powered feature that summarizes product reviews is an example of a Level 2 feature. Instead of sifting through numerous reviews before purchasing, a customer can simply read an AI-generated summary.
In Level 3, AI drives the process and the human is there to fine-tune, edit, decline, and approve things. Netflix’s onboarding experience is an example of this. An individual creates an account, selects examples of the type of content they’re interested in, and then AI generates a personalized homepage with relevant recommendations. From there, the human selects what to watch.
And in Level 4, there is no human involvement and the process is fully automated by AI, like a self-driving car.
Three reasons why the PM is key to the success of AI features
As AI enters the business world at full force, a lot of companies are thinking about how they can use this technology to bolster their product and customer experience. While AI and machine learning are extremely powerful and the possibilities they present are nearly endless, the product manager is still the key to success. Here are three reasons why.
Reason 1: Moving from Level 2 to Level 3 is a product question, not a machine learning question
Let’s explore this first reason with an example from our product team at Pendo. They initially built an AI-powered feature that was at Level 2, because it involved AI assisting by automatically providing key insights about the given data. After working with design partners throughout the process, the team realized they needed to take this functionality to the next level and actually show users what to do with these insights.
The takeaway, though, is that the decision to take the AI-powered feature from Level 2 to Level 3 was driven by our product managers and UX designers, based on customer feedback. For the next iteration of the feature, the machine learning behind it stayed more or less the same, but the user experience was elevated—thanks to the product manager.
Reason 2: Great AI does not necessarily equal a great product
Consider this: What if the AI is 100% accurate, but it’s powering a poor user experience? This still results in a subpar product, no matter how powerful the AI technology is.
In the previous example, the issue didn’t lie in the AI. Rather, our product team needed to provide better tools and capabilities in order for our users to get the most out of the information AI was providing.
And since AI can’t ever really be 100% accurate, there are a few ways to ensure you’re still providing the best possible product experience:
- Be transparent and explain the AI model’s decisions whenever possible
- Allow the user to intervene and fine-tune the model’s decision
- Leverage feedback loops so the model can learn from feedback as you receive it
Reason 3: The product manager controls the roadmap
If you’re a product manager, you’ve probably found yourself thinking or responding to a request with: “We are at capacity.”
While innovation is exciting, it can also be a pain point because you already have an established roadmap, and more innovation often means more disruption to these plans.
But as the product manager, it’s your job to make room for innovation with AI. Here are three ways to go about this:
- Create a balanced roadmap—for example balancing proofs of concept and delivery work, as well as low and high risk projects
- Plan to be surprised and be prepared to change your roadmap
- Foster an innovative culture and encourage those around you to explore new ideas and AI technology
Tips for building AI-powered features
As you get started—or continue—building AI features in your own product, here are four tips from Inbal Budowski-Tal, Pendo’s senior director of machine learning.
First, avoid moonshots. It’s not easy to go straight to AI Level 3, so start with Level 2. Focus on finding ways to provide value to your customers, then build from there.
It’s also important to ensure you have the right people working on AI initiatives. Create working groups with engineers, data scientists, and designers so that there is a dedicated focus on building AI-powered products and features.
Remember to make room for innovation and embrace a culture of experimentation with AI. And when it comes to releasing AI features, leverage feedback—especially in the very early stages—to inform how you’ll tweak the algorithm to provide even more value to customers.
Want to dig deeper into AI’s place in product management? Take the AI for Product Management Course to learn practical AI use cases and best practices.
