Another Horrible Idea: The AI Product Manager!
It’s not a real job!
Background
As best as I can estimate, sometime in 2020 or just before, the term “AI Product Manager” started appearing in popular IT literature and discourse. Personally, I wasn’t too aware of the term until a good data scientist friend of mine brought it up in a recent discussion. I looked it up and discovered that Kellogg at Northwestern has a programme for AI Product Manager. IBM has a professional certification course for it. Clearly there must be a demand for such a role, otherwise, why invent it? Here’s my take: people are so desperate to be part of the AI movement that they will try to enter the “community” in any way they can. IT folks want to be seen as AI experts. Business folks want to position themselves as AI-savvy business advisors. Everyone is trying to extract their “bag of money” from an already seriously inflated domain.
And so I dedicate my 84th article to calling bullshit on this emerging industry nonsense.
(I write a weekly series of articles where I call out bad thinking and bad practices in data analytics / data science which you can find here.)
AI Product Manager
This is what some of the courses on AI product management teach:
- Duke University — “Manage the Design & Development of ML Products. Understand how machine learning works and when and how it can be applied to solve problems.”
- IBM — “Evaluate real-world case studies showcasing the successful integration of AI into existing product management systems.”
- Kellogg — “Identify high-value AI use cases across customer experience, operations, and support functions, tailored to industry needs.”
The website www.datascience-pm.com defines AI product management as the “practice of managing business, technology, and data to develop, launch, and operate AI products”. Seems like a circular definition. This website is typical of the dribble found across many other websites. Isn’t an AI product manager or AI product management just like any other product manager / management? What makes it so special that it needs to have the “AI” prefix attached?
Now, the responsibility of a product manager is to own the product P&L, and oversee the product lifecycle, i.e. it’s development and build, it’s value proposition and market positioning, it’s continuous enhancements and refreshes. Given this description, the meaningful question to ask is: “Is AI a product?” If it isn’t, then an AI product manager is just vapourware. I submit to you that AI is mostly a feature, and not a product. You may then ask: “What’s the difference between a product and a feature?” (Note that this is a historical debate that isn’t fully resolved.)
Product vs Feature
Here are some useful definitions to distinguish a product from a feature:
- A product (can be physical or digital) can be sold on its own, but a feature cannot be sold separately from the underlying product. In a sense, a feature is an attribute of a product; it provides some kind of functionality to the product. The technology that powers the feature can, of course, be licensed, but it cannot exist on its own outside of a similar product.
- A product is built around a situational pain-point, while a feature is built around a task-oriented pain-point. File management is a situational pain-point while file sharing is a task pain-point; file management consists of storage, access, synchronisation, security, sharing, etc. Mobile banking is a situational pain-point while fund remittance is a task pain-point; mobile banking consists of balance inquiry, investment management, bill payments, remittances, etc.
These are just guidelines, and you can always find exceptions. Nonetheless, they do help us understand whether AI is a product or a feature.
AI is a Feature … Mostly
This nonsense needs to stop! Despite what you may have read …
- An AI-powered recommendation engine is NOT a product.
- The self-driving module installed into an EV car is NOT a product.
- The intelligent spam filter on your email is NOT a product.
- A bank’s anti-money laundering monitoring algorithm is NOT a product.
In short, an AI-enabled (i.e. machine learning) solution is NOT a product. In contrast, ChatGPT is a product. But Apple Intelligence is a feature. ChatGPT was created explicitly to be licensed and subscribed to; to be used via a fixed set of interfaces, either as a standalone or integrated into a workflow. Apple Intelligence, on the other hand, is an enhancement to Siri and other Apple applications to add new functionality; with or without Apple Intelligence, those applications can still be operated.
Now, let’s separate the Big Tech companies from everybody else. The job of Big Tech companies is to develop and sell technology, AI being one of them. But most organisations are not building AI products. Instead, they are leveraging AI to enhance their existing set of products and solutions. They are using AI intentionally as a feature. For these organisations, the concept of an AI product manager makes no sense. There is no P&L built exclusively around AI, but there is measurement on how AI adds to the P&L of existing products.
In most organisations, existing product managers work with their data science and data engineering teams to explore opportunities to improve their products with AI. The product manager behaves like a regular product manager would when embracing new technology — no different from when they were looking to internet-enable their products, or to digitalise their products. Of course, they have to read up on what the new technology can solve for, but they don’t have to be a different kind of product manager. The product manager doesn’t and cannot own the AI components embedded into their products.
Conclusion
Creating a course or learning programme for AI product manager makes no sense. These roles simply don’t and won’t exist in most organisations. Yes, product managers should learn about AI, but there is no new practice of product management that emerges from it. And even if you work for a Big Tech company, then AI is just another variation of a technology product manager.
