Why not all Product Managers are fit for Data and AI products?

There is a remarkable increase in the number of Product Managers in the Data and AI field as the convergence of data-driven technologies reshapes industries.

More and more companies are aware that data science teams should not work in isolation. These teams should work as a product team, and make strategic decisions about what is worth building and why.

Oftentimes, we see product managers transitioning directly into data and AI-focused roles, from other domains, which might seem a smooth change for them, but might also cause a challenging start for everyone else involved in the product team. Let’s not make data scientists and/or ml engineers' lives harder than it is, and have a look at why not everyone is suitable for this transition.

Product Manager Role

There are many PMs specialized in different domains. A PM of an e-commerce website has different skills and obstacles to tackle than a PM of a hardware product, or a PM of a data science product.

A product manager is a multidisciplinary individual who identifies what product to build when to build it, and why it needs to be built. They find answers to questions like “Which one is a great idea, and which one is not?” Which ones are feasible given our technical capabilities in the company and which ones are not?

They are at the intersection of 3 key areas:

The technical components of a product. The UX components of a product (user experience), and the business components of the product.

Being a product manager is like “Jack of all trades, master of none.”
They must have a broad knowledge of technical user experience and business domains but they’re not necessarily the experts.

Product Manager in Data and AI

In Data & AI, there is an additional domain: Data.

A product manager in AI and data has the additional responsibility of considering the collection, security, variety, and accuracy of the data being used in the algorithms or models.

As a technical component, they are also expected to understand how AI works at a high level, the foundational practices in data science, and ML algorithms (supervised vs unsupervised learning, etc.).

Product managers learn how to compromise and collaborate between these four components of a product in order to build the best overall product possible, that optimizes for tech, UX, business, and data all at once.

For example, a data scientist can come up with the best ML model for making a user happy, which may not be the best user experience because it could compromise a user’s interest in privacy. It also may not be the best business decision because the data it is trained on might be too costly to collect.

Product managers manage the product, not the people.

Product managers do not manage people, they collaborate as a team members and pull ideas from many parties. It’s easier to share your big bold ideas with a coworker than it is a boss, right? That’s the beauty of managing a product.

2 additional challenges AI and data PMs have to face

• Uncertainty of time and performance of projects. It’s not certain how much time and effort is needed. And it’s never guaranteed that the performance of the product will achieve the goal.
• Communicating AI to your stakeholders. Data science and the technicality of the operation side of it are still very little understood by executive and leadership teams without technical knowledge. PMs are expected to work on educating some people involved in the product on how AI really works.

Example Failure

Imagine a scenario where a PM is assigned to lead a data science product, a recommendation system for an e-commerce platform. Because he/she is already managing some products on e-commerce.

The PM lacks a deep understanding of the ML product development process, despite having good insights into the business strategy and user experience. Possible failures that can occur:

Unrealistic Timeliness: Product managers may inadvertently set inaccurate timelines by asking wrong or inefficient questions. For instance, they might consult data scientists regarding model development but overlook the essential step of model deployment simply because they lack awareness of the specific components involved in the process. This knowledge gap could lead to unrealistic or incomplete timelines for the overall project.

Inappropriate Metrics: The product manager might struggle to define appropriate metrics for measuring the success of the data science product. They may focus on traditional product metrics, overlooking crucial data science-specific performance indicators like accuracy, precision, recall, and F1-score.

Misalignment with Data Team: Data is one of 4 important components. A PM without a technical background in AI, may not realize how big the role of data requirements is in product development. They may not check in with Data teams with the right requirements.

Difficulty in Stakeholder Communication: Explaining the progress and challenges of the data science project to stakeholders may prove challenging for the product manager. Without a deep understanding of the underlying technology, they may struggle to address technical questions or provide meaningful updates to stakeholders.

Remember, it’s not that these product managers are incompetent; they’re just better suited for a different dance floor where data and algorithms don’t lead the rhythm! They can always educate themselves and close the knowledge gap to become a better fit for data and AI.

A nice course I have found on product management in data and AI, which is the inspiration for this article:

https://www.udemy.com/course/the-product-management-for-data-science-ai-course/