Defining the AI Product Manager

“A robot and a human having fun in one line style” by DALL-E 2

In this blog post, Michael Zhao, a Product Manager (PM) in SEEK’s Artificial Intelligence & Platform Services (AIPS), talks about how he’s defined what an AI PM really is, what they do and how it’s really not like the movie I, Robot, after all.

What I thought an AI Product Manager was

“How was I going to recreate the 3 laws of Robotics from I, Robot?” I wondered as I walked through the doors on the first day of my job as a Product Manager in the AIPS team at SEEK.

Here are the laws in case you’re not familiar with them:

1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.
2. A robot must obey orders given to it by human beings, except where such orders would conflict with the First Law.
3. A robot must protect its own existence, as long as such protection does not conflict with the First or Second Laws.

Fortunately, for the world and for myself, I didn’t have to recreate these laws. But in these last six months, I have come face to face with difficult philosophical questions, intense ethical debates and deeply technical models. I am beginning to see why Will Smith was running around so much.

As I entered into this new profession I promised myself that I would attempt to document my learnings to achieve three things:

• To get smarter: Like any good student, I have always enjoyed the process of writing, consolidating and recalling interesting content. However, unlike my trusted answers section at the back of the book, outside of the education system, I have found it difficult to find ‘correct answers’. Instead, I have found that bringing up thoughts and hypotheses and seeing what my peers have to say allows me to quickly improve my understanding.
• To reflect: During my foray into product, it appears (at least at first glance) to be a highly dynamic profession. Each product manager seems to have their own set of strategies and work processes. Some prefer optimising strategic positions, while others prefer driving progress through stakeholder engagement. This difference is what makes the profession so interesting. Product management embodies a set of skills that help people execute tactical responses to daily problems. Through a process of documenting and reflecting on these actions, I hope to discover a set of principles that will help make me a better product manager. Or even better, a set of strategies and processes to help other Product Managers improve their craft.
• To spark conversations: What better way to make friends with other PMs than to ask them to follow my Medium account? Just kidding. These blogs will help document my journey and learnings and allow me to share these thoughts with other AI Product Managers both at SEEK and abroad.

So what is an AI Product Manager?

From my time observing AI Product Managers and from trying to explain my job to my parents — who still find it hard to imagine someone only a few years out of university managing products with the same technology as Skynet — I have noticed 4 ways that people commonly try to define AI product managers:

An AI Product Manager looks after AI Products. Duh.

One of the simplest ways to describe the role of an AI product manager is to first describe the product and then mention that an AI product manager ‘manages’ it. For instance, one conversation I remember went something like this —

AI product managers are responsible for creating and improving digital products that use AI technology. Some popular AI products that you are probably already familiar with are Google’s Search or Netflix’s recommendation algorithm. Behind these products are generally teams of data scientists, engineers and product managers. At SEEK, there are also product managers that work on AI products. For instance, my team works on models that predict the probability that a candidate successfully gets the job.

An AI Product Manager defines product strategy, aligns partners and executes with the team

But simply managing a product doesn’t mean much. Instead, some AI product managers might try to explain what they do day to day. In my case, the average day might include:

1. Product Strategy: Working with the leadership team, partners, and the team to define, frame and communicate the highest-value problems the squad will work on.
2. Partner Collaboration: AI teams work closely with front-end product teams. For instance, teams looking after recommendation algorithms work with the home page team to decide the best way to deliver job recommendations. To maximise the impact of AI services, AI product managers will work to understand partner business objectives, identify opportunities for collaboration and coordinate team engagement.
3. Service Ownership: Defining measures of success to track service performance and working with the experimentation team to run experiments and optimise services.
4. Team facilitation: Maximising the team output by coordinating team processes, identifying and managing dependencies from partner teams and coordinating updates to various stakeholders

An AI Product Manager ‘fills in the gaps’

Another definition of AI Product Managers is that they ‘fill in the gaps’ within the team. They do anything and everything necessary to ensure the success of a product. For instance, while engineers build pipelines and data scientists train models, the product managers might liaise with partner teams, organise team events, and resolve bureaucratic red tape. They are tasked to do everything outside the technical remit of the team.

An AI Product Manager solves for the ‘why’

However, all these definitions are incomplete. They do not fully capture the value that an AI Product Manager provides. An AI Product Manager is ultimately the person responsible for defining the problem. Despite having the ability to allocate responsibilities, problem solve and take meetings with partner teams, an AI Product Manager is ultimately tasked with and assessed by their ability to effectively identify and communicate the problem to be solved. And it is through this function where they provide the highest value. All other tasks that an AI product manager chooses to do are accessories to this primary function. If a product manager was able to effectively determine and communicate the problem to be solved, but refused to take a single meeting more or write up another Jira ticket, I think the team would still be effective.

So what next?

As varied and fluid as the definition of AI product management is, I am excited by the prospect of learning what it takes to be an effective AI product manager. These 6 months have passed incredibly quickly and the amount of information that has been crammed into my brain would make Shamwow jealous.

Despite the immense amount of information I have learnt, I am excited to explore these following topics:

• How does an AI Product Manager differ from a traditional Product Manager?
• What are AI Product Management skills and how to quantify progress?
• How to optimise the average week of an AI Product Manager?
• How do you measure the success of AI Product Managers?

If these sound at all interesting, I look forward to you joining me on this journey. I hope the next 6 months are just as exciting and full of discovery!