Most large enterprises begin their AI transformation by hiring a few data scientists to build prototype models. The AI leaders quickly realize they need to make many other foundational changes to realize the company’s AI vision. The big, obvious changes might be building an MLOps infrastructure and creating a new ML governance model. On top of those changes, they also need to handle the million little details and challenges that surface every day.
AI leaders soon realize they need a quarterback for their AI initiatives. They need an AI product manager to lead the organization through all of the necessary changes. In this blog post, I’ll offer some suggestions to help you build a product-management capability for your AI transformation.
The ideal AI product manager
Your ideal candidate has the following skills and experience:
- A foundational understanding of data science practices such as methodology selection, experimentation, training data acquisition, feature engineering, performance evaluation, and tool selection. This knowledge is critical for communicating with data scientists and putting their results in context for customers.
- Infrastructure experience, ideally experience with deploying and scaling models in modern platforms such as Kubernetes and service meshes.
- The initiative and determination to overcome obstacles such as resistance to change and the ambiguous legal issues that relate to the data.
- Experience with modern product management best practices such as lean startup and customer development. They’ll use these strategies to identify the biggest project risks and systematically overcome them.
- An ability to simplify AI for customers and address their challenges.
- Personal persuasion skills to lead customers through the process of behavior change necessary to take advantage of AI.
In short, an AI product manager needs to understand how to build models, deploy them, drive organizational change, and remove obstacles so everyone can be successful.
Unicorns don’t exist
Unfortunately, the ideal AI product manager doesn’t exist. The few people who have the necessary experience don’t want to be an AI product manager. They start their own companies (like I did, for example), or they would rather lead a company’s AI transformation (like you did, for example).
The scarcity of AI project managers will change as the industry matures and more people have experience with the full data science lifecycle. But demand will continue to outstrip supply for years to come, and the top talent will continue to migrate to the best opportunities at the employers of choice.
Focus on capabilities, not people
Most large companies will need to focus on creating an AI product management capability that doesn’t depend on elite talent. In other words, they must build a team that has the ideal AI product management skills rather than searching for the unicorn hire that can offer all of the skills single-handedly.
As you build your AI product-management team, consider these suggestions:
- Get your structure right.
- Start with your top-tier talent.
- Assess gaps in skills.
- Close those gaps by adding other team members.
Get your structure right
Even unicorns won’t be effective if the structure of your organizational transformation doesn’t set them up for success. So start by aligning the right organizational structure with a clear mandate. I detail this process in my two-part series about how to transform your organization for AI.
Start with the top 20% of your existing talent
I’m always surprised by leaders who expect great results from average performers. No magical framework, governance model, or process is going to work if this team isn’t staffed with elite talent.
Your top 20% of performers will be good at almost anything. So start by selecting your best project manager, product manager, data scientist, or engineer. Use them as the foundation of your AI product management team.
Of course creating this team isn’t going to be easy. The organization will be reluctant to surrender the people who are making everything work. You will need to convince decision makers that the advantages of dedicating top talent to this team outweigh the disadvantages.
Assess the skills gaps
After you identify the best available person for the AI product management job, assess the additional skills your team needs.
Example: The effective traditional product manager
Suppose you can choose from five product managers, and you decide Lilly is the best fit for your AI team. Lilly has been with the company for three years. She’s a master at navigating the organization through complex projects. She has worked primarily on infrastructure projects and she knows data. Wherever data issues arise, business customers call Lilly because she knows how to get things done. Best of all, Lilly is excited about AI and can’t wait to dive into a new challenge.
You and Lilly get together and mutually assess her current skills and experience. You rate her strengths as high (H), medium (M), and low (L).
Lilly is ideal for the product management job. But she knows little about data science or machine learning. Although her infrastructure experience is limited, she has enough experience to get the program started.
Build the team to close the gaps
When you see the gaps in AI product-management skills, the solution is usually obvious. In this case, Lilly needs to partner with a strong data scientist. Ideally, the partner should be able to simplify and communicate complex AI concepts to customers.
Lilly isn’t a unicorn. Many of our clients wouldn’t even offer her an interview. But if you approach project management from the perspective of capabilities, the solution is obvious.
Mistakes to avoid
When I help large enterprises build AI product management teams, I steer them around some common pitfalls that can cause problems later on. These common pitfalls often undermine efforts to build an effective AI project management capability:
- Overweighting data science skills
- Ignoring experience in data processing or statistics
- Team members who lack enthusiasm for the AI transformation
- Team leaders who have big egos
Overweighting data science skills
Data science fluency is necessary for maximum team efficiency. However, efforts that undervalue this attribute to the point of sacrificing soft skills and leadership experience almost always end in disaster. So consider both data science skills and people skills as you assemble your AI project management team.
Ignoring data processing or statistics experience
For the past 20 years, companies have been prioritizing front-end (UX/UI) experience when they hire product managers. Because most major initiatives were based on web or mobile applications, interface competency has been critical for success.
For AI initiatives, however, data and infrastructure challenges are usually a bigger risk than the front-end experience. So prioritize experience in data processing and statistics over the front-end experience when you’re considering your team’s skill sets.
Lack of enthusiasm for AI and change
You want to build a team of passionate AI enthusiasts who share your vision for transforming the company with technology. But your incumbent employees—even the top 20% of your performers—might not share your passion for big change. As you consider candidates for your team, make sure they’re motivated by the idea of AI transformation and that they’re committed to bringing the vision to life.
Product management within an enterprise can be incredibly hard when the organization resists the best practices. For example, everyone loves slogans like “fail fast" until the first failure surfaces and blame starts flying. People who feel burned by failure are often reluctant to take risks again.
To overcome friction in big projects, some organizations seek out egotistical product managers who will bulldoze their way to success, no matter the human cost. Such people are toxic within an AI team, and they won’t help you create the idea meritocracy you need.
So watch out for bravado, and select team members who balance confidence with humility.
When in doubt, go with talent
I hope this blog post helps organize your approach to building your company’s AI product management capability. This capability is hard to create, but a team approach will move you down the path of success.
When in doubt, bet on the most talented resources you can find. Your downside risk will be significantly lower.