Juan Ginzo

Implementing AI products and Machine Learning in your company will bring a lot of changes to your company’s processes and culture. And as we all know: change is hard.

When faced with the decision of implementing new technologies most executives will think “how will this change/affect/impact my business” -mostly probably thinking about ROI-. But when talking about implementing AI and ML models to a business we are not talking just about “improvement” but “evolution”. The difference may seem just a matter of semantics, but it’s rather a perspective on how to invest in your company’s future.

AI products and ML models are the next evolutionary step that follows the analytics trend, they provide the ability to unlock operational efficiencies and new customer value. But like any new area of development, it’s hard, risky, and hard to navigate. No surprises then, to see companies getting burn by trying to successfully implement ML products.

So, who can help you, the business stakeholder, navigate the treacherous waters of deploying A.I. and Machine Learning in your organization, and ultimately unleash its competitive potential? Enter the AI Product Manager (A.I. PM), the navigator that will help you moving towards achieving that evolution in your company.  

A.I. Product management is the emergent discipline that will help you along the journey into the uncharted territory of deploying data products. Not just because AI is a “new” field but also because each company will face its own specific challenges regarding how AI products will develop and work into their workflow and processes, there’s increasing need to this emergent set of skills, that sit across functions. There will be no cookie-cutter answers to any given problem. But as Jeff Bezos put it: “If you only do things where you know the answer in advance, your company goes away.”. It doesn’t mean you need to do alone!

So, why not partner up your data scientists with just a knowledgeable subject matter expert or classic PM? The first thing to recognize is that A.I. brings it’s own challenges when trying to plan and estimate, since they involve a shift from deterministic to probabilistic processes. You need someone that brings not only the usual product manager skills but also a good understanding of the capabilities and limitations of A.I. and Machine Learning. Two minds minds might work best than one, but you still need someone to act as a translator, able to straddle between both worlds.

Importantly the A.I. PM needs to bring not only substantive knowledge of the organization and industry but also of the way A.I. products get develop and the phases of a typical A.I. project goes through, i.e.: ideation -> feature development & data management -> experimentation -> modelling -> deployment/serving.

To take one of the stages as an example, feature development and data management, the A.I. PM needs to understand how the data being collected at the company can support predictive models. Is the data fit for purpose? How costly is to gather labelled data? How many silos need to be stitched together? How many teams need to be brought in to collaborate? Even if the A.I. PM doesn’t have depth experience in the particular company data landscape, it needs to be aware of the typical issues that will raise in the process of developing a data product, the right questions to ask and where the sensible tradeoffs lie.

So, its clear there’s a need for this emergent talent. Then next natural questions is, where to get it from. As with Data Science talent a decade ago, there’s still no well trodden path to become an A.I. PM, but still two natural funnels emerge. Most current A.I. PM in the industry found themselves thrown into the deep end of the discipline, either by interest or business need, from their day to day job as a “classical” PM. The other, more intriguing but less prevalent path is for Data Scientist, Data Engineers and ML Engineers to make the leap into a more business-centric role. I foresee that in the next couple of years we will see this more people talking about A.I. product management as its own discipline, in tandem with companies accruing more experience in developing data products and their human capital devoted to managing that development increasing.

In the meantime, we are still in the heroic age of putting Machine Learning into production, trying to figure out what works best and how data products differ from previous efforts in software development. So experimentation is key, not only with models but with ways of working. And the need for a steady set of hands that can bridge the gap between business realities and Data Science is about to explode. Exciting times to be a A.I. Product Manager.