You are using an outdated browser. For a faster, safer browsing experience, upgrade for free today.
Artificial Intelligence: Bringing Value to Product Management
  • by Giorgiana Dragos

Get up to speed on artificial intelligence in product management and discover the skillset product managers need to succeed.

Artificial Intelligence has become a high-stakes business priority. Companies have spent $360 billion on AI applications in the past three years alone. And it pays off — businesses that scale AI are expected to achieve a triple ROI.

But what does this mean for product managers?

It’s true, not all product companies are currently reaping the benefits of AI. But it’s only a matter of time before AI becomes indispensable to the product management landscape and drives new changes to the product development lifecycle. At this rate, what does the future hold?

Well, AI is only as good as the insights it provides. And this extends to the product management function as well. That’s why product managers working with AI-driven products have to find the right data and use it in order to respond to customer expectations and needs. Patrick Tsao, the  Head of Product at Getsafe, defines AI as “automated decision-making”.

AI and Product Management

If AI is summarized as automated decision-making, how do product management decisions look like in the age of AI? 

To tackle this question, let’s start from the product itself. All products are built by following the same process — understanding the problem, narrowing in on the solution, drawing the plan, building the product, and making adjustments until the product delivers on its problem-solving promise. 

Yet as much as products have to solve problems, the underlying role of the product manager revolves around discovering them. Indeed, many product managers highlight the problem-first mindset as a stepping stone towards designing successful products. Why? Because it speaks to the deeper problem and is likely to generate long-term impact.

I firmly believe that if you’re focused on solving real customer pain points then you’re most likely moving your product in the right direction. [...] Customers can recognize that they have a problem but it’s not always clear how to solve that problem, and what other workflows or knock-on problems might arise as a result of solving their initial problem. Because of this, there just isn’t any real substitute for staying close to customers and listening carefully to what they have to say.

Jeff Gardner on his Product Mindset

Adding AI to the problem-first mindset allows product managers to delve deeper into the problem side and make informed decisions that will add value to the product’s lifecycle. As soon as they have discovered the right problem, product managers can invest in the product itself —  build prototypes, design, write solutions specs, and so on. 

But to deliver on the promise of the problem-first mindset, product managers need to work with data scientists and data engineers in order to map the problem effectively and empower product teams to deliver great output for the right outcomes

The AI-Driven Product Manager

To thrive in the age of AI, product managers need to own the automated decision-making game. To this end, here is the skillset they need, as outlined by Mayukh Bhaowal, Director of Product Management at Salesforce Einstein.

  1. Problem-Mapping

This starts from knowing your problem. Reaching the solution might involve AI — or not. That’s why PMs need to sift through the traditional methods and understand whether AI would add value to the application or not. 

A common pitfall at this point involves relying on the technology alone and ignoring real user needs. AI could be the best way to solve a problem, but product managers must keep user needs top of mind in order to drive successful outcomes.

  1. Data as the New User Interface

Apply AI to your data and you’re on the path to innovation. So are the aspirations of every product manager out there. But data is no silver bullet — it could be clean or messy. And product managers need to be data literate to ask the right questions about the data they have at hand. Once they have a reliable data set, they can build on that and design products that cater to the customers’ needs.

  1. Acceptance Criteria

This involves working with data scientists in order to establish the metrics that need to be optimized and train the machine learning algorithm to work towards the desired outcomes. To this end, the ML system needs rigorous acceptance criteria for the outputs. And product managers need to perform regular quality checks for bugs, missing output precision, incomplete or inconsistent results, and so on.

  1. Explainability, Ethics, and Bias

Gearing up for success is easy. Gearing up for failure is a whole different story. But the reality is, a lot could go wrong when working with AI and most issues are actually difficult to spot. So gearing up for failure might be the best foot forward. For instance, contemplating how the ML model might fail and understanding how your own biases or biases occurring in your training data could distort outcomes.

When it comes to explainability, product managers need to be transparent about the AI algorithms linked to their products. This depends on the use case and the industry they are serving, and to the extent to which the AI algorithm is explainable or not. In highly-regulated industries, such as healthcare, explainability is more important than accuracy.

And ultimately, product managers have to take into account the ethical side of their AI-powered products. Is the data set diverse enough? Is it likely to enforce unethical biases, such as race or gender. These are all important questions that don’t rely on technical answers. That’s why product managers need to consider whether the data is fit for a given application. For example, gender is an important signal for medical diagnosis, but not in the recruitment workflow. 

  1. Scaling from Research to Production

How do you unleash the full power of AI in product management? Product managers need to be strategic with their AI pilots in order to scale the impact of AI to real world applications. In this process, problems will abound, but not all of them will make it into production. 

Nonetheless, some common specifications need to be tackled. Where will the data reside? In the cloud or across mobile devices? How soon will the AI emit its output? How often will the ML models retrain? Getting answers to these questions help product managers multiply the value of their AI efforts and scale successfully, time and again.

What’s Next?

As the now famous adage goes: “AI is here to stay.” Indeed, it is. That’s why it’s difficult to ponder the future of product management without adding AI into the mix.

Prow Conference
Social Updates

Find more info about Prow here:

Brought to you by
Bianca Arhot