Relax PMs, we're ready for AI

Mind-blowing ways to make new products? Bring it on

Apr 09, 2024

Last year was all about the GenAI hype. Now we’re getting to practical deployments for everyone - not just tech giants. Barriers to entry are lowering fast.

It’s an exciting time to be a Product Manager.

It’s also, lets face it, a bit daunting. There’s already so much to know in the craft of Product Management…now we need to be AI experts too?

Yes. We all need to learn the fundamentals because AI is here to stay, it’s going to get better and more usable. There are some great learning resources at the bottom of this post.

But the good news is that Product Management doesn't change just because we can solve our user’s problems in amazing new ways.

We’ve got this. Most of the tools are already in our toolkit.

Here’s my take on three approaches that will stand us in good stead for the AI era:

PMs love the Problem, not the Solution

It’s so easy to get carried away as a technologist when a whole new horizon opens up like this.

But PMs know not to fall into that trap. Our discipline works with Outcomes, Opportunity Trees, Jobs to Be Done and User Story Maps because they help us focus on the problems our users are trying to solve.

Lots of companies are taking existing processes and speeding them up with AI automation. And that’s… ok. But really it’s just a faster horse.

Product Management means understanding the root customer problem and applying AI solutions to solve it.

It’s applicable to problems you might not immediately think of - like the challenge of growing crops. Keeping the weeds away is a problem space as old as agriculture itself, and we’ve come a long way from the first solution - a hoe. John Deere’s See & Spray product uses AI to do something a farmer on their own could never achieve - monitoring entire fields to identify crops from weeds and apply the right treatment to each individual plant. It can reduce pesticide application by two thirds while improving farm productivity and profitability.

AI products are transforming industries, by finding new ways to meet user problems.

Product Risks frame our thinking

Marty Cagan established the 4 Risks in Product Management in his book Inspired. They hold true today - and we can use them as guardrails in AI product development too.

Here are some key questions to ask when you’re thinking about Product Risk in AI projects:

Value Risk

whether customers will buy it or users will choose to use it

What problems are we targeting, and how important are they to our users?
How will deploying AI in the solution affect the user's willingness to choose it? Might they have privacy, ethical or other concerns?
How important is accuracy in the solution? AI works by probability, so will never be 100% accurate on its own.

Usability Risk

whether users can figure out how to use it

What's the most important element of usability for this problem? It may be speed, accuracy and quality of results, or ease of learning how to use it.
Does a user have to know about how AI works to be able to use the solution?
How does the user input information to the product, and how do they do that in a way that’s quick and intuitive?

Feasibility Risk

whether we can build it with the time, skills and technology we have

This is the most important area to focus on as a PM when assessing a new AI solution, and the most complex. It’s highly unlikely a single LLM will be a suitable solution. Compound systems with complex architectures chain together various data sources and generative models. So be prepared to interrogate this risk hard to ensure value.

What data do we intend to use to train the model?
Do we have the appropriate data infrastructure and cloud architecture to sustain an AI solution?
How much will the AI solution cost, initially and in production?
How are we going to ensure accuracy of the responses? (we all want to avoid hallucinations…)
How competent are we to build this solution? Have we built anything like this before?
What technical skills and capabilities will we need in the product development team? You’re likely to need a different make up from a non-AI solution, with roles like data engineers, AI architects and ML engineers.
What’s our partner strategy? Which AI capabilities do we want to develop and grow in-house, and what are we comfortable to buy-in?

Business Viability Risk

whether it works for the various aspects of our business

What are the regulatory, ethical and legal considerations?
What is the data treatment? Consider your users, partners and suppliers as well as your own business when assessing data and privacy needs.
What are the security risks associated with the use of AI for this solution?
How would deploying AI in this product fit with our business strategy?
What’s our go to market approach and does AI impact that positively or negatively?

We build the skateboard before the sportscar

Last year, I worked on an AI travel concierge proof of concept. The product strategy aspect of the role was fun - it’s an amazing problem space. Who wouldn’t want to make it easier to find a holiday without hours of searching and filtering? But the build element of the product was really challenging. We hadn’t done this before. Our stakeholders were in a hurry and the pressure was on. What worked? We stripped the proof of concept right back to the bare minimum slice of vertical functionality - one end-to-end journey for one use case. Were our stakeholders happy? Not hugely. They had bought into the hype. But they were happy enough to let us keep going, keep learning, keep building. Now, we have the real deal.

AI solutions are a prime example of the agile principle that you build the earliest testable product and grow from there. They’re big, complex and new. Teams need to learn and iterate. PMs apply that principle, understand the fundamental problem we’re trying to solve, and get there in the leanest way possible to begin with. Our vision might be for a Ferrari, but first we build a skateboard.

What Product Managers bring to the AI party is the same crucial thing we always bring - a laser focus on value, for our users and our businesses.

As AI goes mainstream, that need for true product thinking will be central to success.

PMs, let’s do this.