top of page

Applying Bob Moesta's lessons to AI



A water color painting with bright colors that look like blobs

When a big trend like AI emerges, we tend to focus heavily on new skills and forget those that have always served us well. Every machine learning expert I've interviewed has stressed the importance of having clear objectives to direct their work. These objectives should be grounded in a solid understanding of what the customer is trying to achieve, then translated into a format that is understandable for the AI/ML team.


I recently watched Bob Moesta, a legend in product innovation, talk about developing your innovation-building muscle at a virtual Business of Software event hosted by the always delightful Mark Littlewood. Moesta is the founder and CEO of The Rewired Group and author of "Learning to Build: Skills of Innovators and Entrepreneurs" and "Demand-Side Sales: Stop Selling and Help Your Customers Make Progress." Moesta will appear at BoS USA 2024, September 23-25, in Raleigh, North Carolina.


The skills that Moesta taught apply to ML and AI projects in the same way they do to food, real estate, and pretty much any product or service you can think of.


Empathetic Perspective

"The irrational thing becomes rational with context," was one of my favorite quotes from Moesta.


Being able to set aside your ideas and truly listen to customers is crucial. We often think we know what customers want, but our internal biases can cloud our judgment. Empathy goes beyond just listening to understanding how the customer connects the dots, what draws them in, and what will put them off.


Often, purchasing decisions are made by more than one person, and understanding the perspectives and jobs of everyone involved helps us make product decisions with limited resources.


An example is lawn care: a customer may not be into gardening, but they don't want to be seen as a crummy neighbor. We can translate the job they want to do, how they want to feel about it, and how they want to be seen by their neighbors into a use case for machine learning. The input could be photos of the current state of their lawn, and the output would be recommendations on what products to use to achieve an acceptable lawn.



Uncovering Demand

One of Moesta's key points is that demand exists inside consumers; it's not something created by a product or company. Every time a customer makes a purchase, you can go back through their story and discover a struggling moment. The goal of first-round interviews with people who have recently purchased your product is to uncover this struggling moment.


When Moesta worked for SNHU, they discovered that students were coming to level up their careers because of a life event like a potential milestone or layoff. Understanding a key customer segment they weren't supporting drove them to invest in redesigning their admission process resulting 10,000 applications a month when they usually got 1,000 in a year.


AI or ML projects can deliver a big win, but they are expensive, time-intensive, and risky. Understanding customer demand helps us invest in the right use cases and gives both the AI team and the product team clear direction on how to serve customers.



Causal Structures

I've coached a lot of skiers over the years, so I'm good at analyzing movements and describing cause and effect (I've even passed exams on it). So I thought I would get this, but it's been a struggle. Apparently, I'm not alone; Moesta says most people get this wrong.


Customers want outcomes, not outputs. We need to design backward, starting with the customer, the outcome they want, and then the system. We also need to focus on what we can control and make it robust enough because there are forces beyond our control.


We need to place AI/ML projects within causal structures to ensure the output moves us towards the outcome the customer is looking for.



Prototyping to Learn

Prototyping has been popular for many years. I've had many moments of discovery when I thought I had it all figured out and most definitely did not. We often try to prototype our best ideas, but Moesta stresses the need to put out the bad ones too to create more contrast between them and to generate more understanding of how the customer is thinking about the differences.


It's possible to prototype an ML or AI project by using outputs generated by an LLM, working with a small dataset, or a reasonable guess from a subject matter expert. The goal is not to validate what we can do with the data but to understand what kind of guidance our customers are looking for. This could save us from spending months or years working on a recommendation that no one would ever use. When I've done this before, we also got valuable insight into how customers weighed results and what actions they want to take next.



Making Trade-Offs

Everything in life and business is a trade-off, and in Moesta's interviews, he's able to surface how a customer makes trade-offs during the purchase process. In second-round interviews, Moesta recommends offering a variety of options to the customer so you can discover how they make trade-offs.


Recently, I interviewed an expert in predictive analytics and machine learning who works on a B2B information product. He doesn't use AI in forecasting even though it could be more accurate because customers need to be able to use their knowledge to modify the inputs into the forecast and explain their recommendations to their clients. A black box generating results would not be adopted by the audience, at least not now.



Communicating What You've Learned Effectively

During Moesta's presentation and my own research with AI/ML teams, a common theme is how challenging it can be to communicate research and influence senior leaders in an organization. Moesta recommends getting leaders or stakeholders involved in the research process.


When I was working with a client on a mobile app, the department head spent four hours doing interviews with students and insisted every engineer sit in on prototype tests. This created a powerful shift in perspective within the organization.


Additionally, I see product managers, designers, and engineers struggle with this often; it's a major gap in their professional training. This is why I focus a lot on storytelling and do a lot of prep sessions with my teams.


Conclusion

Don't get distracted by shiny new things and devalue older skills. These skills that helped Moesta launch more than 3,500 products work on AI products too. They help us truly understand our customers, make smart decisions, and navigate tradeoffs.

Comments


bottom of page