How AI and ML will transform product management

Ashwin Payyandan’s experience ranges from systems hardware, cloud software to building human-computer interactions (HCI) at Amazon, using Generative AI and Large Language Models (LLMs) that help millions of customers have meaningful conversations with AI assistants.

Building on his expertise as a Senior Product Manager, his ELVTR course on Product Management for AI & ML will help learners harness the immense powers of artificial intelligence to build better products and features.

Ashwin spoke with ELVTR about the heightened significance of humans in the AI era, the critical role of data in developing effective AI/ML solutions, and the essential skills that will enhance the future employability of product managers.

Former Google Chief Evangelist Nicolas Darveau-Garneau told us that if you are not a world-class AI marketer, three years from now you will not have a job. Is AI equally important in product management?

AI is no longer a specialty or differentiation point for products, but rather a table-stakes requirement. AI reflects a new technology trend that companies will need to use to stay relevant in the market.

All professions, not just product management, will require AI as a necessary skill. Today we don't talk about doing business on the internet, because that's taken for granted, everybody is online. 

You don't have an internet customer manager or an internet product manager, because that's what everybody does. Similarly, AI/ML will inherently be part of everything that we do. AI fluency is already a core competency for modern product management.

What are the major problems with AI and how do you solve them?

According to the latest estimates, 328 million terabytes of data are created by businesses and individuals each day. 90% of the world's data was generated in the last two years alone. Most of this data is intrinsically unclean and not appropriate for AI systems to consume. 

If you train machine learning models with dirty data, the resulting predictions will be unexpected because the machine is only learning from what it's given: garbage in, garbage out. 

So, for AI systems to make meaningful predictions, data needs to be cleaned, transformed into the right format with the right ingredients so that the machine is learning the right things and making the right decisions. And that responsibility squarely lies with humans who are designing these systems.

The key to AI/ML success lies in the availability and quality of data, as well as the way it is transformed. My course focuses heavily on the importance of data to help you understand how you locate and source data for your AI projects, what kinds of data are available and what kinds of issues within data you have to overcome to be successful with your AI projects.

What practical skills will students learn from your course that they can use on their first day as an AI/ML product manager?

One of the primary skills of a product manager is to recognise problems that can be addressed with AI and being able to articulate them. This requires understanding frustrations potential customers have. 

It's crucial to understand the importance of AI/ML beyond simply applying existing AI solutions to problems. You think of the problem first and then the solution, not the other way around. The course provides tools and frameworks to help you organise your customer backwards thinking.

Students will gain proficiency in implementing AI/ML solutions by learning to source the right data for the business/customer problem, gaining understanding of AI/ML models, experimenting, testing and evaluating their performance. We do this through a practical example, where students experience the actual process of identifying a problem and building an AI solution for it.

During the course, we help them recognise transferable skills that can be carried over from previous roles or their current profession, and identify new ones that they need to be successful AI product managers.

How does AI/ML experimentation lead to a better product? 

First, you need to recognise that you cannot build AI solutions without experimentation. It's such a complex system that you can never get your solution the first time. The first iteration of the model you are building will not perform at the level you expect it to. 

That's ‌the nature of this technology. It has to be configured, adjusted, like tuning the radios in the old days or a musical instrument. Getting it to the right setup immediately was always a struggle.

AI/ML is identical. You need to build and tune it, perhaps many times over until you reach the right setup. So you build a first version, you test it out, you look at your predictions and then you go back and adjust your parameters. You may even have to bring more data and retrain it to improve. Experimentation never stops, because data is changing as well.

The human element of designing solutions is also important. Your approach has to be: “Is my solution safe for humans to use?” If not, how do I make it safe? Is it going to harm human beings or encroach on their privacy? Are they okay with giving us their private information?

The capstone project that students have to do at the end of the course forces you to do that. If you build an AI system that customers don’t understand or know how to interact with, that’s a failed system.

When you recruit product managers at Amazon, do candidates need to have a background in AI?

It helps if you have some background in AI and machine learning. But the key is understanding how to apply AI solutions, the foundations of AI technology, which is what our course is designed to do. 

You don't have to be super technical about these models. You don’t need to know exactly how they work and what mathematical algorithms are applied internally. I don’t know them myself. 

But you have to learn to identify the problems and what kind of model you would apply to those problems, whether that's supervised learning, deep learning, neural network, LLMs. As a product manager, it’s your job to map the business problem to any AI or machine learning solution.

What skill will help you stand out of the crowd, since many people are taking courses on AI/ML?

This might come as a surprise, but it’s customer-oriented thinking and being able to empathise with human problems. All AI/ML solutions need customer-backwards thinking. If you can build products by articulating exactly what the customer needs or what the problem is and then apply the appropriate AI/ML solution to do it, that will always stand out. 

It's being able to have your user or your customer front and centre, that ability to be passionate and obsessive about your customer. Being able to empathise, because machines don't have empathy. They can’t feel pain or disruption in their life, and that empathy has to come from product managers and people who build these products. 

Product inventors need to apply high empathy, taking extreme care of your system’s product design and the product’s data, so that it translates into machine learning solutions that work for humans. Customer obsession, what I would call customer centrism, is what makes you stand out.

What emerging trends in AI/ML will have the most significant impact on product management?

GenAI will disrupt most professions, not just product management. We are seeing a very fast pace of development with AI. Emerging technologies such as AI have the potential to enhance every human capability. 

For example, image recognition is replicating human eyesight. NLP (Natural Language Processing) models are able to recognise or create human voice. Large language models can summarise large bits of information and make decisions for you. 

All these together will help humans create highly intelligent systems. It's already happening, AI systems are doing a lot of the human work. A lot of the ideation, problem discovery and thinking done by product managers and inventors could be supported by Generative AI in the future. 

Product managers are likely to turn into moderators for ideas based on data we already have and pivot towards more disruptive innovation and generating ideas that haven’t previously been explored.