Moravec's Paradox says Product Management is Not Dead

Product folks love to debate about existential topics about their own roles! I had largely ignored everything about whether Product Management is or is not dead, until I learnt about Moravec's Paradox. If you are one of those who believe Product Management is dead, I hope this will give you something to consider.

Hans Peter Moravec was a computer scientist at the Robotics Institute of Carnegie Mellon. In 1988 he wrote that "it is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility".

Psychologist Steven Pinker wrote in 1994 that when it comes to AI research, easy problems are hard and hard problems are easy. What does that mean? What is easy for humans, or what humans do unconsciously, seemingly effortlessly, is hard for computers. What is hard for humans, or what takes lot of effort for humans, can be reverse-engineered and done easily by computers.

Moravec's Paradox can be represented by this graph below.

Now let's apply this to the role of a Product Manager and the discipline of product management.

Applying Moravec's Paradox to Product Management

(1) Easy for PMs & Easy for AI

Tasks that are structured and repetitive, or well-documented, are easy for PMs to perform and also easy for AI. When AI does these tasks, it creates a massive time advantage for PMs, freeing them up to do more valuable work. Of course, it goes without saying, that for AI to be effective at these tasks is dependent on the data and context that already exists.

Some of these PM tasks are:

• Tracking sprint progress
• Summarizing meeting notes
• Setting up and running A/B tests
• Writing basic user stories
• Generating initial drafts of PRDs
• Market research and competitive intel

You might say this is busy work that PMs have to spend a lot of time doing. AI as a handy companion acts as a force multiplier. The PM still owns and is accountable for the PRD and user stories, though! :-)

AI creates a massive time advantage for PMs.

(2) Hard for PMs & Easy for AI

Humans are constrained by cognitive and physical limitations. We get tired. We get overwhelmed. Our body and brain can only be stretched so much. AI, on the other hand, never sleeps, is never hungry, and never gets tired!

So tasks that involve large-scale data crunching, and structured problem-solving, are things that AI can do in it's sleep, or actually, when we sleep! For example:

• Customer sentiment analysis at scale - AI can analyze lots of reviews, support tickets, social media comments, etc., and come up with patterns for us to look at.
• Pattern recognition in user behavior - AI can find correlations in user behavior and how they engage with the product, that might not be obvious for us.
• Optimizing pricing models - AI can simulate different pricing strategies and, given market and customer data, figure out how to optimize for revenue.
• Churn prediction - AI can flag at-risk customers before they leave by analyzing usage and engagement metrics.

When AI does these types of tasks, it almost gives PMs superpowers by not being restricted by physical and cognitive burden and enabling better decision-making.

AI bestows superpowers to PMs.

(3) Hard for PMs & Hard for AI

PMs have to face a lot of uncertainty and ambiguity. We have to operate with incomplete data. We need to abstract and filter the signal from the noise. We have to be good at reasoning. AI is also getting good at reasoning. It is useful to distinguish between the types of reasoning AI is good and at what PMs, or humans in general, are good at.

Deductive Reasoning:

This is about making accurate conclusions based on rules. For example: All birds have feathers. Penguins are birds. Therefore, penguins have feathers (they do, even if we can't see them). AI is very good at this and can outperform humans because it can be trained on massive number of rules, pretty much, all rules about the whole world. Humans don't have the cognitive bandwidth to learn and remember all rules.

Inductive Reasoning:

This is about making probable (but perhaps not accurate) conclusions based on patterns. For example: I only see white swans. All my friends have only seen white swans. Therefore all swans are white (not entirely accurate). AI is very good at this as well and can outperform humans because it can look across large swaths of data and find patterns. Humans, again, don't have the cognitive bandwidth to look across large datasets and find the patterns.

Abductive Reasoning:

This is about making observations, understanding context, looking at exceptions and anomalies and coming up with most likely hypotheses that can be tested. Take the example of Sherlock Holmes. An expensive horse was stolen from the stable at night. But the stable dog did not bark. Therefore the thief must be someone that is known to the dog. In other words, this must be an inside job. Humans excel at this over AI. Crime scene investigators start from the dead body and ask what all must have had to be true for this person to end up here like this. Humans can work backwards from the environment and context and come up with most likely explanations.

So what are the types of things PMs do that are hard and also hard for AI? This is actually the hardest quadrant because there are no clear right or wrong answers. Data alone is insufficient. Consider things like:

• Defining a truly innovative product vision: PMs struggle because vision requires intuition, creativity, ingenuity, and understanding of human behavior. These are things that cannot be derived from data. AI also struggles here. We got the automobile and the iPhone because Henry Ford and Steve Jobs realized the answer was not faster horses and better Blackberrys.
  
  
• Predicting market disruption and black swan events: PMs struggle because unexpected events (COVID-19, regulatory changes) are hard to foresee. Human intuition comes in the way because of biases and overconfidence. AI struggles because it cannot predict truly novel and exception scenarios. The success that Zoom and other collaboration tools had during the shift to remote work could not have been predicted, but only capitalized upon.
  
  
• Achieving Product-Market Fit in nascent markets: Humans struggle because there are no clear precedents or patterns. PMs have to dig deep into customer feedback and iterate constantly. AI struggles because it cannot predict something that never existed. Take the example of AirBnB. AI would have predicted failure based on the existing model of hospitality. Even humans struggled with the concept of a business model around someone staying in a stranger's home.
  
  
• Determining when to pivot vs when to persevere: PMs struggle because of uncertainty, and biases such as sunk-cost bias. It also becomes a gut-wrenching decision. AI also likely struggles because past data cannot predict the future. Slack pivoted from a gaming company to an enterprise collaboration company. This pivot was intuitive based on knowledge of how users were behaving.

Success here often depends on intuition, judgement, experience and tolerance for uncertainty. These are things that AI is not good at.

(4) Easy for PMs & Hard for AI

Success in a PM role comes down to stakeholder relationships, influencing, understanding team dynamics, negotiating priorities and trade-offs, building compelling narratives, and adapting communication to different audiences. These are skills that humans have built through the course of human evolution. They can be learnt. They come easily for some, and some work on building them. They require emotional and social intelligence which is hard for AI.

These are the things that make us deeply human and great PMs excel at:

• Influencing stakeholders: Driving alignment between executives, XFN teams, on key decisions, strategy and roadmap is deeply human.
  
  
• Interpreting ambiguous customer feedback: PMs have to be good at observing frustration, hesitation, sarcasm, indirect inputs, and also read body language.
  
  
• Understanding team dynamics: Recognizing interpersonal conflicts or when a team is overworked is important. Understanding politics and power is also deeply human.
  
  
• Developing a compelling product narrative: Positioning and messaging needs to move people. When Canva says it is a web-based design tool for non-designers, it makes all non-designers feel like they can create anything.
  
  
• Resolving conflict:A PM needs to have the logical left-brain to connect with engineers and finance, the creative and empathetic right-brain to connect with customers, designers and marketing/sales, and the speech of a diplomat to balance conflicting priorities and personalities.

To Summarize

This reframing that comes from understanding Moravec's Paradox allows us to understand that AI isn't going to replace PMs but it can certainly amplify PMs.

Product Management is NOT DEAD, long live PMs!