Strengthening Product Sense through Generative AI

One of the highest-leverage skills that a product manager can cultivate is product sense: the ability to identify which product ideas are most likely to succeed, and the ability to verbalize the rationale for why they’ll likely succeed.

Product managers with strong product sense are able to unlock more value from their products, mitigate more risks, and reduce more costs. It’s no wonder why PM managers deeply value product sense in their PMs.

However, this poses a catch-22: product sense has traditionally been hard to cultivate without prior product management experience, yet most people aren’t given the chance to become PMs if they don’t yet have strong product sense.

Thankfully, with the rise of generative AI, we now have the ability to empower people to practice their product thinking and strengthen their product sense, even without prior PM experience.

Here’s why this wasn’t feasible in the past: without generative AI, product management exercises take a lot of time and effort.

Creating pre-mortems for our products, analyzing customer segments, crafting product one-pagers, writing mock marketing materials, etc. used to take hours and hours to complete a first draft, let alone a polished deliverable.

But now, generative AI can accelerate this iteration time and enables us as PMs to explore many more scenarios and to better hone our product sense without us needing to write “low value words.”

That said, a word of caution: generative AI is not foolproof. Every PM should seriously question the responses that come back from their prompts.

However, even with that caveat in mind, PMs should consider using generative AI to quickly break through blank page syndrome and rapidly iterate through different angles for their product proposals.

In other words, while we’ll still need to view the output of generative AI with a critical eye, we no longer need to struggle through writing every word ourselves. 

In this in-depth guide, I discuss how to leverage generative AI to strengthen product sense. First, I break down product sense into its eight fundamental components. I then provide a brief overview of what generative AI is, and I discuss its potential as well as its pitfalls. Afterwards, I walk through two product management exercises leveraging generative AI. 

What is product sense?

Many product leaders have broken down product sense in a variety of ways:

  • Jules Walter defines product sense as “the overlap between empathy and creativity”

  • Shreyas Doshi defines product sense as “cognitive empathy, domain knowledge, and creativity”

  • Jackie Bavaro defines product sense as “understanding business goals & customer goals and finding opportunities to better serve those goals”

  • Paul Adams defines product sense as “repeated in-depth observation and direct interaction with customers who use your product”

While these mental models are a helpful starting point, they can still be somewhat vague for early-career PMs.

In our coaching practice, we’ve found that breaking down product sense into the following eight components has been the most helpful & actionable for our clients:

  1. Customer segments: Which customer segments exist? Which segment is the most valuable to focus on?

  2. Pains: What kinds of pains will customers likely have? Where are there unmet needs?

  3. Competitors and substitutes: What are customers already using? How well do these solutions solve their pains?

  4. Workflow: What new user behaviors am I proposing? How easy or difficult will these behaviors be for our customer segment to adopt?

  5. Frictions: What are the potential things that might confuse people or slow them down?

  6. Desirability: How will our targeted customer segment likely feel about our solution?

  7. Positioning: How should we market our product to our targeted customer segment?

  8. Return on investment: How much value will our solution yield, how much will it cost, and how long will it take to pay off?

By learning to answer each of the eight questions above for a variety of products, we find that product managers can swiftly deepen their product sense.

Keep in mind that each of these questions builds on the question before; you shouldn’t try to improve your “ROI product sense” if you don’t yet have a strong grasp of “customer segment product sense.”

Problematically, many product managers think that product sense is all about being a product visionary. That couldn’t be further from the truth. Product sense is all about using first principles and using your own experiences as an end user to inform your product proposals.

When you see an experienced PM rapidly identify the areas of improvement of an existing product or quickly assess the viability of a product idea, they’re running through the eight components above.

Because they’ve done it so many times, their product sense looks like instinct, causing people to conclude that product sense is an innate trait and not something that can be trained.

But, consider lawyers. Lawyers can quickly identify the key components of a contract because they’ve had so much repeated exposure with contracts; there’s no such thing as “innate contract sense” that people are born with.

Similarly, due to sheer exposure, experienced PMs can pull off product sense in a way that feels innate, but that shouldn’t stop you from learning product sense yourself. By approaching product sense methodically, you can quickly improve.

Now, let’s discuss what generative AI is, and why it can help accelerate product sense.

What is generative AI?

Generative AI (commonly shortened as “gen AI”) refers to artificial intelligence systems that are designed to learn and mimic patterns from large datasets and then generate new outputs based on those patterns.

These AIs are trained on billions of data points to understand the underlying patterns and structures present in the data. During the training phase, the model learns to capture the statistical relationships and dependencies in the training data, enabling it to generate realistic outputs.

In popular conversation, gen AI typically refers to one of three systems: OpenAI GPT-4, Amazon Bedrock, and Google Bard.

Generative AI isn’t magical, even though the results may appear to be magical. These systems simply predict which words will likely come next, based on its analysis on millions of pieces of literature.

In other words, generative AI models do not possess conscious understanding or intentions. They don't have an inherent understanding of the content they generate. The outputs are purely based on learned statistical patterns and probabilities from the training data.

That means that you need to be careful with the responses it provides! Sometimes, the response will sound very confident, but the details are entirely incorrect. For example, if you ask a generative AI model to add 4 numbers together (e.g. 253 + 219 + 842 + 13), it will likely give you the wrong answer (e.g. 1123 from the AI instead of 1327 as the actual answer).

Furthermore, we should never ask generative AI models to make decisions on our behalf, as these models have no innate reasoning capabilities. The goal of using generative AI is to quickly explore new ideas (i.e. divergent exploration) and not to decide how our products should be built (i.e. convergent decision-making).

Now that we understand what generative AI is, we can thoughtfully wield these powerful tools to accelerate our product sense. Specifically, let’s discuss the two key exercises that product managers use to build product sense, and let’s dive into how to use generative AI to rapidly iterate through these exercises.

The two key exercises to build product sense

Product managers hone their product sense in two main ways: analyzing other people’s products (a.k.a. product teardowns), and crafting their own product specs.

In both exercises, they walk through all eight product sense components above to identify the strengths and areas of improvement for each product idea.

While there are many other product management exercises like retrospectives, sprint planning, and setting OKRs, these don’t hone product sense quite as quickly as the two above because they don’t touch all eight components of product sense. 

Now that we know that product teardowns and product specs are the fastest way to build our product sense muscles, let’s discuss how to do each one below by leveraging generative AI.

Practicing product teardowns through generative AI

In the past, conducting a product teardown could easily take at least 1-3 weeks to complete. But, using generative AI, we can complete this exercise within 1-3 hours. I’ll show you how below.

Let’s jointly tackle a product teardown for Hinge, a popular dating app. The goal of our teardown is to identify where Hinge could further improve its product by shipping a new feature.

I’ll start this exercise by covering product context to make sure that we’re tackling the right product. Then, we’ll run through each of the eight components of product sense: customer segments, pains, competitors, workflow, frictions, desirability, positioning, and ROI. 

For each component, I’ll provide three things: a prompt to submit to our generative AI model, an example response that we might get back from a generative AI model, and the action items that you should take based on that response. Responses typically take the form of multiple paragraphs - so, if you’re interested in the response for each prompt, be sure to tap the arrow to expand the response.

As an aside, if you’d like us to run a custom product teardown with generative AI, just reach out for coaching and we’d be more than happy to help. 

Product context

Prompt: For someone who's never used Hinge (a dating app) before, explain what Hinge is to them. Limit your response to 200 words or less.

Action item: Check to see if this response actually aligns with your knowledge of the product. Sometimes, generative AI might accidentally confuse the product with a different product entirely, or it might make up features that don’t actually exist.

Be sure to compare its response vs. the actual user experience of the product. To help guide a more accurate response, you can add additional details about the product in your prompt.

Customer segments

Prompt: Which customer segments are most likely to use Hinge, and why?

Prompt: Which customer segments are least likely to use Hinge, and why?

Action item: See if you agree with each of the analyses. For each proposed customer segment, back them up using customer testimonials, user reviews, and marketing materials. 

Now that you have an overarching view of viable customer segments, your next step as a product manager is to prioritize customer segments.

Decide whether you’re going to dive deeper into a customer segment that is already using the product, or whether you’re going to go after a new customer segment that isn’t currently using the product. Remember, AI cannot make this decision for you!

For the sake of this exercise, I’m going to focus on an existing customer segment: Relationship Seekers.

Pains

Prompt: Let's focus on "relationship seeker" users. What methods or processes do these people typically use to find relationships, and why?

Prompt: Let's focus on "relationship seeker" users. What kinds of pain points, problems, or frictions might they run into when they use their current methods, and why?

Prompt: Let's continue to focus on "serious relationship seeker" users. What kinds of pain points, problems, or frictions might they run into when they use dating apps specifically, and why?

Action item: Run through the product and check to see if these pain points are real or not. As you run through the product, identify whether there are other pain points that the response may have missed. 

Now we have a targeted customer segment, and we have a set of pain points that these customers experience. Let’s now observe which kinds of competing solutions they currently use today, and where there might be gaps that we can execute against.

Competitors

Prompt: Let's continue to focus on "serious relationship seeker" users. What are the most common dating apps that they use today, and why?

Prompt: Say that we wanted to get one of these individuals to switch away from their current dating app, and get them to use our dating app. How would we convince them to switch? Why do these things matter to them?

Action item: Double check to see whether the identified competitors are correct. Identify any other missed competitors by using Google Search for “[product] competitors” and “[product] alternatives.”

Now that we understand what our targeted customers are using today, we’re going to take the next step and propose a new workflow within our product to solve a specific customer pain.

For this exercise, let’s say I want to solve the pain of “limited context” to decide whether to pursue a romantic relationship. To solve this pain, I propose a new feature: let people be able to share videos of themselves on their profile, rather than just words and pictures.

Workflow

Prompt: As a product manager, I want to address the problem of “limited context” on potential romantic partners. I think giving people the ability to see videos of each other is a good way to address this context problem. How much of this “limited context” pain can I solve by providing a way for people to connect through video, and not just text or pictures?

Prompt: As a product manager, I’m considering launching a new feature within my dating app for serious relationship seekers. I would like to give users the ability to add short videos of themselves to their profiles, not just text and photos. How easy or hard would this feature be for a current user to adopt? And, how easy or hard would this feature be for a new user to adopt?

Action item: On a piece of paper, sketch out a couple of different workflows for your proposed product. These sketches don’t need to be pixel-perfect, as your goal is to explore a variety of design approaches to solving customer pain.

Then, compare your sketches against the responses above. For each challenge identified, how would you account for it by changing the workflow? 

And remember, the responses are not necessarily accurate. Are some of these points overblown? Are there missing points?

Now it’s time to dig deeper and determine where our customer segment might run into issues with our proposed workflow.

Frictions

Prompt: Play devil's advocate. What are some reasons why my users may not want me to launch this video feature for user profiles?

Prompt: Write an investment analyst report on our “video snippets” feature that identifies key weaknesses of our product that competitors might take advantage of.

Action item: Identify whether you agree with these concerns or not. Then, weigh each negative against the potential benefit that your proposed feature would unlock.

As a product manager, it’s up to you to decide whether this feature still makes sense to pursue. No generative AI response will help you make this kind of decision.

Assuming that we decide to move forward with this feature, let’s look for ways to enhance the desirability of our functionality for our targeted customers.

Desirability

Prompt: Say we move forward with launching a video feature for user profiles in our dating app. What is the most likely reaction from our current users & customers, and why?

Prompt: How could we make this feature even more desirable for our users & customers?

Prompt: Identify three ways in which this product might fail to gain traction with our users.

Action item: Evaluate these suggestions and determine which ones make sense in the specific context of our product. 

As an example, “video discovery” doesn’t make sense in the context of a dating app. The goal of our users is not to become the next Instagram influencer or TikTok star! They’re probably not interested in broadcasting their personal video responses to a large group of strangers.

Our customers want to connect with each other more easily and determine whether they’re interested in mutually pursuing a long term relationship together. We should use this filter to think through how to further increase the desirability of our product.

Now that we have a product that solves pain for customers and is something that they’re highly likely to use, our next step is to position our product to the market. After all, people need to find out about your product to use it!

Positioning

Prompt: In preparation for our upcoming launch of "video snippets on user profiles" for our dating app, write a press release.

Prompt: In preparation for our upcoming launch of "video snippets on user profiles" for our dating app, write a teaser on Instagram for this new feature. Limit your response to 100 words.

Prompt: Which go to market strategies make sense for this feature? Which go to market strategies do not make sense for this feature?

Action item: Analyze the proposed positioning. How would you tweak the positioning to better fit our customer segment? Which kinds of positioning might accidentally make users less likely to use your product?

We now have clarity on who our customers are, what their pains are, what a compelling product workflow might look like, and how to position it.

But, every product manager has to think like an investor. We have constrained bandwidth and resources, and we can’t build every feature all at once.

Our next step is to consider return on investment, where we maximize benefit for customers while also minimizing our costs. We need to identify how much to ship within our initial scope and determine how much effort it will take. And, we need to clearly identify the features that we will purposefully cut from our initial scope.

Return on investment

Prompt: Which business metrics would you measure to determine the success of our video snippets feature for user profiles on our dating app? Limit your response to only 3 metrics maximum.

Prompt: How complex is this video feature to implement? Rate it on a scale of 1-5, 1 being "trivial" (it takes less than 1 day of work for 1 engineer) and 5 being "immense" (it takes more than a year for a team of 10 engineers).

Prompt: Identify three unintended negative consequences that might happen if this product is very successful.

Action item: Using the responses above, finalize the scope of your product. How much benefit does each feature unlock, and which feature set is the most powerful to ship first?

As an example, we don’t really need to have integrations to TikTok to pipe in existing videos to Hinge. The benefit isn’t high, because many TikTok videos that users create are likely inappropriate for their dating profiles. And, the cost is non-trivial, because you have to build an integration to pull in their TikTok videos.

Once we’ve set our product scope, we need to measure how successful this product is, or else we won’t know the true ROI of our product. Consider whether the proposed metrics above make sense for proving that your feature is helping users, and identify whether there are better metrics that you could use instead.

We’ve now completed an analysis of a publicly available product that we didn’t personally build. By running through this exercise 3-5 more times, you’ll start to see patterns for the kinds of products that are most likely to succeed, as well as patterns for the kinds of products that are most likely to fail.

Accelerating product specs through generative AI

We can use the same prompts above to accelerate our own product specs. For your convenience, I’ve gathered these prompts for you to use below.

Product spec prompts

Customer segments

  • Which customer segments are most likely to use our product, and why?

  • Which customer segments are least likely to use our product, and why?

Pains

  • What processes do our customers currently use to address the problem that we want our product to solve, and why?

  • What kinds of pain points, problems, or frictions might they run into when they use their current methods, and why?

Competitors

  • What are the most common competitors they use, and why?

  • We want to get a customer to switch away from a competitor and get them to use our product. How would we convince them to switch? Why do these things matter to them?

Workflow

  • I have selected a specific customer pain to solve. I believe that my proposed feature can solve it. How much of this pain does the feature solve?

  • I’m interested in launching a specific feature. How easy or hard would this feature be for a current user to adopt? And, how easy or hard would this feature be for a new user to adopt?

Frictions

  • Play devil's advocate. What are some reasons why my users may not want me to launch this feature?

  • Write an investment analyst report on our feature that identifies key weaknesses of our product that competitors might take advantage of.

Desirability

  • Say we launched our feature. What is the most likely reaction from our current users & customers, and why?

  • How could we make this feature even more desirable for our users & customers?

  • Identify three ways in which this product might fail to gain traction with our users.

Positioning

  • In preparation for our upcoming feature launch, write a press release.

  • In preparation for our upcoming feature launch, write a teaser on Instagram for this new feature. Limit your response to 100 words.

  • Which go to market strategies make sense for this feature? Which go to market strategies do not make sense for this feature?


Return on investment

  • Which business metrics would you measure to determine the success of our product? Limit your response to only 3 metrics maximum.

  • How complex is this feature to implement? Rate it on a scale of 1-5, 1 being "trivial" (it takes less than 1 day of work for 1 engineer) and 5 being "immense" (it takes more than a year for a team of 10 engineers).

  • Identify three unintended negative consequences that might happen if this product is very successful.


The point of these prompts is to break through blank page syndrome. Keep in mind that our goal is not to automate away our own critical thinking!

In fact, given the lightning-fast iteration times that you get from using generative AI, we strongly recommend that you ask generative AI to play devil’s advocate.

Devil’s advocate prompts

Probe each of your product assumptions and look for ways in which your hypotheses might be wrong. Here are some devil’s advocate prompts you can use:

  • How might our proposed feature cannibalize or hurt our other products & features?

  • If we launch this product, would our users feel that we’re distracted & losing focus on our core purpose, or would they feel that we’re making progress?

  • Pretend to be a competitor. How would you reap the same benefits for users, but in a faster and cheaper way?

  • Pretend to be our finance team. What are some good reasons for why they should cut our funding and reallocate this spend to other features or initiatives? 

  • Pretend to be an external consulting firm. What are some specific market trends that would significantly reduce the usefulness of our proposed product?

  • What kind of data & signals should I look for to decide that I should kill off this feature idea and invest in something else? 

You can easily see how using generative AI can help you rapidly improve your product sense, as it can raise blind spots that we may not be aware of! 

As a reminder, you shouldn’t use generative AI to write every word of your product spec. Instead, you should use it to test your assumptions and better flesh out your understanding of your own product, including both its strengths and weaknesses.

As an aside, if your organization is looking to uplevel its product spec processes, we provide group workshops for product teams.

Closing thoughts

We can hone product sense by analyzing other people’s products and by writing up our own product specs. Generative AI helps us build product sense significantly faster with significantly less effort.

If you’re not sure how to get started, here are some products to practice breaking down:

  • Google: Gmail, Google Calendar, Google Maps, Google Drive

  • Facebook: Messenger, Groups, Events, Stories

  • Microsoft: Word, Excel, PowerPoint, OneDrive

  • Amazon: e-commerce recommendations, Amazon Prime, Alexa

Of course, running these product sense exercises in advance of any upcoming PM interviews will increase your chances of passing the PM interview and securing an offer. If you’d like more help with PM interviews, check out our on-demand course.

We hope this guide has provided you with clear guidance on how to successfully leverage generative AI to strengthen your product sense, as well as when not to use generative AI to make product decisions. We’re excited to see generative AI continue to unlock new potential!

And, on the topic of AI - when implementing & scaling AI products, don’t hesitate to leverage the above exercises as well. As a PM, you’ll want to tackle a discovery stage for your AI product, then validate your product, and finally scale its success. This AI implementation framework from our friends at Toptal contains valuable steps to build AI products in a way that aligns with your org’s product vision while maximizing the ROI that your teams can deliver.


Thank you to Pauli Bielewicz, Markus Seebauer, Juliet Chuang, and Kendra Ritterhern for making this guide possible.

Previous
Previous

Masterclass: Customer-Centric Product Adoption

Next
Next

Q&AI: Why Leadership Matters