The U.S.I.D.O. Framework for AI Product Managers
TLDR: AI can make you a better Product Manager by helping you Understand, Strategize, Ideate, Define and Optimize your product (I call this the U.S.I.D.O. Framework for AI Product Management).
In last week’s post, we outlined the four questions every Product Manager should ask to build great AI Products.
To recap they are:
- How can AI make me better in my role?
- What are the core use cases for my product?
- What are the best practices for working with with teams to build AI features?
- How can AI accelerate value/growth for my users/company?
This week, we’re diving deeper into the first question and asking:
How can AI make me better in my role as a Product Manager?
To start, it’s important to note that AI won’t be replacing Product Managers now or in the future.
Product Managers are here to stay. Here’s why…
Being customer-centric, having keen business sense, translating user needs into compelling features, collaborating with teams and managing stakeholders remain dynamic, indispensable skills. AI, in this context, is not a replacement but a powerful ally that can amplify your efforts so that you can continue to slay dragons and sling products. Think of AI as that super-smart intern who underpromises and overdelivers while you take all the credit…but, of course, you don’t do that, right?
Coffee, anyone?
But you still need to put in the work.
Okay, so AI won’t be replacing you now or in the future, but it is a tool you need to learn to help you augment, automate and speed up certain tasks. In this way it can help you build better digital experiences.
To structure our approach, it’s helpful to think about this across 5 key steps of the product lifecycle which I call the U.S.I.D.O Framework for AI Product Management because, yes, I’m obsessive compulsive with product frameworks.
After all:
If a tree falls in the forest and there was no product framework to capture it…
I digress.
Let’s dive in.
1. Understand
Not to pile on to all the hype and hyperbole, but it is undeniable that AI’s ability to recognize patterns and trends in data sets is a game-changer. It can sift through troves of both structured data (think product analytics) and unstructured data (think user reviews). It can spot trends and themes around customer pain points that may have taken you hours or days. And perhaps, even more importantly, it can spot patterns and generate insights that may have gone uncovered without it.
Rather than dedicating hours to combing through vast datasets, product managers can utilize AI to sift through and condense this information, allowing them to go deeper on specific themes that warrant closer inspection.
This can then power decision making and become a very powerful tool for things like product discovery, roadmap planning, and growth strategies.
Action Item #1: Use a tool like appbot to download a csv of all of your user reviews from the App Store. Input those into chatGPT with a custom prompt instructing it to list the top 5 user complaints.
Et voila…
2. Strategize
The next area where AI can help Product People is at the strategy stage. This can be for a new or existing product or feature. It can help you understand the market and competition. As an example, you can feed the LLM information about your company/product/users, etc. so it has context. With this context, it can help you identify your strengths, weaknesses, opportunities and threats (SWOT) and generate a detailed SWOT analysis to guide additional decisions across the product lifecycle like where to play, how to position yourself against competitors, and what opportunities to follow.
Strengths, weaknesses, and opportunities…oh my.
Here’s another potential use case. Imagine you have a video streaming product. You have some pretty big competitors with some pretty big moats. Netflix, Disney, Amazon and the like. A Blue Ocean Differentiation Canvas could be useful tool here to help you understand how you can differentiate yourself, create new demand and carve out a niche amongst these Goliaths.
Traditionally, a Blue Ocean Differentiation Canvas is a tool that can be used to help you measure how well you are differentiating against your competitors on key industry factors. AI can help you generate and list the differentiating factors that are relevant for the industry that your company is in and offer suggestions on where you should and should NOT play.
These are just a few examples, but you get the point:
AI can help you perform market/competitor analysis and generate strategic artefacts and insights to then guide product development efforts in addition to things like positioning, differentiation, and go-to-market strategies.
The key takeaway is to use AI to help you generate strategic insights to then guide product development — saving you hours or even days of work.
This gives you faster initial confidence that the ship is pointed in the right direction.
“chatGPT, please insert pirate phrase here…”
“Ai, Captain!”
(see what I did there?)
Action Item #2: Create a simple word doc that describes your company, product, users, and strategy. Then, build a custom prompt asking chatGPT to read this document and identify your strengths, weaknesses, opportunities, and threats. Ask it to then generate a concise SWOT analysis. Use this as a starting point and dive deeper into the insights that are most interesting/promising.
3. Ideate
This may be my favourite step because we transition from understanding and analyzing to creating. If you’ve followed the Action Items above, then you’ve trained chatGPT on your users, competitors, and market. It understands your user’s biggest pain points by analyzing your App Store reviews and it understands your biggest opportunities by performing a SWOT analysis against your competitors in the market. With this understanding it’s well-primed for some good ol’ LLM magic.
Template:
“chatGPT, based on {insert analysis}, what product features should I improve or build to {insert goal} with {insert user-type}”
Action Item #3: Use the template above and generate a prompt like:
“chatGPT, based on {the top five pain points you identified from user reviews and the top five opportunities from SWOT analysis}, what product features should I improve or build to {drive retention} with {new users}.”
This is just a boiler plate, but you can insert any goal/user-type of your choice and customize the prompt to focus on Adoption, Activation, Engagement, Retention, Revenue, etc…
Define
With a list of product features and enhancements that focus on your biggest pain points and opportunities you can now start prioritizing and defining a roadmap. And guess what, AI can help...
Roadmap
AI can help you prioritize which features to build based on their potential impact on product success. It can help you define your product roadmap by analyzing historical data and predicting the impact of specific features on strategic areas and KPIs like retention, user satisfaction and revenue.
Action Item #4: Prompt chatGPT to generate a roadmap using all of the output we generated from the previous action items as context:
“Based on all of the information provided and generated from previous steps: What are the immediate features/enhancements I should focus on? Can you organize the list of features above into a one year roadmap divided into quarters of the year?”
Personas, User Stories, and Release Notes
With its ability to analyze large amounts of data from multiple data sources, there is a world where AI can generate user stories and personas based on things like product usage, user surveys and more.
Action Item #5: Prompt chatGPT to generate user stories based on all of the context we’ve provided in the chat up until this point.
“Can you generate user stories for the first feature in the Gherkin format?”
This last prompt, I find particularly useful as it can save a huge amount of time.
It’s important to note, all of these steps require an AI-augmented approach. Meaning there needs to be a human in the loop (that’s you!). The outputs won’t be perfect. You will need to edit, refine, re-prompt, iterate and guide the LLM. And that’s a good thing. That’s why the AI overlords still need you and why you still have a job!
Aside: Yeah, but what about the Customer?
It goes without saying, you still need to be gathering real customer feedback across the entire product lifecycle. That should not change.
Another Aside: Persona bot, anyone?
One very interesting use case I can imagine in the not-so-distant-future would be building a persona bot that can take on the actual traits of the persona you are targeting. This bot could help stand-in as a proxy for talking to real customers once you start feeding it more and more customer feedback data. This kind of quick proxy could be really useful and much more rooted in reality that the user personas of today. After all, let’s be honest, maybe it’s just me, but I’ve always found user personas a tad fluffy, largely fabricated, and often disconnected from reality. Now, of course, Susie from Sarasota may disagree... We are all entitled to our opinions.
Last Aside (and then I promise I’m done): You aren’t as customer-centric as you say you are.
I’m not saying a ‘persona bot’ would replace the need to speak to real people. But it could help augment these conversations. Because, let’s be honest, again. As much as every product manager preaches about ‘being customer-centric’ and ‘getting-out-of-the-building’ and having real conversations — and as much as we all claim to be bonefied, certified customer heroes — many of us simply aren’t. We simply don’t have (or make) the time. Yes, I said it. You aren’t as customer-centric as you say you are. Deal with it.
Optimize
Okay, my apologies, back on track…
Using AI in the previous steps frees you up to spend more time optimizing your product. Optimization, in this context, falls into two buckets: Experimentation and Personalization. Experimenting refers to figuring out what works with users. Personalizing refers to building individualized bespoke experiences. Experimentation and Personalization are what will set your product apart as best-in-class. And really, everything we’ve done in previous steps is in service to our end goal of building fully optimized and personalized products that drive innovation, adoption, and growth.
Experimentation
Learn what’s working and implement changes quickly. All the time you’ve saved on combing through data and writing requirements by using AI frees you up to run more experiments. This lets you learn what’s working and implement changes quickly. For example, AI can suggest what to test for in terms of multivariate feature testing. AI can even run tests automatically.
Experimentation tools are crucial for product managers to test new features, understand user behaviour, and make data-driven decisions. These tools can vary based on the platform they support (mobile vs. web vs. multi-platform).
I’m partial to Google Firebase (for mobile experiments) and Google Optimize (for web experiments) given they are free, ubiquitous and easy to use. But there are many out there.
For Mobile: Firebase Remote Config allows developers to change the behaviour and appearance of their apps without requiring users to download an app update, enabling seamless A/B testing. For Web: Google Optimize integrates with Google Analytics and offers A/B testing, multivariate testing, and redirect tests.
Optimizely, Aptimize, and Amplitude Experiment are some other favourites.
When choosing an experimentation tool, consider factors like the platforms you’re targeting, the complexity of experiments you need to run, integration capabilities with your existing stack, and the level of analytics and insights provided. Each tool has its strengths and is designed to meet different needs, so it’s essential to evaluate them based on your specific requirements.
Action Item #6: Try out a few different experimentation and optimization tools for A/B Testing.
Personalization
Being able to analyze more product and user data means more ways to personalize the end product. The benefits of this include more tailored content and experiences as well as customized in-app-messaging and user journeys. All of this makes the user feel like the product was made for them. Being ‘bespoke’ will become table-stakes and a fundamental customer expectation of the future. Not being personalized will feel like a product defect.
Despite being a buzz word, “personalization” is hard and doesn’t come out-of-the-box. It starts with Product Managers identifying core use cases and opportunities for personalization. From there, it’s important to create a data map of the all the data you will need to support personalization, how it will be used, and the infrastructure that needs to be set up to capture, store and process the data in a way that makes it useful and helps you achieve your personalization objectives.
This may involve building models so you’ll need to work the both Data Engineers and Data Scientists to build out the right infrastructure architecture and models as well as MLOps to production-ize these models. And then iterate.
Like most things in tech, there is a lot of buzz around “personalization” but a serious gap around the strategic thinking, planning, and know-how needed to achieve best-in-class personalization. It takes a lot of pre-planning and work to make it work right. And often, the critical pre-planning gets passed over in an effort to pump out more features.
Do yourself a favour, encourage your org and stakeholders to slow down and get this one right. Otherwise the technical debt will be hard to dig yourself out from and you’ll fall behind with a dumb app.
Action Item #7: Identify the top use case for how you would like to see your product personalized. Then create a data-map and work with your Data & Analytics leads to build out your data-architecture, infrastructure and strategy. Implement. Start small.
Summary
Product Managers aren’t going away. Our roles are more crucial than ever. Being customer-centric, having keen business sense, translating user needs into compelling features, collaborating with teams and managing stakeholders remain dynamic indispensable skills. A world devoid of Product Managers results in ‘capability-driven’ products that nobody wants. However, the game has changed. AI is a powerful tool that all Product Managers must learn to leverage to be more effective in their roles. Or risk falling behind.
It’s helpful to think about leveraging AI across five key steps of the Product Lifecycle: Understanding, Strategizing, Ideating, Defining and Optimizing. Doing so can help you build better digital experiences while unlocking product value for both your users and your business.
