“An ounce of action is worth a ton of theory.” — Ralph Waldo Emerson
Today, you can listen to podcasts, watch videos, read blog posts, and have endless discussions on Linkedin and Twitter about AI.
But what really matters is what you put into action and how you actually use it to save you time, help you do things you couldn’t do before, or get better results.
From my observations, the delta between what is possible for PMs and what PMs are doing today is quite wide.
While AI trailblazers are going deep, building tools for themselves, making podcasts and courses to teach others, and creating new, innovative ways to do more, faster, most PMs are busy doing the work and many are falling behind.
As I wrote in my prior post on why good teams struggle to adopt AI, a big part of driving adoption is making it easier to try.
Today’s post is to help you and any PMs on your team quickly and easily catch up to what is possible today with AI.
How to Use AI to be a Better PM Today.
While you don’t hear much about AI hallucinating anymore, the fear still exists. It’s easy to think that spending a lot of time using AI just leads to wasted cycles and questionable results.
Fortunately, each of these suggestions are battle tested and well within the capabilities of AI. And where there are potential issues, I’ll call them out so you can keep them in mind.
Let’s dive in…
1) Improve Your Customer Interviews with AI
The strength of AI is text, so it’s no surprise that customer interviews is one of the best opportunities to use AI to help you throughout the process:
Before your customer interviews, you can use AI to help you do everything from:
- Brainstorm which customers are best to reach out to in your customer base and the right messaging to request they sign up for a call with you
- Generate the best questions to ask to get the insights you need most
- Get help tweaking your script to fill in a few gaps in your last few interviews
During the interview, AI is a tremendous help, too. It’s also the most likely place even PMs lagging behind are already using it thanks to the many note-taking solutions out there.
Gone are the days of needing to have the superhuman skill of being able to take notes while talking, always bringing a coworker with you, or having to re-listen to every single call and take notes yourself later. Instead, just use Granola, Fathom, or any number of other note-taking tools and you’ll have most of the heavy lifting done for you.
But that’s just the beginning.
After you have your 7-10 interviews done, it’s time to analyze what you learned, and that’s where AI can really shine.
Rather than having to spend hours pouring over notes and synthesizing your findings and common patterns, AI can do it for you in seconds, all with citations, extracted customer quotes, and whatever format you prefer most (slides, report, etc)
Remember: You get out what you put in.
You’ll notice that all of these suggestions involve prompting the AI to do work for you. It’s important to remember that you get out what you put in the chat window.
A simple, vague prompt will not get nearly as good of an answer as a detailed, specific one with supporting material.
Think about the first example I gave on the interview questions. Which would be better:
- Prompt 1: “Generate me some questions I can ask in my customer interview about our analytics dashboard”
vs. - Prompt 2: “Attached is my first draft of interview questions for asking our customers about ways we can improve our analytics dashboard. As you can see in the attached file, I have a list of common customer complaints that we want to try and understand how many each of our customers have. I’ve also attached a second document that includes the style of questions I generally like to ask. Please try and stick to what and how-based questions because that gets customers talking vs. yes or no leaves a lot you don’t know. Do you have any questions before we begin?”
This is why having an experimental mindset is so important when you’re a PM in the AI era. You need to regularly try tweaking what you’re doing to unlock new and better abilities. And now with Projects in ChatGPT or Claude (or Gems in Gemini), you can save your best stuff to reference again and again.
2) Prototype faster and find the best solution
Unless you’ve been sleeping under a rock for the last year or two, you know there are great prototyping tools out there, like Lovable, V0, and others. However, what you may not be thinking about is how to leverage that to not just do what you’ve always done, but do much more faster and better.
With AI, you can now generate 10 prototypes or mockups instead of one or two.
This makes you much more likely to find the global maxima (the best solution), not just a pretty good one that’s the first thing that came to mind.
More, faster, better competition = Higher stakes and standards
In our very competitive, fast-moving world of AI, it is incredibly important for the competitiveness of your team and business to create stand out solutions.
You don’t want to just embrace the AI design tools to quickly mock something up and be done with it. You should generate many more variants and then explore what’s actually best and what actually solves your customers’ problems.
This means taking some time to explore ideas that may seem crazy at first, because it’s not as expensive. You don’t need a designer to run off for a week, nor to find two afternoons to play around in Balsamiq.
You can literally test everything you (and AI) can think of simply by making a prompt, then go off and do something else, and come back twenty minutes later and have something worth testing.
From there, you can narrow it down to the best options to test in front of your customers, but the important step is expanding your options before that. Don’t rush this step and miss out on how AI can help you be more creative and innovative. (Note: Brian Balfour of Reforge has a great post on how to do this.)
A note on limitations on prototyping and design tools:
Now, the Achilles of these AI design tools right now is that they struggle to stick to your company’s style guide and existing design patterns.
That means you’re not replacing your designer any time soon. You’ll still need their help to make something pixel perfect and ready to be built by engineering.
Yet, you can explore A LOT before that. In my experience, Lovable and Make have been very capable of creating designs you can test in front of customers. You don’t need pixel perfect mocks to test that new interface idea you have and determine whether a table or card style is preferred.
Don’t let good enough get in the way of uncovering what can be turned into great, innovative solutions. And when you’re ready to go deeper, this (paywalled) post on Lenny’s blog by Colin Matthews shows how you can overcome some of these limitations with some effort.
3) Accelerate Analysis by Doing More Yourself
One of the age-old challenges of being a PM is wanting answers to a question that you can’t get the answer to without being in the backlog of a data analyst on your team.
That kind of handicapping slows down your analysis, or forces you to make decisions with a lot less information than you would like.
Fortunately, with the beauty of AI and MCPs, that limitation is going away. There is a lot you can do on your own, so let’s talk about it.
Claire Vo, host of the How I AI podcast, has a great episode covering exactly this:
Is this replacing your data analyst team? No. But it is allowing you to do a lot more than you used to.
A note on limitations of using AI to analyze your data:
It’s important to recognize that this does require more set up and help compared to the other areas we’ve discussed so far. There is up front effort required for this to work for you, and you may need help from others. That includes things like:
- Getting Read-Only Github access: If you want to use AI to search through deploys to find the cause of a drop you notice in your analytics, you need access to your team’s Github repo. You may need to convince a gatekeeper or two of the value of granting you those permissions, but read-only is a safe way to do so.
- Setting up MCPs: There are a lot of great MCPs for data products like Snowflake, Amplitude, and MixPanel. You will need to work to get that set up in your AI coding environment, and then experiment a bit to get comfortable with using them. You may also have restrictions at your company about what MCPs you can and cannot use, so use your best judgement.
- Telling the AI how your data is structured: This was kind of glossed over in the podcast above, but it’s critical. You need to explain to the AI how your tables and data is labeled if accessing a database. The good news is that you do this once (except for any major database updates), then you and your fellow PMs can all benefit. That’s a better ask of the data team than adding 5 more one-off questions to their backlog.
While this one is a steeper task to start using than other things we’ve discussed so far, the payoff is huge. The more you can own and understand your data, the better and faster you can make decisions. Stop making compromises on what data you can get, or being stuck in analyst backlog purgatory.
As Claire said in the podcast, becoming data fluent transformed her career. With AI, there has never been an easier time for you to become data fluent, too.
4) Write better Product Specs Your Team & Boss will ❤️
Product specs are one of the most common issues I see with product managers across skill levels.
It’s hard to strike that perfect balance of having done sufficient research, distilling all your most important learnings, and clearly and concisely conveying everything your designer and engineers need to know.
Writing is a skill that is hard for some to master, especially if you don’t do a lot of reading or writing on your own (this is part of why I read books regularly, and write here).
But with the rise of AI, there is now zero excuse for writing a bad product spec.
You can now use AI to help you every step of the way in writing your spec:
- Before: It can help you organize all your findings across surveys, interviews, data analysis, etc to find the most important and compelling data to share.
- During: AI can apply any template you give it to help you write a great product spec. You can even use voice to text tools like WisprFlow to give AI your ramblings and have it condense it for you.
- After: Turn a good product spec into a great one by having AI provide feedback on your spec, identifying key gaps, tightening your wording, and fitting your company’s style. AI can also help you pull sources and create a powerful “Further Reading” area to back up your conclusions.
The only limit on the quality of your product spec is your willingness to experiment with AI to get the best possible results.
AI can also help with your other communication.
While your product spec is one of the most important pieces of communication and decision making you have as a PM, it’s obviously not the only one. You have many other stakeholders and situations where you need to communicate in a variety of formats and locations.
Fortunately, AI can help you with these, too.
If you give AI some examples (or a template) of other communications you need to make, it can generate all the other formats you need building on the same information in your product spec, including: (credit to Carl Vellotti for inspiration for this list from his Claude Code for PMs course)
- A quick summary email to send to your boss with a link to the document if they want to go deeper.
- A Slack post letting your team know what’s coming next with some key bullets and highlights.
- A ready-made message to share with marketing, so they can start collaborating with you on a launch strategy for what’s coming.
- A slide that summarizes your plan that your boss can put in their deck that goes to the leadership team.
What might have taken you all afternoon to prepare after writing your product spec can now be done in a matter of minutes.
Obviously, you still need to proof-read it, but especially as you improve your prompts and templates, it should require less and less edits before being ready to send.
That’s the beauty of using AI; your investments compound and help you move faster and faster as you iterate.
5) For the Managers & Product Ops People: Build leverage for all of your PMs with AI
As the saying goes, becoming a leader is about shifting your mindset “from me to we.” That means it’s your job to empower your team and make it easier for them to do their work. That includes how they use AI.
While many of the tactics we’ve gone over so far can be done individually by your PMs, there are many things you can do that would create tremendous leverage for your team.
Here’s a few of the things you can do that would increase AI adoption by your PMs, save them time, and lead to greater velocity of your product teams:
- Share Templates: Standardizing the questions answered in your product specs is a great way to ensure all your PMs are creating clear, concise, thoughtful plans. You can also provide templates for the items we mentioned earlier for communication, and thought partners that review and challenge their docs.
- Make Your Templates Living Documents: It’s easy to share a template one time and be done, but the best approach is to make them living documents. Do this by setting up a Github repo (example here) so it evolves and improves over time as you learn where your team needs more guidance or different structure.
- Create a Manager Readme: These used to be controversial, but now with AI, they make all the sense in the world. Make a document they can upload to their favorite AI that explains your work style, communication preferences, and expectations. This will help them give you more of what you want faster. Learn more about this from Hiten Shah on How I AI here.
- Push for the data access and structure: In point 3, we talked about hurdles with getting access to data that lets your team answer their own questions with AI’s help. You can push for that to be a higher priority. When you ask, you have both the authority of your level in the organization and that it will benefit all the PMs, not just one eager AI PM asking.
- Gather the component libraries to share with everyone: Once again, if items are shared across the org, everyone benefits. Whether you get help from an engineer or designer, or you find one of your enterprising PMs already started it, getting this in the hands of all of your PMs will make everyone’s prototypes faster and better.
Remember: 1 PM saving an hour a week is ~50 hours a year saved, but if you save 8 PMs an hour a week, that’s 400 hours a year saved. These are high ROI efforts for you to take on.
But what if I don’t have time to learn AI?!? 😫
One of the most common things you hear from PMs, especially those in leadership roles, is that they just don’t have time to learn all of this. Their calendar is overflowing, so there’s no room for anything else.
And that’s understandable.
But the reality is, the world is changing and you must evolve or die.
As the saying goes, “AI won’t take your job, but someone who uses it will.”
That means you need to make time for this.
If your calendar is out of control, fix that. Audit your meetings as Peter Yang suggests:
“So what’s the solution? For me, it’s being very intentional about protecting my time by:
- Defaulting to async. I’ve become very good at async Slack threads.
- Auditing recurring meetings. I ask: “Does this meeting still need to exist?”
- Batching meetings. I try to cluster meetings so I have mornings for deep work.”
And if you feel it’s really bad, then consider applying some of these 10 tactics like the “Armeetingeddon”.
Bonus: Skate where the puck is going by becoming a true AI PM.
As I wrote in my last post on AI adoption, it’s important to make it easy to get started, and so far, these recommendations have been things that should feel very accessible. You can do most of these starting today, or for the data ones, at least get the ball rolling towards them without tremendous hurdles.
However, it’s important to keep in mind that everything we’ve discussed so far are table stakes for the best companies in the world right now. They’re expected.
So for those of you that read all of this and so far have thought, “yeah, I know all this…doing that…doing that…doing that….Where’s the advanced stuff?” this is for you.
What is the new AI PM lifecycle?
Ethan Mollick, a great AI researcher worth following, sums up the new AI lifecycle well: it’s all about speed and agility.
It’s about building small, scrappy product teams that can move fast to create great solutions that add value for your customers; it’s everything product management has always been about, but optimized for speed and leveraging AI every step of the way.
This image from an in depth post on Lenny’s blog captures it well:
Notice that you can literally do steps 1-3 in a morning. And if you can build a faster way to get in front of your customers, you may be able to do everything but the “Delivery” step all in one day.
This is the new bar to aspire to:
How do I raise my bar and build faster with AI?
If you want to build with the speed of what you see people bragging about on X, Linkedin, and podcasts, the key muscle you need to build is asking yourself 2 questions:
- How can I use AI to help me do this step in my process?
- What are my biggest bottlenecks to going faster?
If you bring an iterative mindset and a desire to learn and experiment, you will quickly find there’s always a step in the process that could use improving.
A few examples that align with things we’ve already talked about:
- Get higher quality designs faster by getting your component library in your prototyping tool of choice.
- Find customers to talk to faster by improving systems inside your company, or with the help of external tools like Great Question, User Testing, or Voice Panel.
- Get insights and prioritize problems faster by building a pipeline to go from customer interviews to AI analysis and summarization to product spec faster.
- Make your first draft from AI closer to your final output by adding extra notes to your templates and reused prompts based on what you find you most commonly have to iterate or tweak.
The key is to first figure out how AI can help you do something. Then, it’s about how to combine steps, link tools, improve approaches, and iterate.
Whether you like it or not, yes, this means your product development process itself is becoming a second product for you to manage.
If you have a Product Operations team, now is a good time to get to know them better and lean on them for help where you’re getting stuck. If you don’t have one, now has never been a better time to add one, because they can have a massive impact on your product velocity if they’re embracing AI and can help drive adoption across your product org.
Regardless, the message today is simple: AI is making everything faster.
You can get more done, explore more ideas, and delight your customers in ways you couldn’t before. And the bar is steadily rising.
If you love learning and have an experimental mindset, the world is your oyster right now. If you were hoping to coast on the same things you were doing 4-5 years ago, you’re in for a rude awakening as you get passed by.
Now is the time to roll up your sleeves and make the time to catch up and start asking yourself and your team the 2 big questions I outlined.
How will you go faster with AI?
If you or your organization are struggling with this, or simply want to level up from the basics, I can help you.
I’m available for consulting projects, and for the right opportunity, a full time role to work with you. Sign up for a call here, or email me your questions to jason at becustomerdriven dot com.