Productside Webinar

AI Strategy in Product Management

Where It Works, Where It Doesn't, and What Leaders Need to Rethink

Date:

09/30/2026

Time EST:

1:00 pm
Watch Now

Everybody’s run an AI experiment. Almost nobody’s written an AI strategy. 

AI is changing how Product Management organizations operate, how product teams make decisions, and how companies create value for customers. But the results are uneven.  

Some organizations are seeing faster learning, better decisions, and differentiated customer experiences. 
Others are adding cost, automating the wrong work, or failing to see meaningful ROI.  

In this session, we look at what is working, what is failing, what the economics are telling us, and what leaders should rethink heading into 2027. 

What You Will Learn: 

  • Tell AI table stakes apart from real competitive advantage 
  • Judge AI strategy at three levels: your PM organization, your product teams, and your customer-facing product 
  • Learn from recent AI wins, failures, and quiet retreats 
  • Turn economics, assurance, and human judgment into a Now / Next / Watch roadmap for 2027 

Welcome and Introductions

Cynthia Petti | 00:00:00 – 00:03:25
All right, let’s get moving. Those who are late, just don’t get fruit cup.

Cynthia Petti | 00:03:25 – 00:03:50
Welcome, everyone. Thank you so much for joining us here today. I’m Cynthia Petti, the COO at Productside, and yes, I am located in Austin, Texas. And I’m here with the pleasure of Dean Peters, our very own principal consultant and trainer, who is such a genius in regard to topics on AI. So, Dean, tell us a little bit about yourself.

Dean Peters | 00:03:50 – 00:04:34
All right, I’ll tell you a little bit. Although, how do I beat an entry like that? And first of all, let’s give some props to Cynthia, because I’m excited, because we’re going to be talking about organizational strategy. I can’t think of a better person to join me on this conversation. What she knows about operations and executive operating systems is going to be really valuable to us.

As for myself, I’ve been with Productside about four and a half years. Before that, 20 years of product management, and then about 15 years of engineering. And if you look at my LinkedIn profile, you will see that I’ve done quite a bit with AI, both from the ground level as well as the strategic level. And I think these days we keep talking tactics. It’s time to talk strategy. So, I’m very excited about this.

Housekeeping, Recordings, and Community Links

Cynthia Petti | 00:04:34 – 00:07:02
Yeah. And just one thing for everyone real quick as we get through some announcements: there will be polls, poll questions. So please answer them, because then it just kind of adds to where everyone’s at and perhaps gets a deeper learning on today’s topic.

And, as we said, we’re Productside. We know how hard it is to build products that people not only trust but love, and that they use. Whether it’s getting alignment from stakeholders, anticipating what the customer needs, or keeping that backlog from taking over your life, we understand that pain, and that’s why we’re here, because we’ve helped hundreds of teams work through it. That’s what we do. So we don’t think of ourselves as just coming in to help; we are your outcome-driven product partner. So just know that we are there alongside you. We have our playbook. We have so many things out there, and we are your trusted experts within product management. So I just wanted to tell you a little bit about us.

And the number one question we do get is, “Can I watch this later?” Yes, it is being recorded. So, you’re going to get to hear all of my ums and ahs, the questions that you got answered, and just have that learning at your fingertips to have it soak in. So, know that that is going to be taking place.

There is a Q&A feature. We already do have some questions, so thank you for those that submitted some questions, because we’ll be interjecting them, and my job is to say, “Hey, Dean, we got this question.” So please go ahead and use that. And also, there’s the chat, so you can also chat along with the webinar with fellow attendees.

And I think from there, we can go to this other one, which is our LinkedIn community. We are on LinkedIn, right? We are there with the product management community. We share things on best practices. We discuss topics. I’m sure there’s a million other topics in there, but these are real people, real product managers, in the day in, day out. And it’s a network of peers. So, please go ahead and join that if you’re not already. And if you are already, go ahead and share it with someone who might want to join as well.

Dean Peters | 00:07:02 – 00:07:10
Yeah, we really want to give everyone in the community a voice. All right. So, keep those comments and those questions coming. Join us online and keep those coming that way.

Webinar Agenda Overview

Dean Peters | 00:07:10 – 00:07:41
Here, let’s talk about the agenda. I’m not going to read the slide. Nobody likes that. I’m going to tell you, in about the next 40 minutes, we’re going to have an exciting conversation around AI strategy from a viewpoint I think is going to be unique, that you may not have heard. We’re going to tell you what the viewpoint is. We’re going to show you some of the good, the bad, and the ugly of that. We’re even going to give you tactical practicals, so as product managers, you can go ahead and assert your influence without authority to make the magic happen.

The Strategic AI Problem and Failures

The AI Adoption vs. Strategy Gap

Dean Peters | 00:07:41 – 00:08:59
So with that, let’s talk about the problem first. Because we are product managers, we live by the credo of “What problem are we solving, and for whom?” And we identified a problem in our 2026 State of AI Product Management report. So many of us are using AI at work, and so few of us have defined the why and the when, the strategy behind it.

You look at that first stat line here. All right, 80% of product managers use AI regularly in their daily work. Okay, good number. Then you look across the alley there and you say, only 23% in an organization do it with a clear strategy. Probably something that keeps a lot of CEOs and CPOs awake at night.

I think the other stat I was looking at was 43%: they say the use of AI is expected. They’re getting pressure from their bosses saying, “Hey, we gave you a Copilot license. Where’s the payout?” Yet only 34% have a documented, findable policy. So this is going to be a talk where we involve culture, where we involve a little more than just models and various tools.

Conway’s Law and Notable Strategic AI Failures

Dean Peters | 00:08:59 – 00:09:20
So speaking of culture over models and tools, Cynthia, in our conversations, were you aware of this principle called Conway’s Law, where the architecture of a software system mirrors the communication structure of the organization that builds the product?

Cynthia Petti | 00:09:20 – 00:09:29
Yeah. When you were showing me this, Dean, I was like, okay, I need to know this. I really want to understand this even better. And some of us have actually lived this.

Dean Peters | 00:09:29 – 00:14:32
I’ve been in organizations that were Microsoft-shaped, where they call it the guns out, and I’ve been in the Meta or the Facebook type shape. And it’s interesting: the products that actually come out of those organizations actually wind up reflecting the communications of the organization. So as we think about the problem we’re trying to solve there, everyone’s being asked to use AI, and everyone’s using it organizationally. We really haven’t talked about how we use it in our communications, how we use it across our organizations, how we use it strategically, and we have notable failures due to these strategy gaps.

This report here from S&P 500 basically says that out of the survey respondents from the S&P 500, 42% of their organizations abandoned 50% of their AI proof-of-life products. So they were building a proof of life with AI, and they were chugging along, and then somebody just said, “Nope, we’re not funding this anymore.” So 50% of those got the axe from 42% of the S&P 500.

We have this wonder called the AI Pin, where I think they got into the trap of novelty. It has a lot of the same characteristics, or a lot of the same aspects to it, that we saw with Google Glass maybe a decade earlier. Basically, it’s a wearable pin. Imagine going to a party, Cynthia, wearing it. You’re adorning it there, and you sort of turn it on there. Oh, by the way, it projects the screen. So now I’ve got my hand in my face to see the screen while I’m talking to you, which I’m sure created just a wonderful user experience there. So, I think there was an AI fail there.

There was a company called Relay.app, which was basically doing what Zapier did, only they were better. They didn’t quite explain how they were better. In fact, what they did was so copyable. Zapier has written, quite frankly, quite often on how they basically did the same thing.

Gemini, about a year, year and a half ago, had this disaster with image generation. A lot of people were doing historic images, which were a little bit suspect. What was interesting was that they were overoptimizing on diversity and inclusion. We want diversity and inclusion in our AI models, but what we don’t want to do is have it be the only metric and overoptimize on that. In fact, I’ll talk about that in a minute.

Of course, we have the poster child for AI gone wrong strategically here, which is Air Canada. Basically, the AI made up a refund policy on the fly. It inferred it. People took it to court. A person had been told by the AI, “Oh, you can get bereavement refunds, sure.” Of course, Air Canada said, “Nope, not us,” and they passed the buck, and they passed the buck until it went to court. And basically it was like, yep: everybody owned creating the product; nobody owned the outcomes of the product.

So we have these, and the point I want to make with these is that each of these failures suffered a strategic dysfunction that wasn’t really related to the AI model, or wasn’t very closely related to the AI model. Or, if it was based on the AI model, it was because the culture around the model, the people building it, had those dysfunctions. So, we want to talk about how Conway’s Law intersects with AI strategies there, and how it actually trickles down and seeps into the product.

So, here again are these notable failures, and I want to put this table out here. You’ll notice on the left the strategic dysfunction of these various items here. We see many reports of companies spending without finding value in their AI product. It’s not just this S&P Global report. MIT’s had a report on this. McKinsey, there have been several news items on people just not seeing where the value’s got to beat the organization-wide cost. We see with products like Humane that they chase the hype; they chase the novelty instead of the customer job. We see with Relay that a wrapper to a model is not a moat. So basically, the organization never sat down and said, “But how are we different? What makes us unique?” I think, Cynthia, we have a lot of conversations about that at Productside. What makes us unique? What helps us deliver value? So you see three of these strategic points are things we talk about quite often, or the team overoptimizing on a single risk, or everyone being the builder and nobody being the owner. So we’re going to talk about… Yeah, go ahead.

Cynthia Petti | 00:14:32 – 00:14:49
I’m so sorry, Dean. What I was going to say is, these are highlighted things that are on the news, right? But there’s so much push to also put AI in the product. That’s not even here, but that is a reality also.

Dean Peters | 00:14:49 – 00:15:10
Oh, yeah. The HiPPO waddles in and bellows, “Add AI to our product,” or “When are we going to see AI on our roadmap?” or “When are we going to see the AI roadmap?” And we’re getting a lot of downward pressure for that. Meanwhile, the people in the C-suite aren’t talking about the culture, or the communication structure, or the process structure to make that happen.

Audience Poll: Workplace AI Strategy Status

Dean Peters | 00:15:10 – 00:15:38
So I think it’s probably time for a poll. So, we’re going to run a poll here. And really, what I want you to do is identify which AI strategy is the most developed at your place of work. So, we’re going to do it in a positive light. These are anonymous. Your bosses will not get it, but we just want to go ahead and get a little feel for the room before we move on.

Poll Question: Which AI strategy is the most developed at your place of work?

Cynthia Petti | 00:15:38 – 00:15:47
We got people responding. Everyone still has some time to click a button there.

Dean Peters | 00:15:47
And we’re not looking for precision. Like I said, we just want to get…

Cynthia Petti | 00:15:49
Everyone gets an A. Everyone gets an A for answering.

Dean Peters | 00:15:52 – 00:16:01
Yes. Well, the only way not to win the game is not to play.

Cynthia Petti | 00:16:01
Great.

Dean Peters | 00:16:02 – 00:16:07
All righty. Give it about another couple seconds there. We’ll end the poll and see some of the results.

Cynthia Petti | 00:16:07
You let me know when you want to end the poll, Dean.

Dean Peters | 00:16:10
That’s right. I’ve got a button here, too.

Cynthia Petti | 00:16:11
Perfect.

Dean Peters | 00:16:13 – 00:16:21
And by the way, the button is red for everyone. So, it’s kind of like an alarm.

Dean Peters | 00:16:21 – 00:17:08
All righty. We’ve ended the poll here. Let’s go ahead and share the results. Look at this. All right. Let me see here. We’ve got organization-wide, so 22% or so. We’ve got a good mix here. All right. So, it looks like… I mean, what we’re really saying here is we have a mix that’s not fully defined. So, where we have there, look at that. That’s right around that 44% again, which, we were talking about that S&P 44% number here. I think they have some client-facing strategy as well. So, people actually have some strategy here, and that’s encouraging. So, hopefully what we’re going to show you is going to help augment that. And for those of you who don’t have it, don’t worry. We’re going to get you in the game.

Poll Results (as referenced during the session):

  • Organization-wide strategy: approximately 22%
  • Mixed, not fully defined: approximately 44%
  • Client-facing strategy: some respondents (percentage not stated)

Results showed a mix across the room, with the largest share of respondents reporting an AI strategy that is not yet fully defined.

A Layered AI Strategy Framework

Overview of the Three-Tier Strategy Model

Dean Peters | 00:17:08 – 00:19:06
So, let’s talk about the model here. We wanted to create a model that wasn’t going to be heavyweight, that’s going to simply answer the question: as product managers, how might we avoid dysfunctional AI strategies? How might we do that? And how might we do it in a way that doesn’t require a parade of certifications, years and years of training, various levels here, lots of change management and expense? So, a very lightweight model here at three distinct levels:

  • One is the organizational strategy. So think of that like your business unit for your organization if you’re at a very large enterprise. If you’re somewhat of a smaller company, think about your overall business’s strategy there.
  • Think about the next level being your product team strategy. All right? So, as product managers within a group, maybe within a business unit here, what is your strategy?
  • And then, what is the strategy as it relates to the product itself and its impact?

And you’ll notice that in the first two levels, we have the table stakes, the things we must have within any of our governance. And then we have those differentiators on the right-hand side, the things that help us get to the next level. So let’s dig into each one of these.

Right before I do that, you’ll notice there’s some governance here, organization-wide governance across:

  • Privacy
  • Security
  • Safety
  • Compliance
  • Auditability
  • Scalability

So again, what we’re really talking about is a layered version of an AI strategy. So it’s not just one big, what would you call it, blob of strategy here. We’re actually doing it at layers that are light and easy to do at each of these levels here.

Strategy Ownership and Cross-Functional Alignment

Dean Peters | 00:19:06 – 00:19:15
I think we had a question here. I think there was a question that we got about at what level, or where, the strategy is defined. Cynthia, did we have a question like that?

Cynthia Petti | 00:19:15 – 00:19:24
Yeah, it was mainly about who should own the AI strategy. We have product, we have engineering, or is it a separate AI team?

Dean Peters | 00:19:24 – 00:20:05
Right. And I would suggest that, like security, you don’t do software security with just one person. It’s layered. So you’re going to have ownership at different layers. I would want someone at the executive level to have ownership from a macro point of view. I’d want to have someone like maybe a director of product or a VP of product looking at it from the product team level. And I think as we have AI products in our strategies, those who own the products need to have a strategy that ladders up and reflects the principles that have enabled it from the top down. So it’s both top-down and bottom-up here.

Cynthia Petti | 00:20:05 – 00:20:30
Yeah. And I just want to also highlight how you mentioned the assurance across the life cycle, right? The other thing to think about from a governance and compliance capability is laws. What’s coming up? What should we know? How does this impact us? So that’s a huge point, because, Dean, you’re saying that organizationally, you’re going to have legal somewhere in here.

Dean Peters | 00:20:30 – 00:21:48
Right. And that has to be done. For example, in our AI class, we use the PESTLE exercise. So we actually look at those six areas:

  • Political
  • Economic
  • Social
  • Technological
  • Ecological
  • Legal

And what we ask is, “Hey, in these six areas, what do you need to act on right away? What do you need to put a monitor on and keep abreast of?” And that probably needs to be done at each of these different levels, because I would imagine, for example, you’re working on a product, let’s say a product working on maybe upstream oil and gas, that’s going to be heavily impacted. We have a lot of talk about AI legislation happening at the state level. Imagine 50 different states in the US, along with various other countries, all these different levels.

So this is the type of thing we’re trying to talk about. It’s owned at these different layers. While the head of the organization owns the high-level constitutional governance for the organization, at the product team there needs to be ownership, and then at the product level there needs to be ownership.

Organizational, Team, and Product-Level Strategy

Dean Peters | 00:21:48 – 00:23:06
So let’s get into each of these layers, since we can talk about this. And this is where we would hope the C-suite, or maybe some of the VPs, would have a conversation about how this governance, how these table stakes, are going to keep us in the game. You mentioned all this changing legislation, and not keeping abreast of it can very easily disrupt what you’re trying to get out the door, what you’re trying to do over the next year. It’s the differentiators that help keep you in the game.

So at an organizational level, one of the things I love, working here at Productside: Rina has said, “You are responsible for whatever the AI outputs.” You own whatever the output is. So we shouldn’t have an Air Canada-like miss. Those are table stakes. Anything that comes out of our chatbots, anything that comes out of Claude Code, we are held responsible for, and the line is very thick and not muddy. I would imagine we also have some other areas of governance that, as an organization, we need to practice that are table stakes, and then some that are differentiators as well. I don’t know if there’s anything that comes to mind for you, Cynthia.

Cynthia Petti | 00:23:06 – 00:23:25
Really, it’s just that even though it’s written a certain way, that this is what we’re abiding by, it’s always to be checking in on these changes. And overall, we as humans, we own it. We can’t just say, “Okay.” Right? We have that judgment as people.

Dean Peters | 00:23:25 – 00:27:52
Yeah. We can’t do what Air Canada actually did: sat before a judge and said, “Well, the AI said it. We’re not responsible for that.” As we like to say here in North Carolina, that dog don’t hunt.

All right. So, if you have this, you also can then say, “Hey, by the way, downstream, you need to keep looking for data opportunities. You need to keep looking for strategic judgment opportunities.” I look at what Relay did. They were basically getting into a very crowded space, and they kept talking about, “But we add human in the loop. We add human in the loop.” All right. And they had this big campaign about it. And about a week later, Zapier magically had human in the loop.

So what they should have said is, “Hey, we don’t just have human in the loop. We’re going to also give some judgment. We’re going to give you some data. We’re going to give you some information. We’re going to give you some recommendations, so you, as the human in the loop, can make better, more data-informed decisions.” That would have been a differentiator, versus just saying, “Hey, we’re just going to send a text message saying, ‘Hey, human, we think we’ve got an issue here.’” I would like to see some recommendations. I’d like to see some data, so I make the best decision. That would have been a differentiator. That would have been much harder for Zapier to copy.

At a team level, the team inherits what’s coming down, but they also look at this. We talk about better workflows and decision-making here. We talk about that culture of discovery and experiments and evaluations and telemetry. So as a team, and this needs to be a product team sport here, all right, AI should sharpen decisions, not just crank out more docs and more apps. As I like to say, AI is a great amplifier, but it doesn’t really care whether it amplifies function or dysfunction. Or, garbage in and crap out. It’s like, yeah, I can use AI to get crap done 10 times faster. At the end of the day, it’s still going to be 10 times the crap out the door. So again, we want to fix our broken workflows first, then speed them up.

I think everyone experimenting with agents last year found that out the hard way, where they were just throwing out agents. Or again, the Humane AI Pin. All right, they really weren’t tracking how people were or were not using this. So here they were, all, as I like to say, jazzed up about the pizzazz it had. “Ooh, look, I’ve got a little pin. Ooh, look how neat this is.” But they never really had anything that told them, “Well, actually, a lot of people may be wearing it, but they’re not actually using it in a lot of circumstances,” or, “Oh, wow, maybe we’re getting a lot of pushback in the same way Google Glass got pushback.” So that was an area at the team level.

Or Gemini, like I said, with AI-assisted decisions. They were actually using that one metric of diversity to rewrite the prompt after you wrote it, and then send it downstream, which again then sends a biased prompt that’s very narrowly focused. It wasn’t the model’s fault. The model did what it was asked to do.

All righty. And then at the product level here, again, we’re product managers. We need to own this ourselves. If you’re not a VP, if you’re not a director, you need to be owning this. Again, AI is going to amplify your weaknesses or strengths. It’s up to you which one it amplifies there. So:

  • It needs to start with the customer job.
  • It needs to be about building a trusted experience.
  • It needs to be making sure that our data is fresh, up-to-date, and not biased.
  • It means optimizing on outcomes instead of outputs.
  • It means rethinking the business model here.

So again, all these things are things that we need to think about at these different layers. So we don’t have one person owning it all. We have a distributed model where people are all pulling in the same direction. Before I go on, are there any questions that we have here?

Cynthia Petti | 00:27:52 – 00:28:15
Well, I have one question here, Dean. Actually, hold on. Okay. How do smaller product teams… where is it that they should focus? We have this big structure, and budget is one thing, right? But if it was a smaller product team, where would you say to focus?

Aligning AI Strategy with the Product Life Cycle

Dean Peters | 00:28:15 – 00:29:56
Yeah. And I think that’s good. That gets into the next slide here, which is: how do we take a big question like you have and break it into smaller parts? So whether you have a large or a small organization, think about your AI strategy, especially at a smaller organization, in terms of what stage of the product life cycle your product, or the various products in your product line, are at. So you’ll notice I’ve got:

  • Introduction
  • Growth
  • Maturity
  • Decline

We’ve all seen that flow there. What I’m suggesting is, if you notice these rows here below it, what’s the marketing objective? So your strategy isn’t going to remain static, Cynthia. It’s going to evolve and grow as the product evolves and grows. So I think this gives direction to what tactical actions we need to take.

What do I need to do as a product manager on a small team here? Well, again, having a marketing objective might be slightly different here. Understanding what that is. So that means getting the right people in the room and defining that. Same thing with our pricing and promotion; it’s going to have to be a little different here. Same thing with our data strategy, which nobody seems to be thinking about in some of these products. How do I create a virtuous cycle of data? So the more use I get of my product, the more data I collect, which means the better algorithms I can deliver, which means I get more product usage. How do I feed that flywheel? How do I avoid things like drift? So to answer your question there, I think one thing is to take that big question and break it down even further.

Cynthia Petti | 00:29:56
Yeah.

Dean Peters | 00:29:56 – 00:30:32
So let’s talk about the diagnosis, right? So we’re going to assume that people are bought in, like, “All right, we buy the model. We understand it. I might not do it exactly like this. I may not agree with everything, but how do I go about seeing what I can do as a product manager?” And I think what I really want to do is diagnose the usual suspects:

  • The wrong work
  • Runaway costs
  • Bad data
  • No trust

I love making this slide so much, but the usual suspects here.

Metrics and Diagnosing AI Initiatives

Audience Poll: Reasons AI Initiatives Undeliver

Dean Peters | 00:30:32 – 00:31:03
So, oh, I think it’s time for a poll first, right? So, everybody out in our audience here, before we talk about the diagnosis, let’s talk about the usual suspects here. All right, let’s get your view of what you feel are the main reasons AI initiatives underdeliver. Another way to say that is, who do you suspect are the usual suspects? Who are those usual suspects that we probably need to diagnose?

Poll Question: What do you feel are the main reasons AI initiatives underdeliver?

Cynthia Petti | 00:31:03 – 00:31:13
And I’m sure, depending on where you’re at in the adoption and use of it, it’s going to fluctuate for everyone as well.

Dean Peters | 00:31:13
Oh, yeah. One size does not fit all.

Cynthia Petti | 00:31:16
Congratulations, you used all your tokens.

Dean Peters | 00:31:19 – 00:31:30
Oh, yeah, right. Who was that? I forget what organization used up, in a month, their year’s allocation for tokens, because they were just saying, “Oh, yeah, just prototype.”

Cynthia Petti | 00:31:30
No governance, right, on the cost there.

Dean Peters | 00:31:35 – 00:31:43
Yeah. All righty. Let’s see. Okay, let’s end the poll. Let’s see what the results are. Wait, we’ve got a few more. Wait a second.

Cynthia Petti | 00:31:43
That was your chance, everyone.

Dean Peters | 00:31:46 – 00:32:21
All righty. I’m going to end the poll. I’m going to share the results, and let’s see what we got. All righty. Okay. So, cool. So, the wrong work was automated. Yes. What I like to call agentic feral cats. They’re running around our ecosystem now, right? Tipping over trash cans, getting in fights at night. Poor data, or poor integrations. Okay. And then the success metrics are unclear. So let’s talk about metrics.

Poll Results (as referenced during the session):

  • The wrong work was automated
  • Poor data or poor integrations
  • Success metrics are unclear

Automating the wrong work was the leading response, followed by poor data or integrations and unclear success metrics.

Defining Success Metrics (Time-to-Value)

Dean Peters | 00:32:21 – 00:33:01
All right, let’s talk about metrics and how they’re going to impact all this here. So how do we measure if we’re keeping or breaking our promises to our customer? That’s the way I like to frame this. In fact, when we teach our Optimal Product Management class, when we talk about success and health metrics, that’s how I like to talk about it. When I talk about success metrics, I always say, “Look, just ask yourself: are we keeping our promises to the customer?” And it’s interesting, when I have this conversation, some people say, “Well, we’re going to go ahead and improve the UX.” And it’s like, “Yeah, but where does that move the needle?” “Oh, well, we’re going to make people feel better.” “Well, where does that move the needle here?”

Cynthia Petti | 00:33:01 – 00:33:13
How do you judge and rank and get a number for feeling better?

Dean Peters | 00:33:13 – 00:34:59
Well, yeah, right. What’s feeling better? What are these subjective feelings here? And that’s exactly the question I ask. It’s like, “Okay, but those are really hard to measure, and not everyone agrees.” So, why don’t we talk instead about how fast we got to… time to job to be done. In other words, all right, I don’t want to say this, but I’m going to say it this way. Let’s see if I don’t trip over my tongue. The time to the job to be done getting done. All right. So, how fast was it done? We’re not all for speed, but speed is still a factor. Just because many of us are over-rotating on speed doesn’t mean we should remove it from the equation.

All right. Time to the gain: how soon do our customers see a payout from using the product here? Time to trust: how soon do they stop double-checking our work, or, as I put in parentheses, stop slop-rejecting from the AI slop cannons that we’ve created in some of our products here? And then one that’s really overlooked: time to lifetime value. The cost of acquiring a new customer is expensive. Our strategy should say, “Hey, how soon do we help that existing customer get to a point where they’ll buy again?” They’ll buy additionally. They’ll resubscribe here.

So again, with your success metrics, whatever you’re measuring, think about:

  • The time to getting the job to be done
  • The time to achieving the gains they’re seeking
  • The time to where they start trusting what’s coming out of the system
  • The time to where the customer is willing to put their money where their mouth is

Because with that trust, with reaching those outcomes here, we’ve become that trusted partner. So success metrics wind up that way here.

Health Metrics Across the 4 Ps

Dean Peters | 00:34:59 – 00:36:35
And then health metrics. All right. Are our people, our pricing, our positioning, our process getting in the way of our promises? So again:

  • People: who owns it, and how do they use it?
  • Pricing: the SaaS model is not really going to work all that well in a world where we’re getting charged per token. So what’s the pricing model going to look like? When’s it going to get offensive? When are people going to start to churn out because of that?
  • Positioning: how are we going to position why we win, or how we’re going to lose? So how do we position our product in the market there?
  • Process: the handoffs, the rework, the activation, customer enablement.

I’ll even say there’s change management that happens. I was thinking about Relay, the Relay app, and thinking, why would I want to go through all the change management of pulling out all my Zapier infrastructure for a product that’s pretty much practically the same? Maybe it gives me 10, 5% more of what I’m actually looking for. But if Zapier’s getting me 85 to 90% of what I want, is it really worth me pulling out what’s already working here? I don’t know. I think you probably have to answer those questions quite a bit when we come to you saying, “Oh, but we need to change what we’re using. We need this application here.” And you ask really realistic questions to that, right? I mean, you’re like, “Okay, that’s interesting, Dean, but tell me why you need this AI model to do whatever here.” And I think those are healthy conversations to have, right?

Cynthia Petti | 00:36:35 – 00:36:44
Yeah. So, here’s a question, though. You showed some really great things, but the question is, what metrics should a product leader report to the executive team, or to people that are curious about this data?

Dean Peters | 00:36:44 – 00:36:53
So I want to take that question and move it into the future a few slides, if you don’t mind.

Cynthia Petti | 00:36:53
Awesome.

Dean Peters | 00:36:53 – 00:37:09
Because I think we’re… hold on a second here. We’re one, two, three, four, we’re five slides away from answering that question. Excuse me, no, six slides. My bad. But we have an answer for that. We are going to get into specifics. All right.

Tactical Tools to Influence Strategy Without Authority

Getting the Right Stakeholders in the Room

Dean Peters | 00:37:09 – 00:39:02
In fact, let’s get into this, right? Because you’re saying, okay, as product managers, how do we engineer influence without authority? All right. And we’ve got the word “influence” in there a second time, because influence without authority is what product management is about. We don’t have power often, right? We don’t have the ability to hire, fire, and promote people here. What we do have is the ability to assert influence here. So it starts with getting the right people in the room. All right? It starts with getting the right people in the room.

So first of all, let’s leverage some conversational canvases. Remember, we talked about the different layers here. So how do we get alignment with the organization, with the team, with the product strategy there?

So, for example, when do you bring legal and compliance into the room? I can imagine some of you may be bringing an AI product to end of life, and you’re thinking your first stop is with the kind people in engineering and DevOps, and I’m here to tell you, absolutely not. Your first stop is at legal, and your second stop is at finance. Maybe in reverse order, depending on how your organization is run. Okay? So, those are things you have to think about there.

Security and compliance: we were working on a product matching patient records using AI, and I just kept hearing, “Well, you can’t do that. That’s illegal. You can’t do that. That’s illegal.” And so I met with legal, and I met with the security and compliance people, and we found that it was actually legal. We just had to make sure that in our backlogs, our product actually had things that ensure security and compliance.

And of course, we often leave out our product marketing people, and whoever’s involved with our go-to-market motions, until the very last minute, saying, “Hey, by the way, we’ve got a launch in a couple weeks. Maybe you want to get something out on the website.”

Cynthia Petti | 00:39:02
How many times has that happened, right?

Dean Peters | 00:39:05 – 00:39:16
Yeah. Right. So, we want to bring them in sooner because, oh, by the way, they have lots of insights. All right. So, I said, get the right people in the room with conversational canvases.

Storyboarding and Problem Framing Canvases

Dean Peters | 00:39:16 – 00:42:50
Let me share with you some unique conversational canvases that aren’t part of the basic stock of conversation pieces we’ve seen before. One of the things I like doing is storyboarding. We use this when we’re trying to avoid spend without… because we lack defining the value. Well, at an organization level, at a product level, maybe what we really need to do is paint a visual picture with the storyboard.

And you can do this with AI these days. You can actually take a very thin narrative, go into Microsoft Copilot, click on the Create button, and it’ll create for you a nice six-panel, what would you call it, storyboard. You can do the same thing in Gemini Notebook. You can go in there, give it some stories, and then actually have it create for you an explainer video. And it’s just getting people like product managers, subject matter experts, someone in UX and research or design, and getting someone from maybe ops in the room as well, to help you paint the story of the user, a day in the life with your product. The storyline here is:

  • Who has the problem
  • What problem they have
  • What’s the “oh crap” moment when that problem comes to a head
  • And then our product shows up, and they use the product here
  • And then they wind up becoming an evangelist for the product

A simple hero’s journey. But imagine now you can socialize this within your organization, saying, “Hey, here’s the narrative we’re going to pursue. Tell us where the baby is ugly.” You can start meetings. So, for example, some of you who are product owners and you’re about to start your sprint planning: you show the storyboard saying, “Hey, ladies and gentlemen, just a reminder, this is who we’re serving here with what we’re about to talk about. So everything we talk about needs to be in light of this outcome here.” So, a really good communications tool, and not expensive to do anymore. It’s just getting the right people in the room and having some fun with it.

Here’s something that possibly looks familiar but isn’t. What we’ve taken is a couple of famous canvases, the customer circle and the problem framing statement, and by the miracle of AI, we’re going to use some compression here. So, one of the things I teach in our class is, hey, for pains, gains, and jobs to be done, why don’t we have AI produce some sticky notes to put on the board? Not as sources of truth, but so when you bring people in the room, like the subject matter experts, someone from design, someone from marketing, you could say, “This is what a sticky note looks like here within this context. By the way, feel free to reject what AI created or modify it.”

But now think about it. Instead of a blank canvas, we actually have a canvas that has some starter stickies, as I like to call them. How much faster that conversation goes, where we can then go ahead and use AI to help us identify those top pains and gains, and finally draft a statement that we can start socializing with the organization, saying, “Listen, this is what we’re doing,” and this is at a team level and at a product level. So when someone comes in, the Rhino comes in and wants to inject something into our roadmap, you’re saying, “Okay, but how does it align with the problem that our user is up against here?”

Agent Strategy Canvas and Risk Metrics

Dean Peters | 00:42:50 – 00:45:44
Agent strategy canvas. I talked about agentic feral cats. So here’s a chance to define not only differentiation, but also get a little more governance in here. And you’ll notice how we do this. We talk about the agent, right? We get the right people in the room. But you’ll notice this is actually not done left to right. This is done in columns.

So first of all, we want to understand: have we understood the problem space? So the first thing I say is, “Do you know where the customer is?” Okay, go on to context. You can almost think of it like a game board, right? “Oh, congratulations. You understand the context. Let’s think about what’s causing the problem.” Now, and only now, do you go on to the solution space. All right? “Hey, what is the solution going to be? What are the safeguards? What are the steps?” And then the metrics: how do we monitor this? What are the mileposts?

So now, as the agent goes out, and you have several agents here, you’re not having political arguments or persuasive arguments over whose agent’s better for answering emails. You actually have something that has some governance around it. So when Caric, who wrote the agent, goes on to something else, and six months later you need to update it, you actually understand what it’s doing at these different levels here. And so this is actually for defining the strategy of your agent here.

And this gets into which metrics. So it’s not one metric here. It’s success and health metrics across what Marty Cagan calls the four risks:

  • Is this product desirable? And I don’t mean desirable like, “Oh, yeah, that’s nice.” It’s desirable like, “Yes, I’m willing to take money out of my budget and put it into yours. I’m willing to take money out of my bank account and put it into yours.” That type of desirable.
  • Is it usable? I’ve seen so many AI products. You look at it and go, “Okay, now what?” Everyone’s like, “Oh, Obsidian, Obsidian.” And I look at that interface going, “Oh my gosh. Yes, I get it, but my gosh, you need to be from Mars to figure it out.”
  • Health metrics here: a simple viability test. All right? And viability means, is this a good business model for us here? Is this a viable business for us? Is this something that’s going to move the needle on some business metrics, such as: Is it going to help people be more productive? Is it going to help save cost? Is it going to ensure safety and compliance? Is it going to increase our ARR? Is it going to expand our market share, or nourish that lifetime value?
  • And then feasibility. We’re talking about: can we build this, and can we do it at scale? “Yeah, I’ve got the engineers to do it, but I’ve only got two of them.” What happens when we go from 30 people to 30,000 people to 300,000 people here? So, is it feasible?

So, here are some of the metrics that you can put down here.

DAISI Governance Framework & Action Commitments

Dean Peters | 00:45:44 – 00:46:50
And then finally, who’s on first? And this pretty much can be done at all three levels. All right? But probably the organization and team level here. This is the DAISI diagram. Now, some of you may be familiar with the RACI diagram: who’s responsible? All right. I’ve never liked that, because that can get delegated and moved. The reason why I’ve liked DAISI is because that D stands for decision. It also stands for driver. Who is the person who drives the decision? Who is the person who owns the driving of the decision here? So, I like this a little better.

And then you can actually go ahead, and in this column, notice it’s not the job; it’s the workstream, the flow, and the decision area. So it’s like, “Oh, yeah, who’s pushing? Who’s driving this?” And so again, you can use a simple conversation piece like this. But all of these are getting the right people in the room. All of these are using these canvases not as some form of magic, but actually to frame a conversation.

Cynthia Petti | 00:46:50
Dean, can I throw them in the AI?

Dean Peters | 00:46:50 – 00:47:10
Absolutely. Go ahead, knock yourself out. Simulate a few. But just don’t use that as a surrogate or a replacement for the humans in the room having the collaboration, because remember, it’s not the AI who’s going to sit in a deposition. It’s not the AI who’s going to go to Club Fed because you broke the law with your AI.

Cynthia Petti | 00:47:10
So, AI is very agreeable sometimes.

Dean Peters | 00:47:13 – 00:47:54
Yeah, AI is way more agreeable. All right. So we’ve gotten through a lot of the content here, and I think now is the time to ask: starting today, with regards to AI strategy, what will you do differently? What will you do in your position to drive influence without authority? I’m curious if someone answers something else. I would love to know what the something else is.

Poll Question: Starting today, with regards to AI strategy, what will you do differently?

Cynthia Petti | 00:47:54
You’ve got some good answers coming in.

Dean Peters | 00:47:58 – 00:48:11
Yeah, I definitely want to know if there’s something else, if there’s one out there. All righty.

Cynthia Petti | 00:48:11
Looks like everyone’s committing to something.

Dean Peters | 00:48:13 – 00:48:59
Yep. I’m going to go ahead and end the poll. So, hurry up. Get your vote in if you’ve got one. And I’m going to share the results that we got. All right. So, redesign one workflow. I like that. “Look, I’m going to pick one workflow, and I’m going to redesign it.” Clarify the customer job. You can never go wrong with that. And when I’ve inherited new engineering teams, that was one of the first things I did: draw the customer circle, just to see what they put in there, to see where everyone’s head was at. Set an outcome metric here. Can’t go wrong with that either. So, absolutely awesome that people have some idea of what they’re going to do next, how they’re going to practice influence without authority.

Poll Results (as referenced during the session):

  • Redesign one workflow
  • Clarify the customer job
  • Set an outcome metric

Attendees committed to concrete next steps, led by redesigning a single workflow, clarifying the customer job, and setting an outcome metric.

Productside Offers and Q&A Session

Productside Resources, Training, and Events

Dean Peters | 00:48:59 – 00:49:05
And I think this comes to valuable offers. So I think we’re going to do Q&A, but first, valuable offers.

Cynthia Petti | 00:49:05 – 00:50:14
Yeah.

  • So we have this State of AI Product Management, which has a collection of data from the survey on where our community is with AI. It’s so important to see and know the data that’s out there, to get a guide of, “Okay, where are we at?” And I think it’s extremely important to know, given the topic of today’s webinar. So everyone can get even more data. There’s a QR code for you to download that.
  • And then we do have an upcoming training of Optimal Product Management, in person here in Austin, Texas, where I’m at, December 1st through 3rd. And that is with our very own Tom Evans. It’s a great opportunity for you to dive in and just immerse yourself in product management for those three days. Get some great exercises that are AI-enabled within the framework of product management, and have a great time just absorbing it all.

Okay. And then we have an upcoming webinar.

Dean Peters | 00:50:14
Here we go.

Cynthia Petti | 00:50:15 – 00:50:30
There it is. So, we have our next upcoming webinar, Wednesday, October 14th, and it’s Proto to Proven with Tom and Roger. We invite you all to register. It would be great to continue your learning with us.

Dean Peters | 00:50:30
I bet you Tom is also going to unleash a little Lovable action there as well.

Cynthia Petti | 00:50:34
Oh, I’m pretty sure.

Dean Peters | 00:50:35 – 00:50:44
I think I know what he’s going to do. Yeah, I’ve seen it, and it’s very cool. Awesome. All righty. And courses.

Cynthia Petti | 00:50:44 – 00:51:17
Yes. So, not only do we offer courses in person, but we are live online, virtual, and we have these two available:

OPM is appropriate for any level you are at in product management, because it’s such a great refresher, with governance and reminders of all things related to product management. And Dean, I’m going to have you give them a little snippet on AI Product Management.

Dean Peters | 00:51:17 – 00:52:19
Yeah. And if you like some of the conversational canvases that you’re seeing here, and the conversations and excitement around them, we are going to have these. Now, of course, we’re also going to get our hands dirty with some vibe coding. We’re going to get up close and personal with some agents. But along the way, we’re going to actually do it with in-class, hands-on activities. And the same thing with OPM, around these conversational canvases that help you collaborate around this, so you don’t release the agent that eats all your tokens and kicks over the trash cans and just leaves a big mess, or you don’t vibe-prototype something here that goes ahead and commits world destruction next week. So, we’re definitely going to enjoy having you there, because it’s always very active and hands-on. I think it’s time for questions.

Q&A: Core AI Usage and PM Intangibles

Cynthia Petti & Dean Peters | 00:52:19 – 00:56:09

Cynthia Petti:
I have some questions here, but they’re very general, if that’s all right.

Dean Peters:
Glad to answer them.

Question:
What do you think is the one thing that someone needs to be using AI for as a product manager today?

Dean Peters:
So I think you’re asking… go ahead and ask that question again here.

Cynthia Petti:
What’s the one thing that you need to be using AI for as a product manager?

Dean Peters:
See, for me, it’s never going to be down to one thing, right? It’s going to be about what conversations I need to have to understand what’s the right thing to build for my organization, for my customer, that creates that relationship where we sponsor the creation of great products that, in turn, people consume, that they find value in, so much so that they’re willing to go ahead and pay for us to provide it. And we create this cyclical relationship of providing value for people to get their jobs done, to remove pains, all right, to achieve certain gains. So I think the one thing is going to be the intangibles, not the tangibles. No one tool, no one model is going to do it all.

Cynthia Petti:
Yeah. And I think the important thing is to also know that we always have to be refreshing ourselves.

Dean Peters:
Yeah. I would agree with that as well. As much as I love AI, as long as I’ve been doing it, it’s like I said, and those of you who visited my skills library: you’re still offering something intangible there. So the one thing is, how much do you invest in you, to be that great judge, to be that person who makes great decisions and has great conversations? Go ahead.

Question:
And this question might be tough: what do you think is the level of AI knowledge that’s needed?

Dean Peters:
See, again, I see that as tactical. I think we all evolve and grow. All right. So I think it’s not so much what level here. All right. It’s: how am I getting better at asserting influence without authority? How am I getting better at understanding the right thing to build before we build it the right way? So again, I think we want to take a step back and maybe ask slightly different questions of ourselves. It’s not the one thing, not the one level we’re trying to ascend to here. All right? But what do we go ahead and do in terms of building into ourselves, and being able to build things for others that create value?

Cynthia Petti:
I think here, also, just be curious.

Dean Peters:
I think curious is one of those, again, intangibles. In other words, for example, let’s talk about AI strategy. We’re hiring someone for AI. I don’t really care if they can use Cursor versus Claude Code. All right. Does this person have the intangibles? Can I trust them? All right. Do they know where to begin? Do they know where to end? Do I have to micromanage them? Those are the things, regardless. I mean, yeah, we can always walk in with the skills. We can have skills as long as our arm and find ourselves out in the street a week later, because we have no judgment around them. We have no trust around them. We have no influence around them. So those intangible things will be the ones I would look at. Certainly, we can do that in the AI context, right? Certainly, we can do that, and I’m not against learning tools, but again, it’s not the tool that’s going to make the product manager successful or not.

Cynthia Petti:
Great.

Q&A: Managing AI Costs, Workflows, and Wrap-Up

Cynthia Petti & Dean Peters | 00:56:09 – 01:00:14

Cynthia Petti:
We’re still waiting to see if anyone else has other questions.

Dean Peters:
Right. I think we had a couple others from the room there. We got “which metrics” there, and others.

Question:
I thought you answered that, but which metrics to share?

Dean Peters:
Well, again, we had the metrics on the board there, but I think you want to look at layered metrics there, right? So, you want to look at it across those four different risk areas here, and just pick two or three from each of those areas:

  • Whether the product is valuable enough to the customer that they’ll buy it
  • Whether the product is usable, to get a person from activation to day-to-day adoption where they can’t live without it
  • Again, is it viable for us to do this as a business? AI is not cheap to implement, so we need to understand the cost there.
  • And then those metrics across: is it feasible? Do we have what we need as an organization here? This is what leads to other conversations: What do we need to bring in? Who do we need to bring in? What other companies might we need to partner with or acquire to make this happen?

Cynthia Petti:
So, right there, how you’re talking about what we need for cost and budget…

Dean Peters:
Yeah.

Question:
How do you work on your roadmap and consider the costs of AI?

Dean Peters:
Yeah. And I think part of that is maybe having a conversation with our friends in finance. This is not a new conversation. I remember working at a large company on their search, and getting a tap on the shoulder from someone in finance, like, “Hey, we’re burning this much money a month. What are we getting in return?” And I was trying to look at this and thinking, “Oh my gosh, what are we doing?” So I think part of that is just understanding what it costs to get something out the door here. Have some financial sense. Honestly, people in finance, I have always found, are more than glad to tell you how much you’re costing them, especially if you’re in product and you do it in the sense of, “Look, I’m trying to put this product out, but I want to make sure the money is coming this way, not going outwards.” So I want to make sure that I know enough about the unit costs of what I’m delivering.

Cynthia Petti:
Yeah. And finance people are really nice to talk to, as long as you have some good numbers to back what you’re asking for.

Question:
Sarah’s got a question about workflows before accelerating them.

Dean Peters:
Right. So one of the problems I’ve seen with workflows is people try to boil the darn ocean with the mother of all agents. I’ll share this screen with you. It’s someone who used n8n, and this person tried saying, “I want to create the comprehensive agent.” Wrong move. Okay. You really just want narrow, vertical-slice agents that probably focus on one thing and do it very well, and then are orchestrated by another agent or some sort of orchestration level there. So when things break down, you have more signal than noise. Your logs, everything, tell you where the problem is, where you can fix it, if you need to modify it. You don’t have to sit there and make a change here that’s going to domino and break things down there.

So it really is a type of work where you’re just saying, “Okay, I just want to grab the email I have, and then I’m going to hand it off to an agent that maybe does some interpretation certain ways,” right? Or maybe something that just correctly captures customer complaints, and then it does a handoff to some other agent that does that. So, the big thing is: don’t boil the ocean. Create a set of narrow vertical slices that are orchestrated either synchronously or asynchronously, however complex you want to get.

Cynthia Petti:
Yeah, that was a great question.

Dean Peters:
So, I think we’re at the top.

Cynthia Petti:
We are. We’re at the time. Thank you, everyone, for joining. We appreciate you. And feel free to sign up for the next webinar. Until then, take care.

Webinar Panelists

Dean Peters

Known for building AI products decades before they were cool. Learn from Dean’s obsession with product and passion for AI.
LinkedIn

Cynthia Petti

Productside COO Cynthia blends strategy, creativity, and heart—optimizing operations, inspiring teams, and sharing actionable insights that drive success.
LinkedIn

Webinar Q&A

An AI strategy for product management defines where AI should create measurable value across the product organization, product teams, and customer-facing products. It connects AI investment to business outcomes, economics, risk, governance, human judgment, and competitive advantage. A strategy is different from a collection of tools or pilots: it is a deliberate set of choices about where AI earns its place and where it does not. Dean Peters frames this as working on three levels. At the organization level, it covers how AI changes the operating model, decision rights, and coordination costs. At the team level, it covers how product managers use AI for research, analysis, discovery, and delivery. At the product level, it covers how AI shapes the experiences and value customers receive. Strong AI strategy for product management treats these levels as connected, because a capability that speeds up internal research may not translate into customer value, and a customer-facing AI feature may carry risk and assurance costs the team never planned for. The goal is to put AI where it improves outcomes rather than simply automating existing work, and to make those choices explicit enough that leaders can fund, measure, and revisit them.
Product leaders build an effective AI product strategy by identifying where AI creates measurable leverage, separating table-stakes capabilities from true competitive advantage, evaluating economics and risk, and then prioritizing initiatives across the organization, the team, and the customer-facing product. The process starts with outcomes rather than technology. Before approving an AI initiative, leaders should define what success looks like in business terms: faster learning cycles, better decision quality, lower coordination costs, productivity gains, revenue impact, cost reduction, or differentiated customer value. AI usage and adoption rates are not ROI; a team can use AI every day without improving a single business result. Next, each initiative should be tested against its full economics, including model costs, integration effort, governance overhead, and the cost of errors when AI outputs are wrong. Dean recommends treating risk and assurance as part of the ROI calculation rather than a compliance step added later. Finally, initiatives are ranked by expected value and confidence, so the AI product strategy concentrates investment where the evidence is strongest and keeps smaller, cheaper bets running where the upside is real but unproven.
AI adoption measures whether people and teams are using AI. AI strategy determines where and why AI should be used to create measurable business or customer value. The distinction matters because many organizations confuse activity with progress. Rolling out AI assistants, counting active users, or tracking prompts per week shows that adoption is happening, but it says nothing about whether the product organization is making better decisions, shipping more valuable products, or building an advantage competitors cannot copy. Dean Peters puts the core idea plainly: having AI tools is not the same as having an AI strategy. Adoption is necessary, since teams cannot benefit from capabilities they never use, but it is not sufficient. An AI strategy for product management adds the choices adoption leaves out: which problems deserve AI investment, which outcomes will be measured, which risks are acceptable, where human judgment must stay in the loop, and which capabilities the organization needs to build over time. Leaders who track adoption alone often find that AI spending grows while results stay flat. Leaders who start with strategy use adoption metrics as one signal among many, not as the definition of success.
The Now, Next, Watch framework organizes an AI product roadmap into three horizons: what the organization should institutionalize now, what it should experiment with next, and what it should keep watching. The Now horizon holds AI capabilities with proven value and manageable risk. These become standard practice, with clear ownership, governance, and training, so they stop being side experiments and start compounding. The Next horizon holds promising initiatives that need structured experimentation, with defined hypotheses, success criteria, and budgets small enough that failure is affordable. The Watch horizon tracks emerging capabilities that are not yet ready because the technology, economics, or risk profile does not justify investment, while keeping the team ready to move when conditions change. The framework pairs with the distinction between table stakes and competitive advantage. Widely available AI capabilities often belong in Now simply to keep pace, while advantage tends to come from Next initiatives that build proprietary data, learning loops, or customer experiences competitors cannot easily reproduce. Each horizon should be weighed against economics, customer value, risk, assurance, governance, and human judgment, which keeps the AI product strategy grounded rather than trend-driven heading into 2027.
Product leaders need to rethink human judgment, risk, and governance because AI changes who makes decisions, how fast decisions are made, and how errors spread. When AI drafts research summaries, prioritizes backlogs, or generates customer-facing content, a mistake can flow into a roadmap or a product release before anyone questions it. A credible AI strategy for product management defines where human judgment is required, where AI can act with oversight, and where AI should not be used at all. Governance becomes a product leadership responsibility rather than something handed off to legal or IT, because product leaders are the ones deciding which AI capabilities reach customers and how much risk the business accepts. Dean Peters treats assurance as part of the value equation: an AI feature customers cannot trust is a liability, not an advantage. This shift also changes the career path for product leaders. The people who advance will be those who can evaluate AI economics, set measurable outcomes, design governance that does not stall delivery, and explain AI tradeoffs to executives. Building those skills deliberately, through practice and structured AI product management training, is what separates leaders who direct AI strategy from those who react to it.