EP 155 - Reshaping product development for AI’s impact with Gil Broza

Matt Green: M. Hello, product people. Welcome to the Product for Product podcast hosted by Matt Green, data advocate and product manager, and Musha Mikanovsky, product leader and author. Our goal is to serve the product community by helping you find products that can help make your work in product management easier. Thanks for joining us on another episode of the Product for Product podcast.

Moshe Mikanovsky: Welcome back, everyone. On today's episode, Mache and I are happy to welcome back author, founder and coach Gil Broza to speak about reshaping product development for AI's impact. Welcome back. Let's. Let's hop in. How you doing, Gil?

Gil Broza: Uh, I'm doing great. Thank you for having me again.

Moshe Mikanovsky: It's, uh, a pleasure. Hey, M. Hey, Matt. Hey, Gil.

Speaker D: Welcome back. Uh, all the way from Toronto. Uh, you're actually in Toronto, where I'm just in the suburbs. I'm just, uh, actually lying to everyone when I'm telling that I'm from Toronto. I'm from a suburb of Toronto.

Moshe Mikanovsky: So, yes, this is a sneaky thing about Atlanta too. So nobody that knows Atlanta. I just say I'm from Atlanta, but really, I'm on the burbs side.

Gil Broza: Exactly.

Speaker D: Uh, amazing. Uh, so welcome back to the show. We're really excited to have you back. Uh, where have you been and what did you do since the last time we met with you?

Gil Broza: Well, I've definitely been diving into AI, surprise, surprise, like nobody else has. Right? Yeah. It is amazing to see what AI is able to. Able to do. And it is, um, somewhat disheartening to see what corporations are doing about it. And so that's what I'm dealing with. Yeah.

Speaker D: But before that, just remind our audience who you are and what's the core passion of your work is.

Gil Broza: Okay, so, Gil Broza, independent For the past 17 years, my thing is to help product leaders reliably deliver meaningful impact. And nowadays in the age of AI, which complicates things and opens up new avenues. Uh, I've worked with over 100 clients so far. Of course, not all 100 with AI. Right. Uh, so for ever more than two decades, it was very much about agility. And nowadays, uh, it's more about, you know, incorporating AI. But in many cases, my clients are already interested in agility. And it's a matter of, you know, how do we bring, uh, AI into that? So we stay still have agility, but we need to kind of shift from what we had before.

Moshe Mikanovsky: Yeah, great. I've actually been thinking quite a bit about that. The agility and agile processes and all those things that we've taken through the years as a product, as product managers, like that itself is changing with AI. Like do you, do you do all the ritualistic, uh, scrum ceremonies and things like that when you're prototyping so quickly and stuff?

Gil Broza: Yeah, right.

Speaker D: Before we get into that, maybe. Sorry. Another comment that I wanted to share is that agility in some cases, and some teams that I worked with, and some people on those teams felt that it was a fad, it was something that, uh, didn't really work because of the implementation that they had and stuff like that. And I think some people have the same, um, thoughts or feelings about AI, but they're two very different things. Oh yeah, because agility is really a mindset and a process where, uh, AI, it's tools and for some reason people kind of conflate them together or replace them. Like you said, people are interested in AI and they may be interested also in Agile, but there are two different things. The reason that you're doing AI doesn't mean that you don't want to do Agile, and the reason you're doing Agile doesn't mean you don't want to do AI. It's just.

Gil Broza: Oh, absolutely. I mean, look, everybody is incorporating AI no matter what they do outside of software development, of course, and depending on the industry, it will look like this or it will look like that.

Speaker D: Mhm.

Gil Broza: But the thing is that it's not just the shiny new thing.

Speaker D: Mhm.

Gil Broza: So for a while Agile was the shiny new thing. Then it was okrs. JIRA became this amazing thing. Hey, let's use jira. Because everybody seems to be. We don't need the rest of the. The rest is 99%, but we don't need the rest. And now AI. But AI is borderline magical, right? Because it touches everything. We do need to revisit how we work.

Speaker D: Yes, absolutely. Um, and actually, uh, Matt and I are thinking about serious. Matt just came up with this brilliant idea to have this series because we're talking about in the last episodes, few episodes, it came up that there are first principles of product management.

Gil Broza: Mhm.

Speaker D: And then all the tools that you're using that still need to support those first principles, you cannot really work without them. People are again replacing the first principles with those tools without thinking about the first principles.

Moshe Mikanovsky: Mhm.

Speaker D: In the same way, um, how you develop an agility and all of that stuff, it still could be considered a first principle in a way.

Gil Broza: Yeah. When I teach, uh, my courses, when I work with clients, one of the first things I want to understand is that the Fundamentals haven't changed. AI increases leverage, but it does not change the fundamentals of product development. You still need a solid product direction. You need to actually build a product, not just write code. Uh, it doesn't totally remove the need for having learning a judgment. Right. Like research, feedback, data, evidence based decision making. The technical foundation is absolutely necessary. People and culture are still a thing. Governance is still a thing. So the fundamentals haven't changed.

Speaker D: Stakeholders are still a thing.

Gil Broza: Yeah.

Speaker D: Unless you could replace them with AI. Well, we'll see what they will have to say about that.

Moshe Mikanovsky: Agents.

Gil Broza: Yeah.

Speaker D: Okay, cool. So maybe we should start diving into this. Um, so our theme today is reshaping product development for AI impact. And we're having Agile in mind, of course, uh, as product development. Was there a reason why when we discussed the topic today you didn't want use the word agile in the topic?

Gil Broza: Because again, it's not about that. Last time we talked about real agility versus the cargo cult version. Um, today's conversation is about what happens when you bring AI into either of those. Okay. Um, so whether we have listeners work in companies where agility is for real in terms of the adaptation, the collaboration, the minimizing time to value all of those good things, or they're experiencing agile in a totally performative way. Ceremonies, air quotes. Right. Uh, the stand ups, the stories, the practices, the jira, the tickets, all of that stuff. Basically, you know, process without the mindset backing it up. Everybody's going to benefit from this.

Speaker D: Mhm.

Gil Broza: But the thing is, once we use the word agile, that's always been true. People understand it differently. It's like so far we've been saying both the word agile and we've said agility and they do mean different things, um, to a lot of people. Still, 30 years in agile equals scrum. Right. But that's not what we're going after.

Speaker D: Yes.

Gil Broza: Okay. What we're going after is literally the ability to move with agility. Meaning an opportunity arises or a competitor eats your lunch or whatever it is you need to respond. Yes. You're not going to respond by having 15 minute stand ups of three questions. Right? Right. So you do need this type of agility. It's in a way, it's almost non negotiable these days because everything moves so fast. Yeah, that's just the speed side of things. The other side is how do you make good calls?

Moshe Mikanovsky: Mhm.

Gil Broza: Which is super critical in product. Right. It's critical in a different way in engineering. But you know, in product, the product will live or die by that.

Speaker D: Yeah.

Gil Broza: How do you make quick calls when everything moves so fast? When you have inherent delays in the system that mean that decisions are not going to come too quickly? Mhm.

Matt Green: Ah.

Gil Broza: How do you not just keep the human in the loop, but actually rely on the same solid human judgment that brought you thus far and kept you in business without abdicating to the AI? So all of these questions are still relevant.

Moshe Mikanovsky: Yeah, that's an interesting balance. So like with AI, we're moving faster, we're innovating faster, we're making decisions faster. But in, in the old, maybe in the old way without AI. The old way as, like three years ago, uh, you know, we had more time to be more intentional, like.

Gil Broza: Exactly.

Moshe Mikanovsky: So I think that's a, that's a balancing act now. I think, I think with AI, you can certainly drift. People can go off and drift and do things, you know, unintentionally, intentionally. But yeah, so it's, it's an interesting balance that I'm seeing in the market.

Gil Broza: Look, the drift is real, right? Let's say you build something super quick and it has, or it contains all sorts of little problems and bad assumptions and stuff like that. Whether you wrote, you know, a product design or you wrote code, or you wrote a test plan, something about it is a little bit off. You keep iterating and building on that, it can drift farther out unless you have mechanisms that prevent that. And Matt, you're right. I mean, it used to be that things operated at a human pace. So we would have conversations and meetings and workshops and we would revisit things and all of that. But now we have this practical magic wand which speaks to us super compellingly and never admits to making mistakes, that just apologizes profusely, which is stupid. Right? And, but, but you have to catch it, making the mistakes and you know, we can tell it well, find what's wrong in what you said there. But even that doesn't work properly. Maybe it will in a few years, but right now it doesn't. Not consistently.

Speaker D: Yeah, yeah, yeah, absolutely.

Moshe Mikanovsky: Yeah. That's something I try to do. I try to interrogate it. So when I come up with prototypes and things like that, I'm, I'm saying like, I'm using it as a design partner. So I'm saying like, I had an idea. What do you think? Give me your best thoughts, but confirm before proceeding. That's always like a thing I run to is like, don't make, don't take an action until I review your plan. Yeah, Anthropic's gotten really good about that. You see, like, uh, in a lot of the things they do now, they. They actually present a plan before it proceeds. So, you know, I think that's certainly improving. But yeah, if you just hit like approve, approve, approve all the time, well, it's going to do whatever it wants.

Gil Broza: But it does make us lazy, doesn't it? Yeah, I mean, that's one of the

Speaker D: challenges I see that happens to me is that it creates a lot. It creates very long responses, and then you have to go and read them to see that it's correct and it's not mistakes. And then over time you're like, okay, I can trust it or not. And if I start trusting it, then I kind of not reading everything and then it may not. I become lazy. And then, um. But it's only because the responses are so long sometimes that I'm like, I need shorter version of that, or I need just the things that I'm looking for, not everything around. Because it can go into rabbit holes, especially with the question that it asks you at the end where it asks you, do you want to also me to do this or this or that? And I'm like, oh, yeah, I want to do this and I wanted to do this. And I like everything that you suggest me.

Gil Broza: Yeah. And here's the thing some people will tell you, well, you need to prompt it differently, you need to configure it differently, you need to tell it what the boundaries are and whatever. Let's say you've done that. Well, let's say you've been able to configure it so it consistently behaves like a rational human being who knows your domain and works with you. And anything you just said, it immediately comes back with a thoughtful reply that is exhausting in ways we can't even describe.

Speaker D: Exactly. Exactly.

Gil Broza: So that creates another problem because, um, I mean, one thing we were going to talk about was like, what are some of the things that organizations miss? I think organizations are totally missing the effect on the humans.

Speaker D: And that was my next question to you. So let me phrase it. And that's a great segue to our question. Um, you say organizations, but let's talk about product leaders. Uh, they are representing the organization for us. What do you think? Or what are you. What are they missing when it comes to AI adoption and why is it important?

Gil Broza: Okay, so first let's clarify, because when we're talking AI adoption, we're talking two things. One, which we're not talking about today, is just building AI Capabilities. Like I have an agent that answers customers. We're not talking about that. We're going to talk about enabling the whole product development life cycle with AI. So the first thing I would say is that they're equating more output faster with success. And that's a huge blind spot. Okay. Output is going up, but it doesn't mean that value is going up. Okay. Producing specs, codes, tests faster doesn't necessarily minimize time to value.

Speaker D: They're still project oriented. They're still thinking in project, not in product.

Gil Broza: Well, they're thinking output. So it's not so much the managing of the thing, it's the we just need stuff, we just need features. Yeah, yeah. It's related. Right. And the thing is that, you know, the faster code generation today often becomes tomorrow's rework and bugs and tech debt and whatever, even though the models are good and they can handle some of it and so on. But what we're doing is we're producing legacy code years ahead of schedule. So that's one the other thing that almost everybody seems to miss, and that goes to systems thinking, which is not a common thing in our space. AI speeds are part of the system, but the system was never designed for that speed. M It was designed for human speed. Okay. Now if you think about product work, it stretches all the way from idea to delivery and there's a bunch of constraints and feedback loops, right? So you accelerate one part, the effects ripple into other people's work, into feedback loops that reinforce problems and into consequences that show up much later. Right. So it's like you put a turbo engine, uh, in a car with the original brakes and suspension. Good luck making the first turn.

Moshe Mikanovsky: M. Right, yeah. Yeah.

Gil Broza: Mhm.

Speaker D: So that's a great way to think about it. Absolutely. I was scared in that car.

Gil Broza: So if leaders don't account for the system, the system will push back on some changes. Right. Those are balancing loops. And it will also form vicious cycles from other changes and you can end up with some serious problems. And you just need to read the news every week. There's something else going on. Okay. Uh, a third thing I would say that leaders are missing is that AI adoption is not an implementation, it's a transformation. Right. So earlier you said, you know, it's really tools. It is really tools, but there are some pretty wicked tools. Okay. We're not talking about, you know, build script. Okay. We're talking about something totally transformational. So what do companies do? They add the tools, they adjust a few practices in product, in engineering, deployment, Everywhere but they keep the system mostly as is.

Moshe Mikanovsky: Okay.

Gil Broza: Mhm. But AI touches everything. Agility was the same way. And so what happens is it enables effects that we couldn't get before and really the only way to seize those effects is to approach value delivery with a different mindset. You also touched on this earlier which is really the strategy for doing your work. You're not going to get the same quality of work done. Um, if your strategy for Getting work is two week sprints and 15 minute standups. You need something more fundamental than that. And that's where mindset comes in. Values, beliefs, principles, the prime differentiator between successful and unsuccessful Agile implementations. Here again. But the mindset we need now is not the base model for agility. We need something a bit different because now we have AI team members.

Speaker D: Yes, that's true.

Gil Broza: Yeah, that's a good point.

Speaker D: We will get into that because I'm really curious to hear that. How do you deal with that? And, and uh, we'll talk about your course that uh, you developed and you're providing uh, people because you know this is uh, about uh, actionable tools to product leaders so they can actually learn, you know, not just from listening to us but also from uh, you know, taking your course. But um, what I'm um, the, the other thought that I had when you were talking about that was how companies that never realize what agility is all about now with AI have even a bigger problem that they ever had before because they're trying to adjust to those uh, uh, performance agile. Um, and it definitely doesn't work in the same speed because you don't expect the agent to come to your stand up and tell you this is the three, the what I did yesterday, what I'm doing today and where I'm stuck.

Gil Broza: I bet, I bet that there are companies that do exactly that, that when they run their stand up. And you know nowadays most stand ups are virtual and people don't actually stand up but they're still there to talk about the progress in the plan. You can have the AI basically say here's what I moved. Mhm. Okay. What do you get then? You get something that did a task in 10 minutes. Kind of sit there until your 24 hour cadence rolls around to update you about that. That's not a net gain.

Speaker D: Yeah, yeah, that's uh, not utilizing uh, the tools. But I really appreciate that you put it in a speed way and when thinking about the systems and us as human, not designed to work in the same speed as those agents are, are designed to work. So maybe I'm jumping ahead of this, but I'm really curious to hear, is there a way to kind of slow down the agents or speed up us or. That's not the right question to ask. Maybe there is a different way to phrase it.

Gil Broza: I think it's not so much a different way to phrase it, but I think what we need is a different way to think about it. Because here's the thing. We will never catch up to the agents. Never. Right? I mean, humans cannot catch up to machines, period. Right.

Speaker D: We develop them.

Gil Broza: M. Well, but not only that, I mean, we know that they're fast. Right? Um, look, we have security incidents all the time, right? You just read this in the news every week. Okay. You will never have a security team of humans that can catch up to, you know, breaches and such when they're done by AI. Right? So the whole Mythos thing is basically part of that. But it's going to be the same thing when we're just talking about times of peace. We're rolling along, developing a product and we have something that makes changes in 10 minutes that a human would need an hour to review. So what do you do? Mhm. What I want people to do is to actually think about the system they're in and understand several things about it. One is, what are they actually optimizing for? Okay, historically we optimize product development for predictability. Right? That's the waterfall game. Predictability, get it right the first time, standardize, solve everything the same way. Okay. We realize that that's probably not great because we can't really get it right the first time.

Speaker D: Mhm.

Gil Broza: Okay. Nowadays we. We actually. No, I'm jumping the gun on this one. Wait, okay, so that was the waterfall game. Then we changed what we optimized for when agility came along and said, no, no, no, we're optimizing for doing more of the right and less of the wrong. How are we going to do that? By talking to each other. By validating with our customers. Rapid feedback, uh, gradually delivering valuable pieces and all of that stuff. Is this what we're actually optimizing for in a given company? If not, should we? Most organizations don't have that conversation.

Moshe Mikanovsky: Mhm.

Gil Broza: Instead, they're in the business of producing stuff and they optimize usually for schedules and for delivering lots of stuff and for hopefully coming in earlier again. Project management. Yeah. So that is a totally different target of optimization. It's really what I call values. Right. Um, it's a target of optimization that is different from saying, we are optimizing for the best experience, we are optimizing for the highest referral rate. We are optimizing for delight. We are optimizing for uniqueness and innovation. The only difference from saying, no, we optimize for producing a shitload of stuff.

Moshe Mikanovsky: Yeah, yeah, I like the way you frame that because we're using Claude throughout, from product down into engineering at this point. So, you know, it's optimizing the way we work. And personally the way I work with engineers now has changed. So, you know, within a two week sprint there's questions that come up about tickets and it's like, hey, let's hop on a call, prototype something and work together. It's way more collaborative now. And then they can go off and implement them. So I think that's an interesting way to frame it from an uh, optimization. Like what are we trying to achieve here, uh, with all this?

Gil Broza: So when you have that conversation, we sometimes want to uh, distinguish two things. What are we trying to achieve can be just the product vision. That's one. The other thing is what are we optimizing our way of working to achieve? So we can optimize our way of working to achieve innovation. We can optimize our way of working to achieve predictability. We can optimize our way of working to achieve on time and on budget, all for the same product vision. It's our choice. Well, it's senior leadership's choice. What you want to do is after you have determined what you're optimizing for and all your leaders and the teams are aligned towards that because everybody will have different preferences. But we want to agree to operate with the same alignment. Now you can look at your system, how it's set up and say, okay, how does it actually achieve that?

Moshe Mikanovsky: Mhm. Mhm.

Gil Broza: Does it achieve that by involving enough human brains at the right time, in the right place? Does it achieve that by having a whole bunch of specialists hand work off to each other? Does it optimize that by deferring 90% of decision making to the machine? And you can say, well, okay, is that the way we have it? Is it fit for purpose? And now you want to improve that. So it's not about going backwards and saying, look, we like the idea of sprints. How can we have AI in sprints? Oh, we can invite the agent to the standup. Oh, we can have the AI write tickets for us. We can have the AI, um, I don't know, summarize Slack channels and whatever.

Moshe Mikanovsky: Mhm.

Gil Broza: Does it materially Change what you're optimizing for. Usually not. And sometimes it actually moves you backward. For instance, I have this, uh, I started working with a client and one of the things they wanted to use AI for, like, they actually have an okr around this for like six months from now of it will do all the ticket writing. Okay. And I'm thinking, well, but where does it add value? So let's say you're a product manager and you give your ramblings to the AI and it produces this pristine ticket. If all it does is organize your thoughts. Beautiful. Saves you tedium. Totally fine. If what it does is come up with what it thinks the requirements should be, you're now at the mercy of guesswork effectively or copying what it got trained on. Do you want that? Not sure you do. Now, if you use it instead as a thinking partner, to say, all right, I need more insights about the market. I have access to just a few people, uh, and they're busy and whatever. The machine will help me think. Okay, that does seem to help you move closer to what you're trying to accomplish.

Moshe Mikanovsky: Going back to a point you mentioned earlier about system thinking as well, the way, the way this is, the way the systems have worked historically have been defined, and everybody knows it really well. I was talking with someone the other day. They, they asked about. They mentioned Scrum Masters, and I was like, a Scrum Master today would be. I don't even know if that's a role anymore, really. And then that comes into the roles and responsibilities within the system that you're working in. I've talked to PMs too, and I've seen it on LinkedIn. Where do PMs need to start doing pull requests, uh, and things like that. I know Michae and I have talked about it as well, and I've talked to engineers, and engineers, you know, are. They need to be more product focused and like, what am I delivering to the business? So, you know, I think it's a bigger thing. And I think if the organization's not thinking about that, like you said, it's a blind spot for them.

Gil Broza: Yes. And I think what's happening right now is, you know, it's the same human behavior when new things come up. Okay, um, we have an idea. Let's hammer it into place until we can make it work. It's like, okay, so 30 years ago, scrum said we need. Scrum Masters gave a really good, legitimate, uh, value statement for the role. Most organizations did not get the memo on that one, but they still try make it work. Make it work. Based on their previous understanding, previous mindset, which effectively equated the Scrum Master with a project administrator. Okay, so it's not about, you know, do we need to give the Scrum Master more responsibilities? I think you want to take a step back and say, all right, what do we need?

Moshe Mikanovsky: Mhm.

Gil Broza: And something that we needed. And you know, to be fair, it was a blind spot. Something we needed was we needed somebody to help teams succeed and thrive together as teams, not just as individuals. So now you can ask, well, okay, who's going to do that? Well, it could be a tech lead, could be an engineering manager, it could be a coach, it could be anybody, really. But you do need to give a home to the role of, uh, sorry. To the responsibility of helping the team succeed as a team. So again, Scrum suggested that the memo got lost and all of that. And now what we're getting is because of the, uh, tools, all of a sudden we're saying, well, engineers become product people, and product people, they do vibe coding and this, that and the other. And, and I think that's looking at the wrong question. The question is, what do we need? I don't technically need my product manager to vibe code. What I probably do need is faster experimentation. Mhm. How do we get faster experimentation? Is it by giving somebody who's great at reading the market a tool that they can play with M. Or is it by, um, attaching an engineer to them so they pair up? What's wrong with that?

Moshe Mikanovsky: Yeah, that's experimentation. Point, like 5 code something. And then you get in front of a customer very quickly and get really quick feedback, uh, getting feedback and then going back to the team, going back to the engineer, getting them on a call and doing a little more workshop around it. And hey, I learned this. Let's make adjustments.

Gil Broza: So, so it sounds like you're doing this voluntarily and. And you're good at it. So good for you. Keep going. What I'm not cool about is when organizations say, no PMs, you need to figure out this vibe coding thing or you're out. Engineers, you got to figure out the AI thing or you're out.

Moshe Mikanovsky: You do see that? I've seen a lot of chatter around that. Yeah.

Gil Broza: Yes. And again, the thing is, it goes after the wrong problem.

Speaker D: Yeah, yeah. And it comes from some of the best companies, unfortunately.

Gil Broza: So I've been talking to a client where somehow they, uh, decided to start using something called AI. Dlc. You heard about it?

Speaker D: Aitlc.

Gil Broza: Dlc.

Speaker D: Dlc.

Gil Broza: Yeah. Like Delivery lifecycle.

Speaker D: Okay.

Gil Broza: Development lifecycle.

Moshe Mikanovsky: I haven't seen that yet.

Gil Broza: Well, good. So here's the thing. And that actually also brings us back to Real Agile or performative Agile. Okay, so what happened there is that this is a company that brought Agile in, uh, three, four years ago. Kind of late to the game, but fine, okay. And I spoke with them and they like it. And when I asked, okay, what do you like about it? I said, oh, there's the ceremonies. And like, okay, fine, I'm not wild in ceremonies to begin with. Definitely not a word. But I think what they really like is predictability. So when somebody showed them this AI DLC which came from Amazon, they really took a shine to it. When you look at this AI dlc, basically it's waterfall. With AI, it's great for predictability. So it's like, it walks like waterfall and quacks like waterfall, only you use AI to optimize each stage.

Speaker D: It's interesting because I was thinking exactly about that before that when you said about predictability.

Gil Broza: Uh-huh.

Speaker D: That's what AI is for. Predictive, uh, AI is a predictive system. It allows you to predict things based on data. M. So it kind of makes a lot of sense that AI taking you back instead of forward to the days that managers and executives wanted things to be predictive, that you have to develop the right things the first time. We don't want to waste time on developing, uh, anything, but they don't even give you the time to experiment on anything. So they tell you, develop this. This is the right thing. Just develop that. So they predict the future. And I think that they are probably in love with AI because AI will tell them all of this is wrong and this is right and you have to do this.

Gil Broza: And it's very convincing. And it's like the way it speaks to us is just, you know, the most compelling. But so this actually also brings up another interesting point. Okay, I totally get why everybody wants predictability. Heck, I would want predictability. Why not? M. Why aren't we able to get it? Is it a tooling question? Is it an algorithm question? Is it a model question? No, it's not. It's because the world is a super complex system and our correctness doesn't depend 100% on our brains. It depends on what other people will do and how the market will react to things. And product market fit is not something you can predict. Right. We can make some really informed guesses and we have learned that we should probably test them out.

Moshe Mikanovsky: Mhm.

Gil Broza: Right. And we should probably iterate because it's not the problem of, you know, thinking ability or brain power. It's because the world is inherently unpredictable. Yes. And, you know, I just read the other day about, you know, how, um, there are some people who are called super forecasters. They seem able to make predictions much better than others. Okay, fine. Profits. Interesting. Yeah, so they're pretty rare. But they do seem to make better predictions than even the, uh, wisdom of crowds. Okay. And now they're trying to build AI models to be even better than those super forecasters. But the fact remains that when you put a product out there into the world, there are so many factors that are going to affect its success and fit and desirability, the whole ilities, right. With the viability, feasibility, whatever. So many things are going to affect that that you have no way of predicting.

Moshe Mikanovsky: We're all just prediction machines in the end, we're just using the best information we have available to make a prediction about whether something's going to be successful when it comes to a product launch. These are the variables. We know what we know and how we implement it.

Gil Broza: Uh, and how often are we wrong? And how often are we wrong despite having good data? So if you say, okay, maybe things are not inherently predictable. So should I have a development lifecycle that pretends as if things are in fact predictable? Again, that was waterfall, and now it's the, say, dlc. Or do I build mechanisms into the workflow so that we can respond gracefully and quickly and economically when things go wrong? That's the basic case for agility. It manifests differently these days because things move fast and whatnot. It manifests differently, for instance, in that because of AI skipping, prototyping and skipping, experimentation is really not excusable anymore. Okay, if it use. If, you know, in 2022, it took a month to come up with a testable prototype to test things, and you could legitimately say, you know what, we're going to take the risk, we're going to skip it, because we're fairly convinced this and that. Okay, fine. Nowadays, when it takes you a minute. Well, not a minute, uh, let's say a day. Okay, whatever. Peanuts. Fast. Yeah, it's really far less excusable to do that. Now, systems thinking, let's say you've shortened the creation of the experiment from a month to a day. How long will it take you to get useful, actionable feedback? If it's still the same as before, which is, I don't know, a couple of weeks. Um, you're still faster, but you're not. It's not a day versus a month. It's a day plus two weeks as opposed to a month plus two weeks.

Moshe Mikanovsky: It's a good way to look at it.

Gil Broza: Yeah. And so you need to account for that and takes you two weeks. What do you do during those two weeks? Do people sit on their thumbs? Do they try another experiment? These are decisions people have to make and nobody's making them because they're in the business of just doing more faster.

Speaker D: Mhm.

Gil Broza: And what we also know from systems thinking and all of Kanban is based on that is, is that the more you push into the system, the more the WIP grows, things will back up and just wait and then the workload will kind of slam you. And that's exactly what's happening. Like I said earlier, it's also having an adverse effect on people and that is totally unmeasured. The fact that people work from 9:00am uh, with the agents and by 11:00 clock they're wiped because it is whole. The cognitive load is intense and there is no rest. Who's accounting for that? It's not on any P and L. Mhm. And. And it's not cool to say, hey, you know, I actually need some time here because you know, velocity people, we gotta move.

Moshe Mikanovsky: Yeah. I've recently seen this with a company, I won't mention, mention them, but they're in the financial service industry. But yeah, they actually like benchmark. It seems like they're pushing out a lot of innovation in the financial services, really cool stuff in AI but they have like benchmarks now to where it's, it's almost a competitive. They actually mentioned it's, it's a competitive environment now where teams are challenging themselves to like compete against each other for what they like push out the door. And I'm like, that must be like a real pressure cooker if you're not doing enough as your other team, your peers are doing more. That's creating strife inside the company potentially.

Gil Broza: Some people make the argument that nothing is really new in this whole thing. We've always had evolution, revolution, everything got faster, right? I mean software development in the 70s, you could still have competition between teams, only they didn't make a whole lot. And nowadays they can make a whole lot. I think this is in fact different what we have now because it looks to be on par with our human ability. That was never the case before. What we had before was better tools. But this still mattered, right? What went on inside the skull still mattered. And nowadays it seems like well maybe it doesn't need to matter as much. And that has effects on self worth and self importance of course, in the positive way. And also what's next for me, you have people walking around with fear for their jobs and the fear is real and the consequences are real.

Moshe Mikanovsky: I've only done five AI things and the other team has done eight AI things. I'm not hitting the numbers so. But they aren't like external value to the customer. They're like internal projects almost.

Speaker D: I had an interesting argument a year ago with the developer that uh, he, I argued that his job will be eliminated first and he argued that my job will be eliminated first, the product management. And um, I completely understood his argument only because he didn't understand what product manager does does. So from his perspective was like product managers all he does is give him the things to do. He can get that from the AI. He doesn't need a person to tell

Gil Broza: him what you just write tickets, right? Yeah. Okay.

Speaker D: Good luck with that.

Gil Broza: Yeah, of course. Right.

Speaker D: The said of the reality is that uh, that's what most people think and that's why product commitment has been the kicking ball for everyone. But uh, I want to maybe uh, we talked about a lot of things there and maybe I want to just a few suggestions what leaders can do. And here I'm using the word leaders where before I use the word manager and executives. Because not every manager and executive is a leader and not every leader is a manager and executive. So I'm talking about leaders, people that understand this, understand the uh, conundrum that some of companies are in. What can they do to make things better.

Gil Broza: Yeah. So I think it's what I can just kind of give a quick summary of the things I mentioned along the way. So start with systems thinking. Basically understand your value delivery system. Don't just think process, think people, how they get work done, who's inside of it, where the bottlenecks are, what the feedback loops are doing and also where AI speed gains are actually being absorbed as opposed to creating new pressure elsewhere. So understand how the whole thing works. Second, revisit the mindset explicitly again. These are the values, beliefs and principles that shape how the organization makes decisions. Uh, AI really demands us to rethink what we optimize for and what we assume. And this is a lot harder than updating a process or installing tools. Uh, third, actually design the way of working for AI speed conditions. Right. Don't just adopt it into your current process. This is not about enabling your two week sprints. Maybe you need flow instead of sprints. This isn't about writing pristine tickets, because maybe what you need is not quite a backlog and so on. Uh, definitely pay attention to the human side, to engagement, skill development or atrophy. We haven't talked about that, but it is totally real. Um, and also what it actually feels like to work in the organization. Okay. Um, Rising Velocity can trick you into thinking that everything is all right. So here's just one example. We haven't talked much about the people side of things, but I'll just give an example. Now, a lot of leaders are thinking we can have smaller teams because with AI, they can do more. We don't need big teams. You know, there's a coordination tax and all of that. But if you make teams much smaller, people will be even lonelier than they already are because of this hybrid remote business. Okay, but if you're not actively looking, you wouldn't. You won't see until the problem becomes super expensive.

Speaker D: And then they will hire AIs that will behave like humans to be their friends.

Gil Broza: Yeah, sorry.

Speaker D: Sorry for the negativity over there, but you know that this is coming.

Gil Broza: Yeah. Uh, what else leaders can do? So really be strategic and explicit where and how you use AI, because there's really three modes. It can be a pairing partner, like a thinking partner, a coding partner, or whatnot. Um, it can be an autonomous agent doing stuff, basic gen AI. Or it can simply be an admin or an automation component. Like, I heard about this company where the leaders are always invited into multiple meetings. They send their AI to take notes and tell them what the action items were. Okay, that's just. Automation doesn't actually buy you anything more than that. But these three working modes, you want to be explicit about where we use, uh, each one. And the last thing is, uh, don't really wait for a complete plan for rolling out AI. The landscape is moving too fast. So really get your leadership team to a shared mental model and then just move gradually and iteratively like any other

Speaker D: tool that they might implement.

Gil Broza: Yes, yes. I mean, uh, I'm really into change being gradual, but you still need to be strategic about it. Not just, uh, like what a lot of companies are doing. And I hear this from people reaching out. To me, it's like we've spent the last year, we bought the licenses and we just told people, hey, see what you can make with it. That's m. Like, the least strategic you can think of it. No, because what also happens is that some people will love it. Some people Will hate it. Their opinions will become kind of baked solid. And not only that, those who do use it, some will use it poorly, some will use it well, it will create pressures in the system. So your starting point of, you know, hey, try this, see how it works. All of a sudden creates, let's say an overload in pull requests. And you never saw this coming.

Matt Green: Mhm.

Gil Broza: So be strategic.

Speaker D: Absolutely.

Moshe Mikanovsky: Yeah.

Gil Broza: Great.

Speaker D: Uh, so we mentioned your courses uh, earlier. Tell us a bit about them and uh, who they're for, uh, what do you teach them and how do they help the participants.

Gil Broza: Yeah. So, um, I have a whole line of offerings now when it comes to dealing with AI's impact. Uh, the two courses are kind of similar but they're from for different levels in the org. So uh, the one that shares the title of this episode called product development for AI's impact, uh, this is for directors of product engineering, UX, sometimes VPs, basically the people accountable for how the whole system performs.

Speaker D: Mhm.

Gil Broza: So we work on three things. Seeing your delivery system clearly, redesigning the way of working for AI conditions and leading people through it. Um, the second course, uh, it's called Leading AI Enabled Product Teams. So same foundation, but this one is for frontline managers. So engineering managers, PMs, design leads, team leads, really the people in the day to day who feel the impact on themselves and their teams. So we cover designing an AI enabled way of working, planning and managing at AI speed and uh, really leading people when their work is changing under them. So things like engagement and growth and teamwork.

Moshe Mikanovsky: Mhm.

Gil Broza: What happens to teamwork when you have machines doing a good chunk of the work? Right, right, yeah.

Speaker D: And where can they find those courses?

Gil Broza: Uh, on my website, 3pvantage.com um, I can tell you that they seem to be in good demand. I announced them in my newsletter just a couple months ago and there's already a bunch of interest in them because I think they're addressing a real issue.

Speaker D: Yeah, everyone wants to learn about AI. The thing is that everyone is focusing on what is AI and building AI agents and leveraging AI tools and stuff like that. But I've not seen yet. And when you told me about those courses I got really excited. I don't see yet, uh, people focusing on how do you really utilize it properly in our line of work, which

Gil Broza: requires connecting the dots that usually do not get connected. Again, systems thinking, mindset, leadership behaviors, ways of working, not just tools.

Moshe Mikanovsky: Yeah, yeah, totally. Yeah. Going back to your other point where they just give you a tool and say hey, see what you can do with it.

Gil Broza: Yeah.

Moshe Mikanovsky: And I was even talking to the team the other day that AI is just changing so drastically that it is a gross overload. Like, I mean you can go on LinkedIn now and you, it's unavoidable, uh, in what AI is being talked about and you could really get overwhelmed and get lost with it and do the wrong things. And so I think going further back, being a system thinker, coming up with a system, like what are all the teams in the company trying to accomplish with AI and make sure everybody's on board with it and say, hey, this is a magical thing, but it's not a be all, end all and it isn't going to cure everything and we

Gil Broza: can't get it right the first time. So you know, just one more thing, one basic thing in agility was, you know, continuous improvement. So we would get together every couple of weeks, maybe also at the end of the project. And all of that again, not negotiable anymore because everything changes so fast both in terms of what the tools can do and in terms of what we do with them and the trouble we create and what we learn along the way. So continuous improvement. You also need somebody to own it now. Don't just leave it to the teams to. Yeah, yeah, you know, they'll change the ticket format. Okay, fine, but marginal.

Moshe Mikanovsky: Yeah.

Gil Broza: You really need to think a lot about your way of working and do that frequently because it can quickly get out of hand or simply be suboptimal.

Moshe Mikanovsky: Yeah.

Matt Green: Amazing.

Gil Broza: Mhm.

Speaker D: Um, you. Last time that we had you on the show, we talked about your last book and I think it was a couple of years ago because it was more than 50 episodes ago and I,

Gil Broza: uh, think it came out around that time. Yes.

Matt Green: Yeah.

Speaker D: Um, are you planning to write a book about this topic?

Gil Broza: No, no, I think I'm done with books.

Moshe Mikanovsky: Exhausted.

Speaker D: People don't read them anymore. They read their AI prompt, uh, responses.

Gil Broza: Yeah, I think that's the case. I think the economic case for this type of book has changed, uh, materially.

Moshe Mikanovsky: M. It'd be hard to keep up with it these days. Like keep up with it. That's provision.

Gil Broza: Yeah, I mean, all my books, I wrote them in a way that I thought of them as evergreen and by which I meant they would last for like five, seven years with AI.

Speaker D: Um, but uh, the first principles are the same. It's about system thinking. It's about why we're building what we're building. It's about, uh, the humans that are building for the usage of other humans and not for, you know, just for the sake of building. Even though some people do that.

Gil Broza: And leadership.

Speaker D: And leadership.

Gil Broza: Absolutely. Leadership. And you pointed before leadership versus management. Yes, because this, this is not something we just manage into existence. We lead people who then bring it into existence and that's totally different.

Speaker D: Absolutely, absolutely. Matt, do you have any other questions?

Moshe Mikanovsky: No. This has been a great conversation. Uh, it's covered a lot of the things I've thought about as a product leader and thinking about all the different tools and how they're impacting us all. And it's been a pleasure having you on the show again, Gil.

Gil Broza: Uh, same as always? Yes, thank you. In two years we'll talk again.

Speaker D: Hopefully less what will happen with AI by then. Um, maybe then my predicting the developers, uh, will have more risk than product managers will be real. But I don't know.

Moshe Mikanovsky: I'm not sure.

Speaker D: I'm. Uh, so tell me, remind um, our audience where they can, uh, reach out to you.

Gil Broza: Uh, so two places. Again, my website is3pvantage.com and I'm also on LinkedIn. Uh, I have a newsletter that people can sign up for. Again, you can sign up on my website. It's the same content as I put on LinkedIn, but LinkedIn never shows it to anybody.

Moshe Mikanovsky: Yes, you got to put AI in it.

Gil Broza: Or simply the victim of m. Um,

Speaker D: so I just, um, wrote to someone on Friday on his post about LinkedIn that it feels to me now LinkedIn becomes a dead horse. That we are riding a dead horse. Um, we are kind of in a habit of being there because we've been there for so long. I'm not sure what's going on there anymore.

Moshe Mikanovsky: The algorithm has definitely changed, uh, recently, so I can tell.

Speaker D: Yeah, I don't know the algorithm, the, the um, participation, what people are writing. Um, I don't know if there is value in it anymore. As simple as possible as that.

Gil Broza: Yeah. Um, yeah, so. So the newsletter is the better way to catch up on my thoughts. I publish, uh, every week or two.

Speaker D: Perfect.

Gil Broza: It's always articles, right? So it's not, uh.

Moshe Mikanovsky: Yep.

Speaker D: Yeah, perfect.

Gil Broza: Good stuff.

Speaker D: Thank you so much, Gil.

Moshe Mikanovsky: Thank you, Gil. Appreciate it so much. Thank you. Uh, m. Thank you to all the listeners and we'll talk to everybody next time. Take care.

Gil Broza: All right, bye.

Speaker D: Take care everyone.

Moshe Mikanovsky: Thank you to all the listeners. We really appreciate the feedback and support. Please leave us a review to help others find the show on Apple or Spotify or anywhere else. You're listening to the show.

Gil Broza: Mhm.

Matt Green: Sa.