Who Should Be the Most Nervous? Product Managers, Designers or Engineers?

Alex Radu, Pamela Mead, Aastha Yadav, Rory Madden at UXDX EMEA. Video: https://youtu.be/ExWZL5Mm-po

Readable transcript: edited from the recording's captions for readability (fillers and false starts removed, punctuation and section headings added). Wording is the speaker's own. Timestamps are positions in the video. Names marked [?] could not be verified against the audio.

The Mexican standoff in the product triad

[00:00:09] Rory: Great. We've designed this as well: it's product, design, engineering, just so it helps you out.

[00:00:16] Pamela: I feel great already. [laughter] In the middle.

[00:00:20] Rory: All right. As I gave in the introduction, the title of this is: who should be the most nervous? Is it product? Is it design? Is it engineering? I don't know if you saw the thing a few weeks ago, Marc Andreessen said there's a bit of a Mexican standoff going on in the product triad, because product managers think, "Oh, I can just develop things. I can prompt. I don't need designers and engineers." And designers go, "I don't need coders anymore." And engineers are like, "What do we need all those product people for?" There's this Mexican standoff happening. What we want to talk about is: is this real? What's actually happening in the industry, and how are things changing?

Will AI replace people?

[00:01:00] My first question, and I'm going to go straight in for it: do you think AI is going to replace people? I'm going to start with you, Aastha.

[00:01:11] Aastha: I think AI is not going to replace people. It might replace or take over some of the tasks from us. How I look at it, it also depends very personally on the people. All of us are built differently. Some of us like digging into things, and some of us like to be more broad. For every single individual, where they find their strength and where they find maybe a lacking gap is probably where they start leveraging AI first. I think it might not replace people, but it might make it more competitive to be good, because now everybody has the resources to be good.

[00:01:49] Rory: And do you see that competitiveness within a function? Or do you see there's going to be a little bit of a cross-functional...

[00:01:58] Aastha: I think definitely both, especially when it comes down to ownership. Previously, organizations had siloed ownership. You had product: okay, you're supposed to deliver so many products. With engineers, it might be a different KPI or metric that they're evaluated on. For product designers, it's different. But I think the goals are merging as well, because the responsibilities are merging. It's like, okay, it doesn't matter who's doing what, let's get the thing done. In these situations there might be an inter-function competitiveness as well. But I think it depends on the organization itself. If you have a bigger organization and multiple teams are competing for resources, then it might happen within the function as well.

[00:02:42] Rory: Excellent. And Pamela, that was leveraging your gaps so you can use the tool to improve. Do you see it more within the function, or more cross-functional?

[00:02:55] Pamela: I do definitely think that we are going to be more cross-functional, but there is pressure on design. This is not a new pressure, as I'm sitting in between, because I think there's always been a little bit of a sense of "of course everybody can design." I think that's the phase we're currently in, where 56% of non-designers are doing some form of design task, according to a Figma report from 2025.

[00:03:22] The pressure, I think, is around design, but from my perspective it brings us back to where maybe we were 10 years ago, with a level of quality and strategy for design. I think that's where the shift will be: in refocusing us on the strategic questions. I think there are some skill sets that designers have that are pretty unique that I believe will become more important. That's what I spoke about last year: this core depth of knowledge that we need to be bringing to the table to be able to make better judgments, better prompting, better decisions around what to build. There is pressure, but I also have optimism that once we're through this mess, it will become a little bit clearer.

[00:04:11] Rory: How I'm hearing that is: yes, you think AI will take on more of the UI design aspects of it, but there are still the right questions to ask.

[00:04:23] Pamela: It will be able to. It's a tool. That's how we're using it right now, I think. I think it has more capacity than that, but we need to learn how to use it well, and I think we're still in the middle of figuring that out.

[00:04:36] Aastha: I just want to add to this. Forget applications. I have seen some documents and decks that product managers and engineers have built, and trust me, we need designers. [laughter] Not even going to the level of applications, there's definitely an eye for the aesthetic and what pleases, which is a special skill.

[00:04:55] Pamela: Yes, I've seen that throughout my career, unfortunately. In that regard it's not new, but the tension is more intense.

[00:05:07] Rory: Excellent. I'm going to go to engineering, because I think with AI, engineering is probably where it's most advanced at the moment. It's catching up in design, but in engineering a lot of people are saying it's writing 100% of my code. Do you think AI is going to start replacing developers?

[00:05:25] Alex: I don't think so. I think we still need engineers and we still need developers, but we need them for different things. I know some AI company CEOs have said coding is solved, and maybe it is, but good coding, good software, I'm not so sure about that.

[00:05:44] When you're having an AI write a function that's maybe 100 times larger than it should be, where a senior engineer would look at it and say, "I can actually solve that in three lines," that's a problem, because you're in a way making your own problems. If the code base becomes so large, and if your engineers don't actually read and review and approve the code that's being written, or rewrite it, or ask the AI to rewrite it, then you start to have problems with unmanageable amounts of code that maybe do the right thing now, but you don't know what other edge cases have been introduced in the code, because you haven't actually read it and you don't understand it. You could ask the AI, but AI likes to please. AI likes to be chatty. It might come back and say, "Well, yeah, I think this is all right. I think I did what you asked me to."

[00:06:37] But you can never really be sure. In some industries that's not good enough, and it will continue to not be good enough. Would you get on a plane that an AI wrote the code for? Today? But at the same time, humans are not very good in high-risk situations, so we do need to automate. However, there's a decision between: do we just prompt an AI to say, "Fix this," or do we actually spend a lot of time in research and design and planning and build something that's going to be good software or good technology, good engineering, good design, good product? Which takes a lot more time and effort than just asking an AI to make you an app.

Good and bad patterns of AI use

[00:07:22] Rory: That leads into my next question perfectly, because there are different ways that people are adopting AI. I can't remember the phrasing that he used, but it's abdicating thought, where you're just going, "Build me this," and you're not even thinking about it. How are you seeing different people using AI, and what are the good patterns and what are the bad patterns?

[00:07:48] Alex: Weirdly enough, I think the good patterns are the ones where we do have cross-functional collaboration. The patterns where an engineer is going to build themselves an AI PM and an AI designer don't really work, because they're not entirely sure what they're looking for from those agents or from those capabilities. They're just asking, "Based on your best knowledge of the world, build me a PM to replace my PM." That's not really a good option.

[00:08:16] The other part, and I was talking to Pamela about this earlier, is that you might have more emerging technologists, younger folks that maybe are just out of university, or in university, or in school, or not in school. They might be using these tools to build things. That doesn't mean they understand what they're building, and understand the problem, and understand the customer that they want to maybe sell this to. That takes time, and you need to build foundations. But what they're doing now is using AI to do their homework and their exam preparation and their submissions, which means that they're not actually building their critical thinking and the skills that are going to make them valuable builders once they graduate and they want a job. We're shortchanging ourselves, by ourselves. I don't really know how we fix that.

[00:09:07] Rory: We'll come back to that, because I love that topic about losing the skill building that used to happen in the past. I want to come back to that. But just for now, you're saying that the best patterns are still having human experts, not just trying to replace everybody with an AI, and being able to then prompt it correctly.

[00:09:30] Alex: Or being able to put, and this is another example, let's say you have an application or a product: being able to put all of that knowledge into one place that is accessible to all of your different team members and roles, for them to be able to self-serve that information. Whether they're an engineer that wants clarity on what the customer actually wants, or a product manager trying to understand, do we already have this feature, why is it not working, or a researcher trying to see what the usage is of some of the features we released in the past couple of months, do we need to do a bit more digging there. Having all of that together, and then giving people the ability through these tools to understand and self-serve and deep dive and get that additional knowledge of how we do this well. And then being able to come back to the team and say, "Hey, I have this idea based on what I saw in our Uber knowledge base of our product."

[00:10:28] Rory: Yep. Because there was a lot on documentation and product: we've talked about the engineering side, but from a product perspective, what are the best ways that people are using it, and maybe some of the bad ways?

[00:10:41] Aastha: I think the best ways, of course, are these micro pockets of efficiency that people are finding. I started focusing a lot more on product growth and business development only recently, and was mainly focusing on product strategy and planning. What I see as the main gap is that earlier, we might want to build products fast, but we would do it with a lot less context, because either we didn't have the capability or the capacity to dig into that information and those data points.

[00:11:22] What AI allows us right now is, instead of building the first prototype only to break it, you can already access a lot of your information at your fingertips a lot faster, instead of scratching everything out from first principles. You should always still use first principles, adding to the skill aspect, because you have to have that creativity that starts it.

[00:11:44] Some of the ways that people are not using it well: one is that original thoughts are becoming more and more rare. The aspect of reading something and actually getting the "aha, this is something cool and interesting, and this actually differentiates my product from the competitor." Because you are not building your product in a silo, unless I guess you're Nvidia and you do not have any competitor in the world. Most of us are working in competitive industries. It's not just internal; from the user's perspective there's another product which is competing as well. I feel like what's happening is that this differentiation is becoming more blurred, and this is where people need to bring in their creativity to differentiate the products.

[00:12:33] Another thing where I think people are not using it that well is accountability. At some point, a year or so ago, I started having some of my team members, if there was something missing or wrong, let's say in a data point, say, "Oh, AI gave me this." I think it has become a more common excuse now, to say, "This happened because AI gave me this." There's this aspect of accountability which also needs to be embedded, which is: "Hey, it's okay for you to leverage AI as a tool, in coding or building or process management, but at the end of the day you still need to have that human in the loop to say, okay, I have verified this, and I'm accountable for my output."

[00:13:15] The great ways are everything from prototyping, fast iteration, listening to your customers better, and also building personalization for your customers better. I think this is a sector where AI definitely helps users a lot more on their side, because for products where you do need better personalization, it's easier to build. And then the wrong sides are the accountability aspects, and finding this gap between what is really needed.

Designers, strategy and tension with product

[00:13:44] Rory: Great. I'm going to try to be controversial there, just to get a little bit of controversy on the stage. Pamela, you mentioned at the start you wanted design to move away from, I know a person who called it the coloring-in department, from that concept of it just being about UI. But what we just heard there was product differentiation, personalization. These are, I guess, moving into design strategy. Or do you see it differently? I guess the question to you is: how are designers using AI to move into that strategy, and do you see there being a little bit of conflict with product?

[00:14:23] Pamela: That's almost a yes to that. The loaded question. There's been conflict with product for a long time. Right now, the way designers are using AI is more as tools. There is, of course, the expediting of processes and documentation and some of the more standard tasks that we have to do on a daily basis, or an annual basis for that matter. I think that's getting much more simplified.

[00:14:51] But the design tools, or the prototyping tools, are where I see some designers really exploring: how can we understand the problem a bit better, and then solve it better, to have a first pass at what the solution could look like, and how do we learn from that? I think there's always a risk. You can build things and not really fully understand them. If you don't really understand what the problem is, there's some thought process you have to put into the solution. That's where good designers still use their judgment and make better decisions. And designers who are not so familiar, or don't have the depth of the skills, will prototype and not be any better.

[00:15:29] Rory: It looks right, and that's the deceiving part.

[00:15:32] Pamela: That's been deceiving for a long time. I think it's just faster that it becomes deceiving in that regard. But what I've seen a fair amount, when I talk to some of the more senior people on my teams, is that they know when not to use AI. They know when it's just going to take much longer for us to prompt and get the prompts right and do all of that, rather than just use Figma, or use the AI tools that we have, to just do it right. Getting really smart and good at knowing what tools to use and when to use them is really the magic answer, I think.

[00:16:04] You also asked the question about strategy and PM. I have always looked at the role of designers as being strategic, at least as you get more senior. And yes, that has often caused some level of tension. But I think that's where it's important, with this acceleration of product and builders, for teams to work together, trust each other, not be territorial, and really make that become almost the real version of agility that was envisioned such a long time ago. I think that potential exists now. But you need to have the knowledge, you need to have the experience, and you need to have the trust of each other.

Do we still need deep expertise?

[00:16:47] Rory: And that ties back to your point, Alex, about needing that expertise. There's a risk that if you're in college and you're trying to learn, and you go, "I'll just get AI to answer it," you're not actually building that skill and that judgment that was mentioned. Alex, do we need that deep expertise, or is it a case of we're just the old fuddy-duddies who think the horseshoes are still needed when we're moving to cars?

[00:17:18] Alex: I think it depends, because there will be a period of transition as we decide what we should be delegating and relegating to automation and AI, and having it help us do what we want to do faster and more efficiently, and have time for other things. We were also talking earlier about how humans are builders by nature. Think about all of you, maybe at some point when you were kids, you had some wooden blocks or some paper blocks, or you went to a beach and built a sandcastle out of nothing. Somewhere inside of us there's always this desire to build.

[00:18:01] But to be able to build a sandcastle takes very little expertise or thought, and it takes a lot of creativity. To be able to build an airplane, probably not one person can build that, unless it's a very small airplane. It requires a lot of expertise and time and understanding of systems. Maybe that's what we still need to have. We still need systems thinking and understanding, and we still need to understand how to build the harness, even though we don't have horses anymore.

[00:18:35] That's the part where you still need to be able to do it, and the only way you can build systems thinking is if you understand systems and you study systems. If you can't do that, how are you going to learn? If you can't study, and you can't learn and unlearn and relearn, how are you ever going to become better? In Japan, which is one of my favorite countries, people will have one job that they dedicate their life to becoming excellent at. I don't think there are enough people in the world that want to be excellent at something, and therefore I don't believe that our use of AI, if we're not excellent, will be excellent.

[00:19:16] Unless we actually dedicate time to learning, and understand that you still need practice, you still need work, you still need ownership, you still need partnership and collaboration. These things don't just disappear because now you have someone you can talk to and ask anything you want. But most of the time there's no one else there to tell you, "Hey, that's not actually right," or, "I think there's a better way."

[00:19:41] Again, Pamela and I were talking about this before. Communication is a problem; we all know that. We send a text, we send a message, we call someone, and things get lost in translation. Before, I would email Rory, and Rory would come back to tell me, "I didn't get that," or, "Can you explain? Can you tell me?" Now I can go to an AI chatbot, and I can tell it anything.

[00:20:04] Rory: Brilliant. You're spot-on.

[00:20:07] Alex: And I tell it to the chatbot, and the chatbot's like, "Oh, this is amazing, Alex. You rock. Let's do this. This is awesome." And I'm like, well, actually, maybe it wasn't amazing, but how am I ever going to know, because all you do is tell me yes? Obviously I can prompt it and tell it, "Tell me no." But if I'm a student who's maybe 16, who loves the fact that someone told me I'm awesome, will I then think, "Well, what if I'm not awesome?" Probably not. Is that a good thing to teach people? I don't know. I don't know what the final answer is.

[00:20:42] Rory: I guess there's the concept that we need to have the expertise because the AIs aren't smart enough yet. They'll just do what you ask and tell you you're amazing, and it might be completely wrong. Is that just a time thing, where we think that next month there's going to be Mythos, or whatever model gets released, and then it'll know that? From a product perspective, what are your thoughts on deep expertise? Do we really need it, or is that just a temporary bump now until AI has gotten over that step?

[00:21:22] Aastha: I think at least right now we definitely need more expertise, because of what is happening, which is that things are becoming more and more general. Unless we have somebody who has domain expertise, or actually deep knowledge of it, we wouldn't find what is missing.

[00:21:40] I would think of it as a two-by-two. There's the skill and capability aspect, and then there is the expertise and taste aspect. Depending on which area you fall into, and the work you need to do, you can leverage the AI. For example, let's say you're a product manager who is great at working with engineers and thinking about the product and writing a PRD, but not so great at doing competition research or UX research. You know that this is needed. You do not have the expertise. You do have the taste for it, because you know what a good product should look like, but you do not know how to do this particular level of task in the overall outcome.

[00:22:34] I would definitely leverage AI there, because I understand that I would get something done faster where I do not have the skill, and I can save the to-and-fro time asking people what they mean. On the other side, what needs to be thought about is how much, because if you're doing everything with AI, that means you need a verification layer for all of these things as well, which takes an equal amount of time. Which brings us to the topic of: should I rather have just done it myself than go to the AI?

[00:23:06] For this, at least in my head, I have a simple framework, because of course I also sometimes cannot spend all the time evaluating every outcome. I realize, okay, the output here pushes me in a direction for a decision which is either reversible or irreversible. We know that for irreversible decisions you would rather put a lot more time into validating everything you haven't done yourself, and for reversible decisions, if it's an internal conversation, you can definitely use the AI.

[00:23:40] I don't know if that answers your question exactly, because a little bit of a difference is also that a lot of us here are maybe more on the SaaS product side, and I'm coming from a domain which is very much physical, in-person, real-life operational interaction, while also building a software layer on top of it. It's a little different, because you do need very much of a human layer out there to validate the experience as well.

[00:24:05] Alex: Can I ask a question on that? I agree that some decisions are reversible and maybe you're not checking them. But if I'm on the receiving end, and I don't know if Pamela is, and I keep getting things wrong from you, I will lose trust in you at some point.

[00:24:21] Aastha: Exactly.

[00:24:21] Alex: I don't know if I necessarily agree with "if it's not really that important, don't spend all the time checking it." If you get it wrong again and again, not you but the AI, ultimately the ownership is yours.

[00:24:35] Aastha: I think there would be...

[00:24:35] Alex: Your partners will probably be upset.

[00:24:38] Aastha: There would be some guardrails, and of course a minimum quality level. You would not send something which is completely wrong. What I'm talking about is a little bit more the structure, the design, the specificity. Am I cutting out some of the extra fluff that the AI adds to it? How much time am I putting into editing it, so it doesn't come across as an AI-generated thing? Because people put a lot less value on AI-generated things these days.

[00:25:05] This also happens on the growth side. If your user on the other side believes, from a customer support angle, that this is not a human but an AI, their likelihood of feeling satisfied, even if it resolves their problem in a few extra steps, is lower. They might give it four out of five and not five out of five, because they feel, "I deserved a human interaction at this level, maybe, and I did not get it." So of course, having some minimum quality, and not giving just subpar work.

Q&A

[00:25:36] Rory: We could keep talking for ages, but we're actually out of time, and I only got halfway through the questions I wanted to ask. If we can flip to the Q&A slide. You probably didn't realize, but some people have been asking questions. You're able to ask questions throughout the entire talk, so that by the time we get to Q&A we have some questions loaded up.

[00:25:59] We'll go with the first question that we have here: "How do you show CEOs and boards that firing thousands of employees to replace them with AI is a mistake, without waiting years?" I guess the consensus here was that we do still need people, because AI is making mistakes, it's damaging trust, it doesn't have the right skill level to ask the right questions and the context. Yet we're hearing Meta laid off 10%, and all of these layoffs. Does anybody have one that they want to answer on that?

[00:26:35] Aastha: I can take this up. This first question is very relevant, because apparently last week there's another company called Bolt, which is a fintech company, bolt.new, and they apparently laid off their HR. They didn't lay off their HR team, but what they did was say, okay, we got external contractors and we got rid of our internal people teams. I know this because we got a huge backlash on LinkedIn on our profile, because we share names as companies. Our company then took a stance, literally on LinkedIn, publicly saying we are not getting rid of our people department. We still believe they're quite important to bring culture and different aspects of the company.

[00:27:20] I don't know if I'm allowed to say it: some CEOs might still do it. They will get rid of a lot of people in the name of getting AI specialists, but the likelihood is, if they do it without thinking, they might fail. It's the game: fool around and find out.

[00:27:39] Rory: Yeah, I don't know. I think it's probably a little bit of AI washing. I think these CEOs are laying off people and they're looking for the nicest excuse that will get them the least backlash.

[00:27:52] We'll move on to the next question, because it has a lot of upvotes. "What about the environmental, ethical and privacy implications of AI? Is anyone prioritizing this at their companies? It seems to be mostly ignored in favor of business growth." I might go to you, Alex, because on the banking side of things there's a lot more privacy and a lot more regulation.

[00:28:13] Alex: I would say the same rules apply to AI that apply to everything. AI is not an exception. We still have the same standards, the same laws, the same regulations that govern us as a highly regulated industry and company. I would say that there is the desire from the companies, and then there's the ability of the tool-provider companies to satisfy it. Similar to what we did when cloud started becoming a thing, 10 to 15 years ago, it took a while for FinOps for cloud to reach the level that we needed it to be, to be able to make sustainable decisions, or to understand how we shift consumption patterns to be more sustainable. It took the vendors a long time to be able to provide that information and those tools.

[00:29:07] So I would say yes, especially in bigger companies, this is still a very real concern. There are still very real questions about it. But unfortunately, some of the companies that are providing this are companies that are 100 to 200 people strong, and that have other things that they want to do first, as opposed to actually prioritizing the data that should be coming behind the tooling. I think we're going to get there. Since their tools are so great, probably sooner than we did for cloud.

[00:29:47] But yes, I think people are still concerned, and the same rules still apply. It's not like anything changed in the industry in terms of regulations because AI came out. It will probably change in the next couple of years; there are new AI regulations coming out, and they will probably be more strict. I don't think it's going to get any lower. I think it's only going to get stricter.

[00:30:11] Rory: Great. We're out of time, but there's one question, so I'll get the quickest answer from all three of you. Who is the most nervous? Do you think it's product, design? What are your thoughts, quickly?

[00:30:22] Aastha: I think we're all nervous, and overwhelmed with the amount of information and updates that we're getting. And imposter syndrome.

[00:30:30] Pamela: I think I agree with you. I think we're all nervous. However, I do think that right now design, or design-related or creative fields, are the hardest hit. I think that group needs to really get back to its strategic roots and its depth and knowledge, and bring that back to the table.

[00:30:51] Alex: I will just say that, at a personal level, I think every human right now exists between excitement and fear. I don't think that we can speak for everyone, but I think everyone is a little bit nervous somewhere deep down.

[00:31:04] Rory: Brilliant. Well, that's all we have time for. If you can put your hands together, please, for our panel. Thank you very much.