Today, I'm chatting with Stuart Winter-Tear about AI product management. We're getting into the nitty-gritty of what it takes to build and launch LLM-powered products for the commercial market that actually produce value. Among other things in this rich conversation, Stuart surprised me with the level of importance he believes UX has in making LLM-powered products successful, even for technical audiences.

After spending significant time on the forefront of AI’s breakthroughs, Stuart believes many of the products we’re seeing today are the result of FOMO above all else. He shares a belief that I’ve emphasized time and time again on the podcast–product is about the problem, not the solution. This design philosophy has informed Staurt’s 20-plus year-long career, and it is pivotal to understanding how to best use AI to build products that meet users’ needs.

Highlights/ Skip to

• Why Stuart was asked to speak to the House of Lords about AI (2:04)
• The LLM-powered products has Stuart been building recently (4:20)
• Finding product-market fit with AI products (7:44)
• Lessons Stuart has learned over the past two years working with LLM-power products (10:54)
•  Figuring out how to build user trust in your AI products (14:40)
• The differences between being a digital product manager vs. AI product manager (18:13)
• Who is best suited for an AI product management role (25:42)
• Why Stuart thinks user experience matters greatly with AI products (32:18)
• The formula needed to create a business-viable AI product (38:22)
•  Stuart describes the skills and roles he thinks are essential in an AI product team and who he brings on first (50:53)
• Conversations that need to be had with academics and data scientists when building AI-powered products (54:04)
• Final thoughts from Stuart and where you can find more from him (58:07)

Quotes from Today’s Episode

• “I think that the core dream with GenAI is getting data out of IT hands and back to the business. Finding a way to overlay all this disparate, unstructured data and [translate it] to the human language is revolutionary. We’re finding industries that you would think were more conservative (i.e. medical, legal, etc.) are probably the most interested because of the large volumes of unstructured data they have to deal with. People wouldn’t expect large language models to be used for fact-checking… they’re actually very powerful, especially if you can have your own proprietary data or pipelines. Same with security–although large language models introduce a terrifying amount of security problems, they can also be used in reverse to augment security. There’s a lovely contradiction with this technology that I do enjoy.” - Stuart Winter-Tear (5:58)
• “[LLM-powered products] gave me the wow factor, and I think that’s part of what’s caused the problem. If we focus on technology, we build more technology, but if we focus on business and customers, we’re probably going to end up with more business and customers. This is why we end up with so many products that are effectively solutions in search of problems. We’re in this rush and [these products] are [based on] FOMO. We’re leaving behind what we understood about [building] products—as if [an LLM-powered product] is a special piece of technology. It’s not. It’s another piece of technology. [Designers] should look at this technology from the prism of the business and from the prism of the problem. We love to solutionize, but is the problem the problem? What’s the context of the problem? What’s the problem under the problem? Is this problem worth solving, and is GenAI a desirable way to solve it? We’re putting the cart before the horse.” - Stuart Winter-Tear (11:11)
• “[LLM-powered products] feel most amazing when you’re not a domain expert in whatever you’re using it for. I’ll give you an example: I’m terrible at coding. When I got my hands on Cursor, I felt like a superhero. It was unbelievable what I could build. Although [LLM products] look most amazing in the hands of non-experts, it’s actually most powerful in the hands of experts who do understand the domain they’re using this technology. Perhaps I want to do a product strategy, so I ask [the product] for some assistance, and it can get me 70% of the way there. [LLM products] are great as a jumping off point… but ultimately [they are] only powerful because I have certain domain expertise.” - Stuart Winter-Tear (13:01)
• “We’re so used to the digital paradigm. The deterministic nature of you put in X, you get out Y; it’s the same every time. Probabilistic changes every time. There is a huge difference between what results you might be getting in the lab compared to what happens in the real world. You effectively find yourself building [AI products] live, and in order to do that, you need good communities and good feedback available to you. You need these fast feedback loops. From a pure product management perspective, we used to just have the [engineering] timeline… Now, we have [the data research timeline]. If you’re dealing with cutting-edge products, you’ve got these two timelines that you’re trying to put together, and the data research one is very unpredictable. It’s the nature of research. We don’t necessarily know when we’re going to get to where we want to be.” - Stuart Winter-Tear (22:25)
• “I believe that UX will become the #1 priority for large language model products. I firmly believe whoever wins in UX will win in this large language model product world. I’m against fully autonomous agents without human intervention for knowledge work. We need that human in the loop. What was the intent of the user? How do we get that right push back from the large language model to understand even the level of the person that they’re dealing with? These are fundamental UX problems that are going to push UX to the forefront… This is going to be on UX to educate the user, to be able to inject the user in at the right time to be able to make this stuff work. The UX folk who do figure this out are going to create the breakthrough and create the mass adoption.” - Stuart Winter-Tear (33:42)

Links

• Stuart's LinkedIn

Transcript

Brian: Welcome back to Experiencing Data. This is Brian T. O’Neill today. I finally, finally have—I did ask a while ago, Stuart—[laugh].

Stuart: I know, I understand.

Brian: I finally have the Stuart Winter-Tear on the line today.

Stuart: The Stuart Winter-Tear. Thank you for having me.

Brian: Yes. Not a Stuart Winter-Tear [laugh]. I think you’re the only one, probably, right?

Stuart: Probably, yeah.

Brian: Excellent, excellent. So, I think of you as someone definitely in the AI product management category, in my head. I’ve enjoyed talking to you on LinkedIn and all that, so I’m really happy to get your perspectives here on data products, and specifically AI products, and how we make them, and is this all a fad, are we doing anything different, like, what is going on in this space? So, I’m just really excited to have you here, first of all. So, thanks for joining me.

Stuart: Thank you for inviting me. I hope I can bring some value or edutainment from experience, but we’ll see how it goes [laugh].

Brian: We will endeavor to entertain and hope that the value comes second [laugh].

Stuart: [laugh]. Amen.

Brian: You know, some people don’t understand that. Like, you know, no one wants to listen to a fact-filled, boring-ass podcast. So, part of this is, like, it should be entertaining, and the stories and stuff that helps us retain the information. So, there’s a purpose to having it be fun. And that’s why I say it’s like having a beer. These should feel like we’re having a drink or something like that together at the pub. So.

Stuart: I’m really glad you said that. I much prefer that [laugh].

Brian: Yeah, that sounds good. Well, cheers here [laugh]. And the first thing I wanted to ask you was, you got to speak to the House of Lords. Like, tell me about this story. Like, and for—maybe tell people really quickly, for, like, our western listeners here that maybe don’t know what the House of Lords is, it’s like the US Senate-ish, like, to give them a sense of, like, the stature of that. That’s pretty amazing.

Stuart: Yeah, so the House of Lords is part of the parliament building. It’s where they wear all the fancy wigs and make the real decisions on law, and regulation, and what have you, the really meaty stuff. And I became an ambassador for GenAIOps, which is a sort of global initiative based in the UK, out of the UK, to do this AI stuff securely, safely, at scale, and get ROI on it, and just kind of have a team that evangelizes that. And they asked me to come over to the House of Lords to give a talk to the, you know, industry leaders and what have you, and I must admit, when I was standing at the back before I was about to give my talk, I was pretty nervous. They were all suited and booted, and, you know, and I thought, oh my goodness, you know, what can I tell these people?

And something came over me. I just decided, like, most people come up and give their credibility, establish their credibility who they are. And I suddenly found myself at the mic saying, “My name is Stuart Winter-Tear. I’m not going to introduce myself because, frankly, it’s boring to me and it’s boring to you.” And that kind of cracked the ice at that point, and I thought I’m just going to carry on riffing like this.

And I don’t think they’ve probably ever seen anything like it or expected it, and I think that’s probably what worked. And there was a lot of laughter. I didn’t realize I could be an AI comedian, but hey, I seemed to do it that day, and it’s exactly like you just said, Brian. It’s that, you know, you’re dropping the value, you’re dropping the experience and how you see things, but doing it in a way that is memorable. And I was surprised because I saw lots of LinkedIn posts go up afterwards where people were kind of quoting bits that had landed with them.

And you sort of thought, hold on, people, don’t want the dry, always packed with this knowledge. You know, we live in a world where we do need to have a smile as well. And I kind of try to reflect that in LinkedIn posts, to kind of have a bit of humor. Why not?

Brian: Excellent. I love it. Well, Stuart, that’s really amazing to get the—an honor, I’m sure, to go and speak in front of the House of Lords, so congrats on that. So, tell me, just tell the audience, since they don’t know, kind of, quickly summarize some of your recent work and product you’ve been building LLM-powered products for the commercial market most recently, correct?

Stuart: Yeah, absolutely. So, not new to artificial intelligence because I spent so long in the cybersecurity world, but now we have to refer to that as, sort of, traditional machine learning, you know, to sort of differentiate it. I think a lot of people just think GenAI is the totality of AI. But I must admit, I was very much like everybody else. I mean, when ChatGPT came out, it was the first time I’ve been excited about technology for about 20 years, to be honest. Everything had got quite boring to me, and spending so long in cybersecurity, you tend to get a bit jaded.

So, when it came out, I was lucky because I had some time off—literally—as it came out, so I really embarked on these courses to learn as much as I could. And in the end, I was drowning in Greek symbols, and Python, not my specialty, but it does give you the ability to be able to speak to researchers, and I mean data researchers, you know, data scientists. So, you do get a bit of kudos and credibility for that, and you’re not starting from scratch. And so, I joined a startup, and it was building specifically large language model-powered products for the commercial market. So, you certainly learn a lot when you’re putting these things out to the world. Put it that way.

Brian: Yeah, I’m curious, well, some of the things you learned, and maybe if you can give us some context about, like, why would someone pay for one of these products? Like, what was the value that they were-the benefits that they were producing?

Stuart: I think the majority thing I can say is to be able to consume enormous amounts of unstructured data, and make some—surface some patterns, surfing something from it of value, insights, whatever that is. I think that the core dream, as I see it, with generative AI is, firstly, getting data back out of IT hands, as it were, and back to the business. And finding a way to overlay all this disparate data, this unstructured data, and be able to speak to it in a way that is in the human language is pretty revolutionary. And we were finding that industries that perhaps you would think were more conservative, like medical, like legal, were probably the most interested in this kind of technology, and I think it’s because of the large volumes of unstructured data that they have to deal with.

Obviously, that brings up a whole host of other things around hallucinations and what have you, but they were, interestingly, the first that could see the benefit from it. And there were other things. People wouldn’t expect large language models be used for fact-checking, given I just mentioned hallucinations, but they’re actually very powerful at that, especially if you can have your own proprietary data or pipelines that go out to the internet that can make checks and do comparisons on the things that you’re seeing, they can actually be quite powerful. And I’ll say the same about security; although large language models introduce a terrifying amount of security problems, which we probably won’t go into, they can also be used in the reverse, to be able to augment security. So, there’s that lovely contradiction with this technology that I do enjoy.

Brian: Yeah, yeah. And were these products that you’ve been working on most recently, were these, like, startups where you were trying to find product-market fit with these LLM powered solutions, or were these like established products and you were doing feature development on top of an established solution that was in the market already, and adding capability? Can you talk a little—I’m mostly curious about, did you find some product-market fit there? Were there any financial outcomes from that? Even if you can’t tell us what the numbers are, I’m just kind of curious to know. I think this whole… there is a lot of hype here, but there’s also a lot of potential here as well, so I’m always curious about, like, are these products getting out, are they making money, besides just the foundation model [laugh] companies that are out there? How successful is this right now?

Stuart: Yeah so this is predominantly startups looking for product-market fit, okay. It’s interesting because, in part of doing that, you’re speaking to larger companies that looking to implement that technology and get value from it themselves. And yes, it’s such an interesting thing, this value. I did a talk on, you know, data and AI value. But what does value mean? I think that when most people say value, I think they’re talking bottom line; they’re talking revenue.

And that’s going to be both cost-cutting on the operational side and also increasing revenue on the market side. Some of it, not so successful. Some of it’s successful. I can’t go into the details of it, but yeah, both. And that’s part of—I don’t blame myself, from a product perspective, for that because I think over the last two years, which is pretty much, you know, the time I’ve been dealing with this, predominantly, it has been a learning curve, put it that way. So, I’m not going to make any claims that we made billions or anything like that, but we did see what I would call value, in terms of, you know, return on investment. And looking around at the reports at the moment, we can see that many companies are still struggling to turn this into a true return on investment.

Brian: I—yeah. I mean, I’m generally hearing the same thing. Everybody’s kind of in experimentation mode. They’re, you know, getting some proofs of concepts out there. Most of the stuff I encountered, frankly, is not very helpful. They look like bolt-on features on top of existing products. They’re not particularly adding much value, either in a financial sen—like, not things I would pay for, and they’re also not having the other side, which is what I call the user experience outcomes, right? Like, did you make my life better by introducing some kind of generative AI experience into the product? Did you basically save me time, save me frustration, delight me in some way?

For the large quantity of the experiences I’ve had so far, they haven’t [laugh]. Most of the benefits still comes directly working with foundation models, and doing—I find them mostly personally useful for zero-to-one idea generation, legal contract stuff, I find it really helpful for that, especially places where you can call bullshit, you know? When you know enough about the domain to know, like, I need to go verify these facts here, I find they can be really useful. But what I want to know from you is, what have you learned in two years, like, in terms of, have you changed any of your approach, or the way you’re working? Like, oh, okay, don’t start with that. Don’t start with a data set and then back it into a product. Or don’t blah, blah, blah. Like, anything you’ve changed?

Stuart: The first thing is, as I explained, this gave me the wow factor, this technology, and I think that’s part of what’s caused the problem. So, I run by a mantra, which is, you know, if we focus on technology, we end up building technology, and we have more technology, but if we focus on the business and the customers, we’re probably going to end up with more business and customers. You can see why this is so alluring and is ending up with so many products built that are effectively solutions in search of problems, okay? It’s understandable. But in that is a clue to us, and that clue is that we’re in this rush, in this FOMO.

We’re leaving behind the true, tried-and-tested things that we understood about products, and understood about UX, and understood about security and compliance, all the rest of it. We’re throwing it away as if this is a special piece of technology. It’s not. It is revolutionary, in my opinion. I am bullish on it, but at the end of the day, it is another piece of technology. And there are fundamental differences between how we should view this. I’m happy to go into that, but that’s the core of the problem is focusing on, you know, the technology, and then focusing on what it can solve. It must be the other way around.

We still got to look at this technology from the prism of the business and from the prism of the problem. We love to solutionize. Is the problem the problem? What’s the context of the problem? What’s the problem under the problem? Is this problem worth solving? And from that perspective, then, is generative AI a viable, and feasible, and desirable way to solve that problem? We’re putting the cart before the horse.

And I understand why, and I’ve been guilty of it myself. That’s why I could speak quite plainly about these lessons learned. I really enjoyed your point, by the way—I just want to pick up on it, Brian—and that’s the thing is, this technology feels most amazing when you’re not a domain expert in whatever you’re using it for. I’ll give you an example. I’m terrible at coding, okay, so when I got my hands on Cursor, wow, I felt like a superhero. It was unbelievable what I could build.

But I do not have the knowledge, the expert knowledge, in coding, to know if that code is good code, you know? Is it scalable? Is it secure? You know, all of these things, I would have to be an expert to know that, okay? Now, I believe that although this looks most amazing in the hands of non-experts, it’s actually most powerful in the hands of experts that do understand the domain with which they’re using this technology.

So, bring it into my industry, and perhaps I want to do a product strategy, and I’m going to ask it for some assistance to do so, it can get me 70% of the way there, but I’m expert enough to know, hold on, that’s rubbish; hold on, that needs to be checked. And so, great as a sparring partner, great as a jumping off point, great, as you know, not starting with a blank sheet, great at absorbing immense amounts of disparate information to get me along the path, but ultimately only powerful because I have certain domain expertise. So, I just wanted to pick up on that. I hope you don’t mind.

Brian: Yeah. No, I, at first I thought you had said you thought it was most powerful in the hands of amateurs, and my point was, I find it—yeah, with you—it’s more powerful when you know something about the domain that you’re working in such that you can call BS on this stuff—

Stuart: Exactly.

Brian: And when you don’t know anything, it’s like, boil down quantum computing at a high school level for me. I don’t know if what it’s telling me is accurate or not without doing a lot of extra homework. But that kind of gets into, like, how do you know that—I mean, I was thinking about this today—how do you know anything is true? It’s just, like, how many sales did we have last week? Well, it came from the database.

It’s like, well, was it queried right? Like, at what point do you believe because you believe? Like, it’s good enough and you’re going to believe the information, and I thought that was [laugh] kind of an interesting thing. It’s like, at what point do you really believe any information? You validate with some level of trust of the people who provided the information to you or whatever. It’s kind of an interesting space we talk about, like, facts and confabulations and all of this. It’s like, well, how do you know anything that you read on the internet is true, [laugh] you know?

Stuart: Yeah, I couldn’t agree more. And there’s no doubt for the last, sort of, decade or two, we’ve been in that just good enough digital space. It seems like a lot of products were, you know, put there to be just good enough. My only worry in that regard is around this trust aspect of this technology. Because I was always looking at, oh, how do we inculcate trust with this technology in users? Oh, it must be context driven, oh, it must be explainable, and et cetera, et cetera.

To discover that as long as it’s fast and confident-sounding, people will trust it, to the point that we’re seeing a massive uptick in usage, a decline in trust, but also—and this is the key point—a decline in validation of the output. Now, that’s a weird paradox to be in. And I’m not an expert enough to explain why psychologically this is happening, but it does mean that for some reason—and rightly so—people are trusting it less, but they’re validating it less, which is just—I don’t know that’s a weird one to me, but there you go.

Brian: Do you mean there are studies saying that, on average, users are not spending as much time challenging the results they’re getting?

Stuart: Exactly.

Brian: They’re not trying to dig deeper? They’re just taking surface level?

Stuart: Yeah, but at the same time, people are becoming more aware of confabulations. People are becoming more aware that the output might be erroneous—

Brian: May be wrong.

Stuart: Yeah. And at the same time, there’s a surge in the use of the technology, but the people validating the output has declined. And that is a strange paradox to me. I don’t quite understand the psychology behind it. But it does touch on things like, as I said, when it’s pushing out information that’s very fast, very confident, that inspires some element of trust and less validation in us.

But I think it also touches on this sort of anthropomorphizing this technology. You know that OpenAI had to roll back their ‘sycophancy’ level? Because the problem is that, you know, we do want to be agreed with psychologically. We do want our world views to be affirmed. And there’s nothing I mean, I made a confession on a LinkedIn post I’d become one of ‘those people.’

And what I meant by that was because, you know, one of the biggest use case of large language models in the world today is for therapy and companionship. And I was going through a particularly tough time myself, and I found myself drifting into that as well. And there’s nothing like, you know, something, even if it’s a mirror in a machine that’s distorted, saying to you, “You’ve got this champ.” You know, “You’re absolutely right.” “Oh, you go for it.”

Brian: “You spoke at the House of Lords, man.”

Stuart: You know—

Brian: I mean—[laugh].

Stuart: —yeah, I know. So, I should know. I should know if you go up there completely confident and break every rule, the likelihood you’ll be a smash hit with humans. So anyway, just throwing that out. There’s some interesting paradoxes with this technology at the moment.

Brian: Yeah. I like what you said, too. I think—and this is where I feel like… I’m going to generalize data teams that don’t have, like, specifically thinking about the internal data teams that are out there that might be starting to crawl slowly into the space of the product operating model, that without the experience of first principles, it’s like you’re still back to trying to take, what could we do with this GenAI technology that the board or the executive team wants to show that we’re not standing still, so you’re really moving it forward from a technology standpoint, instead of a first-principles product standpoint, which is, is there a demonstrated need? Is it an urgent need? Is it one someone would pay something for or give something up in order to get access to because it’s going to be so much better for them?

Those things don’t change because of the technology, so I’m kind of curious, like, what is—since those first—I think you’re agreeing that those first principles don’t change just because we have AI. I think most of our mature product audience out there would probably know that. What is different, if anything, when we talk about AI product management? And I’m not talking about using AI to do product work better; I’m talking about building primarily B2B AI-driven products. What is different, if first principles isn’t all there is? Have you learned anything that you’d say is different with the digital, classic software product management from AI product management?

Stuart: Yeah, I have, and I will cover that, but I just want to quickly touch on the, sort of, desperation of data and AI teams to show they’re doing something with this technology. It is known and it is painful that, you know, globally, companies are feeling they’re not getting the ROI from their data and AI teams. Now, I’ve said this before, and I’ll say it again: I don’t blame the data and AI teams. They’ve been stuck in a windowless basement, underground, cut off from the market and the business, and then berated when they don’t provide value to the business and the market. And it’s interesting because that problem was, kind of, seen by MIT and McKinsey, and independently, they said there should be a new role to solve this problem: the translator or the connector.

And you sort of think—I was sitting there thinking to myself, firstly, they’re absolutely right, you know? I don’t think they’ve had a seat at the table, and perhaps they’re talking two different languages to the business. And, you know, the business is up at night worrying about the bottom line, and that’s where we need to start, and technologists are thinking about what cool things can we do, and then looking for the problems to solve with it, which is sort of back-to-front from each other. But ultimately, I think that the irony with the translator and the connector role is that I actually see that as perfect for the product role. That’s what the product role should have been doing all along, ironically, being that connector to the market, to the data and AI teams, and the business.

I don’t know—well, I do know, and I shouldn’t make these confessions, but I do simply think the product management role has been morphed into something that it just shouldn’t have been. It’s sort of more about, you know, speed and, you know, delivery and features and, goodness knows what else, and that’s why we have digital pollution. So, I’m just going to come out there and say that. And that’s the fault of business and a misunderstanding of the role itself, to be honest. But looking at what’s different, I sometimes—I’ve only done it once or twice, so I steer away from it now-, and sometimes on LinkedIn, I’ll put up something around this, should it be AI product manager, or should it just be a product manager that does AI? I lean towards AI product manager, and that causes a lot of consternation in the industry because you will have half say, “Absolutely not. Would you have a product manager for toilets or something absurd like that?” And the other half will be in full agreement.

Now, I’m just going to come out and say it. The ones that are in full agreement, when I look at their profiles and who they are, they are deep in large language models themselves. They are dealing with this on a day-to-day basis. Those that come against it tend not to be. And I know that’s controversial. I probably should never have said that.

But there are substantial differences. The first is that I feel like the digital world is more known and understood. So, if I have a particular, be it pricing, or if I’m looking at UX or anything like that, I can probably look at a reasonably similar product and get an idea from it as to how I should, you know, best practices and what have you. Whereas, to me, they don’t yet—those best practices don’t yet exist in the large language model world, and so we are figuring it out as we go along.

This whole thing of the probabilistic nature, we’re so used to, from the digital paradigm, the deterministic nature of you put in X, you get out Y; it’s same every time. Probabilistic changes every time. That is substantially different. And there is a huge difference between what results you might be getting in the lab to what you might potentially see and what happens in the real world. And that takes some real product difference, in that effectively you find yourself building live.

And in order to do that, you need good communities and good feedback available to you, and you have to be careful of things that you might not have seen or might not have expected, which simply were not the reality in the digital world. You need these fast feedback loops. And from a pure product management perspective, we used to just have the one timeline, which was engineering, and that’s why so many product folk have unfortunately become, you know, sort of delivery monkeys trying to speed up delivery, speed up delivery, all the time. Now, we have two timelines, and the other timeline is obviously the data research side. And especially if you’re dealing with cutting-edge products, which is what I’ve been involved with, you’ve got these two timelines that you’re trying to put together, and the data research one is very unpredictable.

It’s the nature of research. We don’t necessarily know when we’re going to get to where we want to be. And in there is another trap, and that is that these models are very quick and relatively simple to get to, something like 80% accuracy. And that can bring in a very false sense of—it can lull you into thinking, okay, this is going to move really, really fast. But unfortunately, that last 20% is not linear; it’s more logarithmic, so it slows down, and slows down, and slows down. And also you’re trying to pair this at the same time with the engineering, and not necessarily knowing the best way to do this from a UX perspective.

So, I know I rambled on quite a lot there. I mean, I could even go into pricing. Even that’s all changed completely for generative AI. So, there are—that’s just a, sort of, taste of some of the fundamental differences that I think you have to really work heavily with it, to taste and experience the pain of it, to appreciate, and then start—and then I think folk will then start thinking, hold on, perhaps there is something to having product folk that are specialists in AI, in specifically large language models and how we put those together.

Brian: Do you have a sense of, in terms of the skill sets, to move into that space, whether or not data knowledge, engineering knowledge, or prior product experience are probably the biggest accelerants to being successful in that role? Is it easier to learn the data stuff if you’re, like, a seasoned PM or is it easier to learn the PM stuff if you have some kind of data background? Like, do you have a sense of that?

Stuart: That’s such an interesting question, almost impossible for me to answer. I can only answer it from my own perspective from the people I’ve spoken to. It really surprises me to see the level of enthusiasm from data scientists and folk of that ilk that really do seem very, very interested in the product side. I’m seeing more of a morph that way, in many respects, than I am the other way. From my personal perspective, I was fortunate in that I dealt with the traditional machine learning in cybersecurity, you know, obviously for full detection, anomaly detection, and how to predict what will be the next piece in a secure architecture, and what the security and mitigations will be, et cetera, so I had that background, but I was fortunate because when ChatGPT came and hit, I had a nice space and a nice runway to be able to deep dive this, and as I say, it truly excited me.

Now, I went too far. There’s no doubt about that. Before I knew it, I was drowning in courses, drowning in Greek symbols and mathematical equations that didn’t have any numbers in them, and complex Python way beyond my—but I did persevere with this. And the one thing that gave me, at the very least, was the ability to be able to have informed conversations with PhD data scientists and data researchers. And for that, I’m grateful.

It’s a bit like I live in Spain. I’m terrible at Spanish, but they appreciate you having an effort to speak their language, you know? And they’re very forgiving and very—you know? It’s the same with data science. So, I wouldn’t suggest people go in that deep as I did, but you do need that degree of fluency to be able to—otherwise, you’re starting from scratch, and you probably won’t get that respect.

Now, this goes the other way as well, by the way, okay? Because if you’re dealing with academic researchers, they tend to perfection. No, no, we mustn’t put this out to the world until it’s 99.9% accurate. You know, it’s currently 97.9% accurate, by the way. And they’ll have a big freak out about that.

From a product perspective, we’re like, “Let’s get it out. Let’s iterate. Let’s see what happens. Just, you know, get it to the wild.” As long as, obviously, it’s not going to be something that’s going to, you know, be a critical part of, you know, something critical or life thr—I don’t mean that, of course, but you know, we’re more, you know, let’s get this out and iterate and, you know, experiment as fast as possible, and it goes back to that kind of, you know, having to build live.

And so, there’s the other way where you’re trying to encourage them to be a little more daring in getting this technology out to the world. And also there is an education piece, and it’s a symbiotic education piece, between the feasibility of the technology and the models and what’s possible, the art of the possible, but also going back with this is what the business needs. This is what our users’ needs. This is the heart of the problem that we’re trying to solve. Because researchers are just as prone—not user researchers, data scientists, data researchers, are just as prone to the wow, this is really cool; let’s build it as everybody else. So, it is a symbiotic relationship.

Brian: Yeah, yeah. My general feel is—and again, I’ve been generalizing just personal opinion—is that learning the product skills is harder because product is wide and shallow, the feedback is not immediate. You can’t get a score about how well you’re doing at any moment [laugh] by, you know, run your model through the test, it gives a number back, and it tells you whether you’re good or not, you know? Like, tell me right now how would you score, Stuart on a product management—there’s like, there’s nothing you can do to go validate that information. So, by definition, both design and product management, we live in a gray area kind of all the time, and you have to be comfortable with that kind of mess.

And I find, like, getting comfortable with that part of it, and like, shipping a C-minus quality academic work, which might be A-plus quality product work because you just need to figure out, does anyone even care right now—

Stuart: Yeah, exactly.

Brian: —what we’re building, and that a C-minus technical work is good enough to figure out if they care. If you can’t let go of your connection to my work is defined by the quality of my technical output, that’s a really hard thing to let go of. And I feel like the other stuff that technical skills, while that’s not easy, you can take tests, you can get feedback, it’s fairly binary in terms of, like, are you learning the subject matter or not? It’s much more scorable, it’s more black and white, so it’s easier to know how you’re progressing. And because of that, I feel like it’s a little bit harder to move into the other role, but I’m not sure. I mean, I’m just generalizing based on my own exposure, you know? Which is not—

Stuart: No, honestly, Brian, I think that’s really insightful. And I can only speak from my own personal perspective as well, right? So, we could be completely wrong. That’s fine, too. But no, I think you nailed it. We’re dealing with the kind of messy business, messy human side, which is, you know, you’d say gray middle area, absolutely.

And let’s be honest. I think what makes good product people, if I’m really, really honest, is their ability to deal with ambiguity, their ability to deal with the unknown, their ability to especially if you’re looking at cutting-edge stuff, especially if we’re looking at true innovation, it is more sense and respond. It’s do the next right thing based on limited information. It’s that let’s place our bet here. We haven’t got all the information, but we think this. And this is why we think this, but we don’t actually know until after we’ve done it. And that takes quite a lot of courage in its own right, and I can certainly identify with that kind of more scientific rigor, academic background, looking for that more definite proof. I think that’s a good insight. Thanks, Brian. Yeah.

Brian: Yeah. At some point when we first met and we had our little pre-talk here, you had talked about how you think user experience is going to start mattering more or a lot [laugh] with some of these products. That’s not something I’m hearing outside of the UX community, who tends to think user experience is always really important, all the time, everywhere. Can you talk to me a little bit about why you think that matters? Particularly, I don’t see outside of the big technical companies, a lot of investment in this space, let alone in traditional data products, forget AI products, especially when we start thinking less interface, you’re talking to a chat thing. What do I need designers for? I’m just hypothesizing what’s going on in people’s minds here, but why do you think user experience matters at all here? Or a lot?

Stuart: Yeah, thanks for asking that. It is obvious to any, I’m not a UX expert. I’m not going to pretend to be, okay, so I just want to make that plain. It is obvious to anybody that’s in the community on social media, the UX community has been on a downer, okay? And you know what? It’s completely understandable. I mean, it infuriates me to see—and I’m just going to come out and say this—product people saying, “Oh, we don’t need UX experts anymore. We will just simply use AI.”

Okay, well, you know that narrative can be flipped around as well, so be careful. But um, so it’s completely understandable to me. And I will make two bold comments here, which I hope I can substantiate, at least from my gut, as it were. The first is, we will look back on the chat interface rather like we would look back on the DOS or command line interface for the personal computer before the GUI appeared, okay? That’s my first—and these are predictions, right, so you know, they are prone to error.

My second is, and I’ve said this boldly on LinkedIn, and it’s a gut thing again, so I do apologize for that, but I believe that UX will move from whatever position it’s in at the moment to the first position, the number one priority for large language model products. It’s going to become some—and I’ve also gone as far as to say whoever wins in UX will win in this large language model product world, and I firmly believe it. I think you need to do an awful lot of research around the psychology of all of this—and it’s far too much to go into in kind of one small podcast—the psychology of what we’re dealing with, the degree of change that we’re dealing with to start to appreciate, UX is a hard road coming in trying to figure this out, but the rewards that will be rightly due to them, you know, when they do. I mean we’re talking about, in my opinion, my layman opinion, a change design paradigm… situation.

Much was known, again, as I said earlier, in the digital space—and I’m going to sort of change them up—the old digital space and the new probabilistic space, just for linguistic ease. And this, to me, is not the same at all. I mean, if we just take something like agents, I am extremely bullish on agents, even though I post a lot on their current limitations, okay? And I’ll state that I’m against fully autonomous agents without human intervention for the knowledge work, and that’s because of things like, as complexity increases, so do cascading failure modes, okay? So, you know, one agent is not necessarily checking the next agent’s output down the chain, okay?

So, we are going to need that human in the loop. How do we do that? What was the intent of the user? How do we get that right push back from the large language model to understand even the level of the person that they’re dealing with? These are UX problems. These are fundamental UX problems that are going to push UX to the forefront.

And too, what you saw—I know this is my experience—you think, wow, I’ve cracked it with this product, to find the very next company you go into, it’s all different. It’s almost as if you’re starting from scratch. It’s just unbelievable. From a UX perspective, I’m talking about. This level of education, this level of—I mean, I did it, I fell into the classic mistakes of, oh, we’ll just give them explainability. Oh, we’ll just educate the user on the fact that this is fallible. That’d be brilliant. They’ll all read that, and they’ll be really clued up.

To find that when the model disappointed them, the first thing they did was bitterly complain as if they had no idea how it reached its decision, and they were reported it could make a wrong decision, which, of course, probabilistic models cannot be a hundred percent correct all the time. This is going to be on UX to educate the user, to be able to inject the user in at the right time, to be able to make this stuff work. And I haven’t got it figured out, I’ve not seen anybody else get it figured out, but boy, the UX folk that do figure this out are going to create the breakthrough. They’re actually going to create the mass adoption. And I know people think, “Well, we’ve already got mass adoption.”

No, we haven’t. Not really. The percentage of people using this technology compared to the 7 billion on the earth at the moment is a small percentage, you know? We just happen to be in this world, and so we think everybody’s the same as us. Everybody’s using this technology all the time. No, they’re not. But that’s when mass adoption will come, and that will be through UX. That’s just my standing on my pulpit moment. And I’m not a UX expert. I just need to say that again.

Brian: Yeah, that’s okay. I mean, It’s good to see how someone that would be utilizing those services on a team or as a resource, how that’s being seen. I’m much more interested in that than what [laugh] what UX people think about it.

Stuart: [laugh].

Brian: That’s just like data people talking about data. Of course you think it’s important because it’s your job, you know? Like, you don’t want to say, “Yeah, it’s really not that important.” [laugh]. I kind of have this theory that the successful new products in this space are going to be: you’ve got your foundation model stuff, you’re going to have access as a business to some repository of knowledge or data that gives you a competitive edge, and when you combine the power of the foundation model with your local data that you have, and you wrap it up into a custom software solution—that’s not entirely AI; it’s not just another chat box window—but it’s a purpose—built—like, you talked about cybersecurity.

So, you think about root cause analysis, remediation, steps, all of this, you don’t want just a text input for that. You may need graphics and visualizations of data. You need evidence to display at the right time. Having some chat interface could be useful there, but it’s a combination of deterministic software with AI capabilities, all powered by this extra data that gives you leverage over what a foundation model could. This is just a hypothesis I have in my head about, kind of like, what is a formula or a recipe to give you a start at potentially having a business-viable AI product going forward. Challenge that, pick it apart. I don’t know, what do you think?

Stuart: Oh man, I thought this is going to be really terrible coming on this podcast because I was a bit scared of your knowledge, but I completely agree. I hate to agree. I’m so sorry.

Brian: Don’t agree. Stop.

Stuart: But what you’ve effectively spoken about there when you started to touch upon determinism with large language models is what myself and Jon Cooke call ‘Large Language Models On Rails.’ There’s absolutely no doubt the fusion of the two, there’s going to have to be that rules-based determinism locked inside—locked against, sorry—probabilistic models. That’s where the power is going to be. You also mentioned small language models. I am a firm believer that the true ability of agents will be highly niched, highly scoped, trained on large language models. Feeding those agents will be trained on proprietary data.

So, it’s kind of—I see a real space—if you want to talk about real business—I see a real space for folk making considerable money from the mid-size companies, not looking at the smaller ones or the bigger ones. Because they’re in a trap, and that trap is this takes something like a CRM. They can go to Salesforce, they can spend an absolute fortune, and it’s going to be too much what they need anyway. Or they can go right down and spend a small amount of money on it, and, you know, it’s not going to do what they need it to. And they’re caught.

They don’t have the money to build their own except for perhaps, now there might be ways to do that. And I’ve actually seen evidence of it, and it was quite staggering, doing this through large language models exactly as you describe, which is that, using their own data, wherever that comes from, bringing all that together, be it structured or unstructured, and being able to present that in a way, like a CRM, that could gather insights from that information that was actually of tangible benefit to the salespeople. They were no longer copying and pasting from here, copying and pasting. And by the way, that’s been the whole problem of why I say digital pollution. Because we’ve got an application for everything, productivity is actually going down as a result of it, ironically because we’ve got so many damn applications.

And that’s the dream, isn’t it, is for generative AI to be that orchestrating panel across the top, bringing that disparate data together along with our own proprietary data. And I’ve seen, as I say, that CRM, it reduced the salespeople’s time—and by the way that CRM was built, I believe, mostly with a tool, an AI tool—it was incredibly modular. It was incredibly easy to configure for any company. It gave the salespeople incredible insights, and they could even set up pipelines to go out and research prospects online and bring that insight back, put it all together, and even from that point, create things, like, very targeted, niche-specific marketing campaigns for those particular prospects.

And it was seeing that you start to see, poof, you know, there is tremendous value here. When we start to combine and surface insights from our unstructured data, our disparate applications, external sources, fusing them together in an intelligent way, not only in this instance did it reduce the sales time spent on these applications by 20%, which meant they could focus in on actual prospects, 20% more in a much more intelligent way, but that had a significant, significant impact on the bottom line, on their revenue, on the things that they care about. And if you can achieve that, suddenly you see a great enthusiasm for the technology.

And I think companies are sick and tired of, oh, how should we do AI? Well, let’s get a data [lead 00:43:33]. No, let’s do it this way. Let’s do a massive transformation, and let’s get a consultant in, and they can have 600 PowerPoints to tell us what to do because we—you know, it’s just not happening anymore. People want to see relatively quick results that matter to them.

And if you can achieve that—and I have seen this—if you can achieve that, you can achieve one or two versions of that, maybe in different departments, that’s absolutely fine, people start to see that this has real, tangible benefits for them in their life and their own revenue, their own income, that’s when you can start to spread organically. Sorry, that was a bit of a sermon, wasn’t it? But anyway, you set me off on one. So.

Brian: That’s okay. But I feel like we—even going back to the conversation about UX, and like, we don’t need design all this kind of stuff, what’s happening is we’re getting faster delivery work.

Stuart: [sigh] ohh…

Brian: Like, delivery work is accelerated. And you can accelerate it, and I think the whole practice of research is going to change. It just, you don’t need to do it the same way. You can accelerate some of the delivery work, but figuring out what should we make and how would we know if we’re creating a benefit or not, these are, like, back to first principle product things. This is where I don’t see a lot of change.

Like, that’s the hard—unfortunately, that’s that gray area, kind of, hard human factors work. It’s knowing the market. It’s some gut. It’s running hypothesis and testing them. And there’s, you know, so you have some of the analytical part of it, there’s some of the product gut that comes from experience and all of that, but mostly what I’m just seeing is faster output delivery.

Like, oh, I can do a strategy now with an AI. Now, I can use, you know, v0.dev, and generate UI. “Give me a dashboard with financials.” [berrup]. It’s like, but is that the right dashboard for the CFO? Is the CFO going to use that or not? Yes, it looks like it has nice fonts and colors, and it has bar charts, and it’s all CSS and HTML. That’s great, but it doesn’t tell you anything about whether or not someone would use that or not.

And that’s all that matters in the actual game of product is, is someone going to use it, and will they use it enough that it creates a benefit that’s worth paying for or exchanging something of value? Like, I’ll let go of my old thing to use this new one because it’s so good. I don’t see that changing too much. But do you see that changing at all? Like—

Stuart: Yeah. This deliver faster, deliver fa—you know, it’s the old adage, isn’t it, that, you know, it doesn’t matter how fast I drive, if I’m going the wrong direction, it doesn’t make a—you know?

Brian: Right. [laugh].

Stuart: It really—and this obsession, I mean, like, I do a lot of mentoring, okay, privately, and I love to do it, and virtually everybody you know is looking to, you know, sort of move up in their career, or perhaps get a new job, something like that. And when I talk to them about their experiences, the reality on the ground of where they are, where they’ve been, it’s always the same. It was, they didn’t have control over all these things that we think they should have control over. Some of them aren’t even allowed to talk to customers because they don’t own the relationship. Somebody else—or, you know, oh no, strategy is just sent down from on high. Or, you know, it was all about, you know, just getting the engineers to push out, push out, push out.

And this knocks on to poor engineers as well, by the way, who feel that, oh my goodness, it’s come to the end of the sprint; I’ve got to demonstrate that, you know, our team did nine of these things. You know, what the hell are these things? Did they matter? Sometimes you think it would be better, you know, let’s concentrate on just wh—firstly, let’s get rid of the 70% of the crap we’ve got in the product because that’s a load of rubbish anyway.

Because, you know, it’s a well known fact—it should be a well known fact—that 80% of the features drive 20% the value anyway. So, we could probably cut 80% out, and that would create more value, and become more niche, and actually solve what we’re supposed to be solving better. I’m on a rant now. But you know, that when I talk to you know, these folk, they’re not doing what you described anyway. So, there’s already a fundamental problem, you know, of this solutionizing and pushing out stuff as quickly as possible.

And you’re absolutely right. It’d be a race to the bottom, unless it can be tied to actually doing this work diligently. And as I say, if folk have not got the experience in doing things like strategy or user research or design or if they’ve not had that real, real time, that ability to do it properly, they don’t have that experience, chops, to be able to understand if the large language model is right or wrong, and that is only going to exacerbate a problem that already exists. So, I’m not going to blame the technology on that, and that is a fundamental misunderstanding of what these roles are, what value they should be delivering, and where they should be positioned in the company, and what they should actually be doing. And that’s something I’ve stopped posting about. I did rant about that a year or two ago, and they became viral in the way you don’t want them to become viral. You know what I mean.

So yeah, you know, if people do listen to this that, you know, they’re going to sense that frustration that was there then and is there now, you know? Because I look at job specs, I’m asked to look at job specs, and you look at them, and you think, this is absolute nonsense. You know, this is all performative, the company is going to get on an interview with the chap I’m mentoring. They’re going to pretend they’re outcome driven, user driven, oh, it’s all about the problem, you know, and the poor interviewee is going to be saying, “That’s exactly what I am. I delivered this value. I delivered that value.” And then the reality hits when they start the job, you know: disempowered delivery jockeys.

I don’t know how we solved that, Brian, genuinely. I don’t. I don’t. Genuinely. And not enough people are talking about it. There you go.

Brian: Yeah. I mean, it’s a leadership culture, right?

Stuart: Yeah.

Brian: Like, I think that the best framing I know is, like, if you’re oriented around outcomes and benefits as the game that’s being played and not just the delivery of stuff on a schedule, then that changes because now just delivering doesn’t count; there has to be some measurement that there was an impact of something when you shipped something. And it’s a harder ga—it’s really hard, but like, it’s really all that actually matters—

Stuart: It is.

Brian: In the end, it’s really all that matters. And there are a lot of people, I think, that got well paid in this space to do mostly delivery work, like, create mock-ups, create a product deck, create a status update for something, generate stuff faster. And for a while that was good, but it’s like the tech is quickly making the delivery part the easy part. Like [laugh].

Stuart: Yeah, absolutely. So, that’s the cruel irony, right there is, as soon as we moved into where money was no longer free in that kind of, you know, growth at all costs rubbish that we were in, and then all of a sudden, the C-suite is looking around at who they can chop, and who they really do believe is bringing value and who isn’t, all of a sudden, UX and product are on the chopping board, you know, because of this perversion of what it should be. And exactly as you say, this is what then makes them ripe for disruption from large language models who can do 80% of this stuff.

So, yeah, it’s a cruel irony. But I don’t blame either product or UX for ending up in this position. I don’t necessarily—I’m not going to apportion blame to anybody. It’s just is what it is, and it needs to change, especially as we’re moving into this new era. And I think it’s going to be—I personally am very optimistic for UX and product in this new era.

Brian: Interesting. Two, kind of, last questions here. There’s been a great conversation, as I expected. First one I wanted to ask is, we didn’t really cover this, but what is your, like, dream team makeup, in terms of roles? Like, if you’re going to go work on a new—especially into the B2B space, some kind of a product that we think is going to take advantage of AI, it’s going to help really make it stand out or disrupt the market that it’s in, what skills and roles do you, like, minimally want to have on your team? Is there a baseline, like, I got to have these, you know, three, four or five skill sets, and that’s, like, your—if I don’t know anything else, I’m going to start with that? Tell me a little bit about what that looks like.

Stuart: Yeah, I can tell you, from experience, who I actually recruit. My first is UX, right? Simple. I need it. I’m rubbish at it, I claim no expertise, I need that. That’s my number one. I need a fantastic data researcher, a data scientist. I really, really do. I need a great data engineer, okay? I need those. And then I need a superb front-end, you know?

And if you can give me that small team, it’s remarkable what you can do. Obviously, you know, you want more data researchers so you can do more; you want more UX so you can do—of course you do. But that core team, you know, can do an incredible amount, you know, with that kind of product head, if I’m allowed to call it that. So yeah, you know, I’ve not included the go-to-market side there. I’m just sort of talking about that core innovation team that can kind of really deal with users, that can really deal with the research side, that really can deal with the pipeline side, you know?

And then I need a great design front-end part to this. And I just find that if you can put those together and you can put a good team together, you can do an incredible amount. That’s—yeah, would you—what about you? What would you tell me your opinion on that one?

Brian: Well, it always depends on the context of—

Stuart: Of course.

Brian: —the project, but like, my, you know, the typical triad: so you’re going to have a user experience lead, a technical lead, and a product lead. Now, a technical lead, that can vary based on what kind of product that is, and sometimes my general feeling is you might have engineering and data science—particularly if you’re building a product that’s going to be powered by AI—you may need separate disciplines because one technical person may not cover all of that range there. I prefer smaller teams. It doesn’t mean you don’t need data engineering and some of that, but I kind of separate some of the strategic roles from the delivery roles, so I tend to think of data engineering as being a little bit more like, once we know what to build, then that stuff matters more, but the first thing is like, do we have anything viable here or not? When we’re in the ideation phase, we don’t know if this is going to work, it’s a different skill set, so I tend to not think about data engineering as being required in that kind of what I call inner ring group. That’s typically how I tend to think about it.

And then obviously there’s subject matter experts. Sometimes you need a separate domain. If you don’t have, like, a data scientist who also knows the medical domain, for example, then you may need an SME in there, especially if the product manager doesn’t have that. Those can be really invaluable.

And I did two—okay, I lied. I have two more questions. I want to give you the last word about anything that you’d like to share to the audience, but the last thing was, you had talked on our pre-call about needing to work with academics and data scientists, especially when you’re building AI-powered products. Tell me a little bit about what that looks like. Like, how much technol—like, PMs have always had to work with engineering, and you need to know, basically, how systems work, you need to know what an API is, you need to know what data is available in a system.

You know there’s a fair amount you can know without actually writing code and doing implementation to understand why something, quote, “Might be hard,” or why it might take months instead of weeks. What’s different or similar in terms of the data science aspect, or an AI researcher, for example? What do those conversations look like?

Stuart: Yeah, when I’m talking about very academic people that spend an awful lot of time with their heads buried in research papers, and have written them themselves, they’re tremendously good fun. They need—[laugh] they just always seem to be a certain way, which I kind of touched on earlier, which is this academic rigor which can be hard to get past. I don’t blame them. But yeah, from their perspective, it’s very hard to get straight answers from them, and you can understand why because they’re looking at—and it’s very difficult for them. Let me—because, you know, tomorrow’s paper can undo what they were proposing yesterday as an approach, and it can be really, really difficult for them. And I do admire them.

I’ve been very, very fortunate. The ones that I’ve worked with have also had this strange ability to be able to skin up rough POCs in about two-and-a-half seconds. I don’t know why they’ve had that ability. I don’t know if it’s just, you know, very fortunate, sort of, encounters with them, but that is tremendous fun. But as I say, you know, fundamentally, they are drawn more to the cautious side, and you might, from the product side, be gently trying to coax them towards a sort of braver, let’s go for it and see what happens side, which is quite alien to them.

And equally, from the product side, you have to learn patience and the fact that you cannot have that degree of confidence with your timeline that you might have been able to present to stakeholders with just an engineering timeline. And that has to be communicated over and over and over again. And also, you know, you want to keep your engineers occupied not in trivia, if there’s a sort of stall on the research side.

And then you have the opposite. When research makes a breakthrough, and it’s this sort of panic on poor engineering to come on, come on, quickly. We want to put this out of the world. But that, to me, is the essence of, kind of, what that product role should have been anyway, which is that kind of negotiator, that peace keeper, that calmer, that encourager, that product, that, you know, I mean all that in a very, very sweet way, by the way, without an ounce of criticism on anybody. It’s just we need these different disciplines to make these things happen, but it inculcates a different kind of mindset, and so you have to learn to negotiate those, and work with them, and bring them together as one team. But magic is there, if you can.

Brian: It’s a huge relationship job, and I think, like, managing up to business stakeholders, and then changing your language to work with a researcher about what they find valuable, and honoring the work and domain knowledge they have, like, you do have to change. You have to speak all these different languages. And again, this gets into that messy part of what product management is about, it’s like wearing all these different hats, just enough to have productive—to move things forward [laugh].

I love it. I think it’s a fun space to be in. It’s not for everybody. But, Stuart, this has been a fantastic conversation. I wanted to give you the last word. Where can people find you? And secondly, do you have any final words or just advice or anything for, like, the data and AI product management community that’s listening to the show? Anything you’d like to share, just closing thoughts?

Stuart: I don’t know. I don’t feel I have anything important to communicate specifically to the data and product, but most important thing is that I’m a LinkedIn addict. It’s a terrible weakness. If I get a thought, I can’t help but post it. I don’t have an actual social life, so that’s probably the reason. I actually met some fantastic people—including you, Brian—on there, so I’m very grateful. So, that’s where to find me.

You can’t go wrong. As you said right at the beginning, there is no other Stuart Winter-Tear, so it’s not going to be a problem to find me. And truthfully, I just want to say about a—I do post sometimes, several times a day. I get up—I’ve become like my parents; I get up at, sort of, four o’clock in the morning now to do the research, but I don’t always get it right.

Do you know what? I was asked to do something, and I thought, “Oh, I did a post about that about a year ago.” And I was looking back and I cringed, Brian, at some of those posts from a year ago. They looked outdated, and wrong, but hey, how else do we learn? This is nascent, you know? We’re all figuring this out, so we have to give each other a very large degree of forgiveness and understanding, you know, as we kind of feel our way forward with this technology.

So yeah, just go easy on yourself. Be curious. Be curious. And I would say to product folk, you know, do use the technology. It’s a really cool way to get to know it. Don’t be frightened of it. And if it grabs you, and you feel like, you know, this is something you want to sort of build, you know, learn the language of the data engineers, of the data scientists. They won’t expect you to be at their level. Man, they went to university for goodness knows how many years to do that, but you know they appreciate the fact that you’ve made the effort [laugh].

Brian: [laugh]. Well, awesome. Thank you for closing with that. Stuart Winter-Tear, Chief Product Officer and Partner at Artimus. It’s been great talking to you. People listening to the show, LinkedIn is definitely the place to find Stuart. He’s always—I don’t know how you read so many articles and white papers. I’m just like, dude, when does this guy sleep? Like, [laugh] I’m like, “You read another, like, paper with Greek symbols in it?” Like… [laugh]. All the—

Stuart: I know. I learn so much from them. I can’t—

Brian: Data people are laughing at us, but that’s okay.

Stuart: Yeah. I love them. I find they’re excellent writers. I don’t know why, but there you go.

Brian: [laugh]. Excellent. Well, thank you so much for coming on, Experiencing Data.

Stuart: Lovely. Thank you for having me, Brian, genuinely.