top of page
Search

Purposeful AI & Cultural Intelligence | Ian Cook of Qloo | The Brainiac Blueprint Podcast

Writer: Kyle Lambert
Kyle Lambert
Apr 5
42 min read

In this episode of The Brainiac Blueprint Podcast, Kyle Lambert sits down with Ian Cook, SVP of AI Technologies at Qloo, to explore what it really means to use AI with intention.


Ian breaks down why he believes “AI is overhyped but underappreciated,” how Qloo uses cultural data to power hyperpersonalization, and why experimentation matters more than chasing hype. From knowledge graphs and recommendation systems to adoption challenges and rapid prototyping, this conversation dives deep into practical, human centered AI.


Full transcript below. 


🎧 Watch or listen to The Brainiac Blueprint Podcast: 

Apple Connect: https://bit.ly/3VMNtlH 


⏱ In this episode, we discuss:

00:00 | Intro

01:30 | What Qloo does + cultural intelligence explained

07:15 | “I think AI is”

09:30 | Why AI is the next major life shift

21:08 | How Ian decides when (and when not) to use AI

22:00 | Experimentation mindset + low cost automation

35:15 | How Qloo drives personalization for real world applications

50:03 | Rapid fire questions

57:43 | Open forum + Fierce Highways + closing


🔗 Ian Cook


🔗 Qloo

Website →  https://qloo.com


📲 Connect with Left Brain AI



📣 Subscribe & Share

 If this episode inspired you, taught you something new, or gave you a different lens on AI, personalization, or experimentation,, share it, leave a comment, or tag us. 


Let’s help more people stay brilliant.


Know someone who would make a great guest?


Episode Full Transcript:


Kyle: Welcome back to another episode of The Brainiac Blueprint where we discuss the intersection of AI and how it impacts business and the world around us with our esteemed guests.


I'm your host, Kyle Lambert, founder of LeftBrain AI and Action Hero Marketing. In today's episode, we're going to discuss driving outcomes and being purposeful with AI.


With that being said, today's Brainiac is Ian Cook. Welcome to the show, Ian.


Ian: Thanks so much for having me, Kyle. It's great to be here.


Kyle: Awesome. Yeah, I know we were just connecting. We're ready to go, starting 2026 off with a positive note here and ready to share some of your exciting insights.


Ian: Yeah, this is my first step back into anything in 2026 that requires me to engage about work. So it's nice to get this chance to have a discussion and just chat about it before I have to show up on Monday morning and feel kind of…


Kyle: We'll get the wheels turning. I'm on my second one of these energy drinks to kick myself back into gear with you here.


So here we go. Ian, you are SVP of AI Technologies at Qloo. You do a lot of other things involving AI. I know you've had some projects that you work on yourself. You're involved in Pittsburgh at AI at Work. You have a lot of different things going on in terms of AI.


So if you don't mind, just tell everybody, you know, who you are, what you work on, all the cool stuff behind the scenes you got going on.


Ian: Absolutely. So the thing that pays my bills every day is working at this fantastic company called Qloo. Q-L-O-O, for anybody who wants to look it up. It's not going to be spelled the way the board game is.


I'm the SVP of AI, which generally means I help take on a lot of the AI-forward projects, helping some of the other developers build, refine, evaluate models, find new uses and applications of our technology within AI and AI tools.


What is that technology? I think of ourselves as having the knowledge graph for social and cultural data. So you want to go out to a restaurant, a hotel, go to a new town, find a new book. We are sort of the way that companies do hyperpersonalization.


So it's a little bit like recommendation, but understanding that you like certain things or that you've interacted with certain things means we can help figure out what else you might like.


So we help companies do all of that personalization to say, you know, you're a big fan of this kind of hotel. Have you tried these hotels in these other cities, or more close to home, book recommendations or influencers or movies, and putting all of that together.


So behind the scenes, Qloo has spent the past nearly 14 years putting together this massive knowledge graph on all of this data across all of these cultural topics. So we can help explain why people who are in a particular neighborhood are shopping in a particular retailer or going to a particular restaurant. Or you can go the other direction and say, "Hey, in this neighborhood for people who are, you know, say my age, men of a certain age who happen to be, you know, fathers and have dogs, what are they looking at in terms of the stuff that they buy?" And help retailers find great places to put stores or brands and how to connect with us, or even online folks, and how to partner up with brands that we're interested in.

That's the day-to-day work. But, you know, I've been doing AI and machine learning–related things for 15 years now. So I have some of my own projects and other companies I've worked with.


There's a fantastic, if you're at all an angler, there's an amazing app called OnWater Fish that I work with to help their AI efforts. Specifically things like measuring fish through the camera and going out and recording your catch.

Now, of course, most anglers are going to want to say they had a personal best and don't, "No, I happen to not get a picture. I didn't, I forgot to measure it." But it was at least…



Kyle: Of course. Of course.


Ian: But this one, if you want to take a picture, capture the length and weight, we do all that with AI. It's fantastic.


And then the last one you mentioned is this really great community we've got going here in Pittsburgh called AI at Work. We are a kind of small, select group of folks who are doing work with AI every day. Some of us are a little bit more technically focused and putting things to use and building models, but there are folks who are, you know, doing this in their design work, in their marketing work, in their legal operations work in tons of different places.


We come together, talk about what's working, not working, our new advances. One of the active channels in our Slack is "news about what was just released." You know, DeepSeek just released some amazing new research, so we're all probably going to scramble and chat about that.


But we get together. One of the things I like about it is it's in real life, which means I get to see other faces. My last several jobs have been all remote, which is fantastic for work-life balance. But you start wanting to see some people up close and personal, and it's just been… It's called AI at Work, and people can find that online as well. Though if you're not in Pittsburgh, it's not going to be as helpful, but you should reach out if you're interested in the kind of thing that we do.


Kyle: Very cool. Very cool. And I feel you there. I've been doing a lot of research for next year of finding ways to get out and be around like-minded people just to kind of talk about even just, you know, entrepreneurship or whatever it is.

It's too easy to just lock yourself away, especially, you know, we're in cold areas now, so it's like, I don't want to leave the house.


Ian: Well, one of the things we've tried, and this has been a kind of a very recent addition, but what I liked a whole lot is we found a local co-working place. And what we said was, like, a whole bunch of people from AI at Work just came and worked at the co-working place kind of together.


We're all together, working on our own things, totally different companies. Some of us are founders working on our own companies, some of us are employees. But we had just a really nice time kind of sitting and chatting and talking with people. But then, you know, heads-down time and getting lunch with folks.

So I highly recommend it. Like, even if you just have to assemble a small group yourself and just say, "Hey, you know, I know you're not doing the exact same thing each day, but there's a similarity there and we all are, you know, like-minded folks." It's been a great thing. So we're going to continue that into 2026.


Kyle: Very cool. I think most people would agree your environment strongly dictates your, you know, your mood, your output. And so just being around people like that, that are productive and grinding away is contagious. It makes you want to, you know, I don't like to use the term one-up them, but it makes you want to at least kind of be like, "All right, I'm grinding away. I'm getting stuff done. They are too. We're kind of going in lockstep here."


Ian: Yeah. You want to perform at the level of people around you. That's great.


Kyle: Exactly. Exactly.


Well, cool. Lots for us to dive into. But as you know, I like to have everybody finish the prompt, "I think AI is…" So we've reached that portion. I'm going to pass it on to you and see what you think about AI.


Ian: All right. I admit I took a little time with this one. I kept thinking, "Oh, it needs to be something completely, not absurd, but totally out of left field and nothing that people would expect."


But what it came down to was I think AI is overhyped but underappreciated.

I am from a tradition that is much more like scientifically rigorous and research-minded. So I did a whole PhD a few years ago, but don't hold that against me.

And I think so much of what we talk about with AI right now is all about, "Well, it's going to have human-level abilities or superhuman-level abilities and take everybody's job and it's the most amazing piece of magic to come along." And I think what that ends up doing is scaring people away from it and they end up not appreciating that this is an amazing technology.


We'll say it's definitely the closest thing to magic I've seen in my career. But it's not magic. And I think focusing on this idea of reaching AGI, which I happen to not believe in at all, is totally distracting people from the fact that this is a force multiplier for whatever you're doing.


And it is going to be a huge help, and it's going to seep its way into everybody's work. But we're seeing right now that tons of people aren't taking this up. They, you know, open up ChatGPT, they just see a blank page and they're like, "Well, what's this supposed to do for me again?"


And they hear about it, but they don't have the opportunity to work with it the way that would reveal to them the best stuff about it. So that's my long version of the simple answer. I think AI is overhyped and underappreciated.


Kyle: You know, I love that answer. I think regardless of what we're talking about, anything that generates change or disruption is automatically going to bring fear, right? That's just, you know, for some people anyway. Others, you know, I think the two of us included, kind of embrace that type of change.


I was actually just having a conversation the other day, and this just resonated with me from your answer. I'm 38. I was born in 1987 and I think there's three instances in my life that we've had, there's a clear pre and post, like my life has changed from these things.


One is 9/11. Two is the smartphone, and three is COVID. And I think in five years, ten years, I'm going to be saying number four is AI. Our life is going to be drastically different before and after.


I choose to think about the underappreciated, not scary portion of it, thinking that hopefully it's going to, you know, make it for the best for us. So I think that'll be number four in my list very soon.


Ian: That is a great list. I definitely think that, you know, 2025 for me, since a good portion of my day is coding, or was coding, and I do, you know, being a senior leader, a lot of times I do things like mentorship and team alignment and large-scale planning. But, you know, we're a scrappy startup, which means if I want to see some prototype, I'm going to build it myself and see if it works.


The change in the cost in what it takes to generate code from the beginning of 2025 to the end of 2025 is just such a dramatic shift in this particular industry. And, you know, it's why it's focused on coding for good reason, but I think there are going to be other industries that see similar things.


It simply is too powerful to just not use and to think, "Well, this is just something that, you know, is annoying and it's going to pass." And I don't think anybody thinks AI is a fad, but they, you know, think, "Well, it's just mostly for the technical folks and it's just going to be, you know, people writing code and that's where it's going to stay."


Kyle: You know, you mentioned that you're a small, scrappy startup too. I think it resonates with me as well that, you know, AI in a way is an equalizer, right? We don't have the resources that - of course, I'm choosing Amazon and them because they're the biggest, but like, you know, even take Fortune 5000 businesses, whatever it is. They have employees and they have resources and, you know, we don't necessarily have that. We have to be scrappy, and having these tools gives us the ability to have the output of multiple people or multiple teams. And it's definitely leveling or equalizing, if you will, the game a little bit for us little guys.


Ian: I would definitely agree. And, you know, with the ethos of startups, one of the things that I've always been, you know, a fan of and pushing is like, fail fast, right?

Like, that's a common term. It's kind of a cliché. But the speed at which you can now test something, try it, see what failed, see what worked, and move on is blinding compared to what it used to be.


So that, I mean, startup notion is "Look, get something out, get response to it, figure out if it's working, fix what isn't." That does not fundamentally change. It just goes a lot faster. That speed of fuel can move so much faster and with fewer people at this point.


Now, my hope is that that means more people try to start their own companies and start their own ideas. That's not really the way technology ends up playing out. Once there's something disruptive, not everybody just immediately turns into, you know, entrepreneurs.


The movie industry is definitely one of these examples that, like, you know, making short videos now is, compared to what it used to be, nearly costless and I don't have to be a cinematographer. I don't have to… but it's not like everybody just turned into a movie maker and suddenly, you know, movies are flourishing.

So we do have some real concerns, and we got to figure out real ways that we are going to apply this and be careful about what it means when folks look at this and see their future changing fairly drastically.


So I certainly appreciate that these things can feel disruptive in a negative way. But it's one of the reasons I always push is, "Well, be the person who knows how it's going to work and how it's going to disrupt what you do."


Kyle: I love that mindset.


All right, let's jump back to the beginning a little bit here, a little origin story here. So, I know that you have kind of been in the stats and, I think, economic world, or econometrics, I think is the word that you used. Yeah?


Ian: Yeah. So years ago I worked for an organization called the RAND Corporation, which is this fantastic think tank for lack of a better term. It is an interesting blend of both like pure research and applied management consulting, where it will take time to research big questions but then work with its customers to say, "What does this really mean for what you do?" And I did this largely for the Department of Defense.


But a lot of my work doing that was statistical. It was gathering data, reviewing the data, coming to estimates of cost, projecting spend out into the future, and looking at budgets. So lots of statistics, lots of, you know, for the economists in the room, they get, they often learn it as econometrics, but essentially data and modeling. So extrapolation from incomplete data.


But I ended up deciding that I wanted to go even further, so I enrolled and got a PhD actually in political science. And I got a real interest in language as text.

So this is the place I decided, you know, "We have massive amounts of information about the American political system sitting in text documents, congressional testimony, speeches," and I wanted to use that as text.


And at the time, natural language processing and natural language programming was becoming a tangible thing to do. This was back in the early days of not even BERT or the early models, but things like word2vec, doc2vec. So just understanding the very early stages of "the words that are around a word help give it meaning," and can we do something statistically with that when we count it up and figure out what's going on?


So I started there, taught myself Python, did my dissertation on communication using a lot of that knowledge and a lot of that skill set, and then I ended up following the path of startups.


You know, both the startup or both the academic and market looked at each other askance. They weren't sure what to do with me, and I wasn't sure that it was the best fit for me. And I got a chance to head into a startup as a machine learning data scientist kind of person and fell in love with it and have never looked back.



Kyle: Very cool. So you've been in the game for a while now.


You know, when you're talking about Qloo in terms of personalization and everything, that, as a marketer, relates directly to me and makes, you know, your typical audience member or customer is going to be in the marketing department. Correct me if I'm making a false assumption there.


But I'm curious from your perspective, I know this might be a little bit of a difficult question, but over the years, where do you think machine learning and AI has been the most impactful and the most game-changing in maybe a specific department or a specific role or, you know, specific task? Have you seen it be like, "Oh, wow, we are just done doing it this way altogether now because we have this"? Like, has there been a moment like that?


Ian: I would say most financial services have been, you know, rocked with the ability to do such detailed forecasting and estimation and analysis and root-cause analysis and just deep things about how and why money is moving and changing.


But, you know, financial services, banking industry, the credit industry, all of those things have gone from very specialized people who take a long time to come up with some very heavy, thick report, to the world of dashboards and quick reports and then being able to model and predict.


And now we're getting into this world where I can, you know, just go to an LLM and ask it to think through a couple of things, and it'll have access to maybe some W2s and some datasets out in the open web and grab an analysis and say, "All right, you've got a thesis about what's going to happen. Here's some data pro and con," and start thinking through it like, truly, a thinking partner in another one.


And maybe I can get you to verify this. I think probably marketing and advertising, you know, once the shift came from the typical ad buying into things like going through Facebook and AdWords. All of that, even if the people who aren't themselves all that mathematically interested, are effectively relying on a massive machine-learning statistical system in the background to understand who they're targeting and how to get there.


And it just wasn't possible to do it that way until we had these models and had compute at this size. And now we can do it, you know, at such a fine level on such small things that, you know, think people like Facebook just learning constantly what's coming through the feed and what's being interacted with, so that their definition of a person that you can target is so fine-grain that it's a little bit disconcerting to some, but also just this explosion in knowledge that I don't think would have happened without the growth and speed of ML.


Kyle: I 100% agree from the marketing standpoint.


I remember, maybe it was like 2017 or something like that, running Google Ads campaigns and it would be manual bidding per keyword. And I would go in there and just kind of sort it and be like, "Where am I overspending?" and just try to like lower it, still kind of in bulk. But it was like, you know, just like a large spreadsheet.

And now just like flip on this switch, have the algorithm, you know, give it a couple guidelines or, you know, some settings and stuff like that, and it just does it all for you. It's like, "Oh, wow."


I mean, and again, that was 2017, something like that. You know, nowadays that's fully built into everything that Google and Meta does, like you mentioned. And I have my own tools that I'm doing on the side. So it's completely changed everything. I love your reference about financial analysis too.


I even was just reading some articles about somebody who took all of their own bank statements and credit card statements and retirement files and uploaded it into NotebookLM and turned that into like a financial planner for them. And like it was like, "You're overspending here and you should be thinking about this." It's crazy, you know, just from a personal standpoint with one free tool, you upload a couple docs and you're right, it's like it's just analyzing everything for you and telling you what you should and shouldn't be doing. It's crazy.


Ian: Well, for so long, the gravity for the super-quantitative, high-end folks who - I mean literally like particle physics, quantum physics kinds of people - was hedge funds because they wanted to get into high-speed trading, high-frequency trading - well, I guess that's the same thing - test out new methods. When I was learning a lot of the statistics, that's where a lot of people were aiming. It was like, "Well, I'm going through this PhD, but my, sure, it can be in economics or statistics, but, you know," they're like really looking for the hedge fund at the other end of finishing that PhD.


Whereas now, all of those folks are going straight to the major labs for LLMs and AI. So it's clearly the impact was felt in financial services and the financial markets, and now we're, you know, democratizing that a little bit because LLMs can be used by anybody.


But yeah, you start getting those folks, it's like, "All right, here's my entire tax return. What did I do wrong? What can I maximize for this year?" Yeah, I hear about plenty of people doing that.


Kyle: Oh yeah, 100%. So I think this is a good segue into the main topic that I wanted to discuss with you, which is deciding how or when to use AI.


Again, kind of our initial conversation and your background shows, I'm sure you've dealt with failures in terms of it not working the way that you intended it to, right? These are new technologies, these are new tools. And so I'm curious, from everything that you've learned, do you have a process? Is it a mindset? What is it like to say, you know, how do you determine, okay, this is where we should be putting AI next? Or, you know, this should be staying the way it is, but we need to disrupt here?


Like, what is your thought process when you're introducing, whether, again, Qloo, any of these projects you're working on, how do you think, "Okay, yep, this is where we go next with AI"?


Ian: It's a really good question. I suppose I should become much more systematic about it. So I think maybe the answer is a bit of a mindset. And a lot of this, and I'll just keep referring back to sort of my academic training, is experimentation.

I think, or I find, that it is almost costless right now to try something with AI and see, "Did that help me? Did that speed things up?" Or maybe I just spent, like, way more time trying to get the LLM to understand what I was doing and I could have just done it myself.


So, when I look across the types of things we have internally at Qloo - so think about internal operations, meaning the number of people who are using code: our, you know, Chief Financial Officer or our marketing folks - I start with the things that they do repeatedly every day that might take a couple of extra steps.

I look for anything that is something that you want the end result for, but it takes you several steps to keep doing it. That's always a target of automation, but a lot of these departments don't have people who are going to be like, "You know what, I'm going to write myself a little Python script to go out and grab these things," and do… like, they just, that's not the way they're often wired. You know, different skill sets, different mindsets.


But I think of that as a great first step. It's like, all right, if there's something that could be automated, the cost of automation these days with LLMs is so low, let's try it there.


Or anytime… So that's one area. Or anytime there is a green field in front of you - "I am thinking about new ideas" - these things become great ways to test that out. "Here's an idea. Come back at me with reasons this isn't going to work, ways it could work, what are the counterarguments."


What I want to see are people who are engaging and thinking through this rather than just trying to have, like, somebody totally outsource and forget about what it is.


The real risk is, like, giving up the thinking is the worst part. It's like, right, we still have taste. We still have our understanding and we have the justification.

Right now, AI can't do "why it does how" incredibly well. But why are you looking at certain ads? Why are you even starting this research? You've got some problem that you need solved, and AI is there to help you do that, but it had no reason, or had no understanding that that problem existed.


So I look for that, and then I get people to try it. Keep using it a couple of times, see if it helps. If there isn't really an improvement, that's fine. Maybe there are other things that you can start looking through.


For code, I find you don't have to convince a whole lot of people. For a little while there, I would say mid–early/late spring to summer 2025, there were still a bunch of people that were still on the whole, like, "Look, LLMs are fun for, you know, helping out for tab completion and coding, but it can't write, you know, big amounts of code." And, you know, I have released plenty of LLM-generated code into production environments, and it works fantastically well.


So coders, I don't find you have to convince as much. But the reason I go back to experimentation is, bigger companies will want to say, "Let's have an AI initiative," and we're going to put one person in charge, and that person's going to oversee AI initiative. And they have to go out and force people to, like say, "You're going to try this. We're going to totally revamp that."


I want it to go the other way. I want people who know their work intimately to say, "This is something I do all the time, and it takes up my time and it's, you know, something that I don't think about. I just have to keep doing it. Maybe I should try AI for that."


Terrific. Tell me about it. You know, bigger companies will have some guidelines. You know, you've got telemetry on some of these tools that you want to turn off.

One big tip that I tell a lot of people who are concerned about it is, when you enter into enterprise deals with these big companies - OpenAI, Anthropic, and like other companies - they will often offer deals on things like not saving the data for training, not keeping the data cached for as long. So there's a whole lot of concern about like, "Well, you're just giving away all our IP to this company to train on." You can get around that if you've got some enterprise deals.


So one of the things that you might start with is, yeah, it's a little more expensive than just the 20 bucks a month, but try that out because that keeps the protection. But give the tool to a bunch of people and see how they use it. Where do they find the value?


Because I think you'll get a lot more buy-in when people say, "Oh, I found a way to use this that's really helping me. Terrific. Now let's go across a group and say, were there a lot of commonalities in how people used it? Great. Now everybody's using it in the marketing team for doing some initial research or for tracking."

You know, you can connect these things to your CRM. So maybe you've got your version of Claude hooked up to Salesforce, and you just want to go, like, "Give me the reminder about the last several meetings I had with this client," and it'll be like, "Hey, you know, I put in notes. They were a little hesitant. They didn't like this. They were really interested in that."


Then you'll find the marketing team or the customer success team starts finding uses for these that are not just code. So I very much like seeing that experimentation and have the information bubble up and then just say like, "All right, great. Now that the people who need to be using it have told us the best way to use it, let's make it more routine. Let's make it part of our operating principles."


Kyle: Interesting. I like that thought process.


I, you know, I think it's, again, you can always have your hypothesis and it could go in a completely different direction. It doesn't mean that if it was a failure, that was wrong, but it's good to have that idea and to experiment.


And I like what you were saying about also that it's good at the "how" and not necessarily the "why." And I think that's an interesting point. And I think that's probably something you have to deal with internally as well when you're discussing, "We should be trying this tool or experimenting with it and whatnot."

So I'm curious how you approach that from an implementation standpoint. We're all slaves to metrics, right? So, you know, it's not necessarily easy to say, you know, especially qualitatively, that this is, you know, qualitatively better. Or, you know, I guess you could say, "Hey, this used to take me five hours. Now it's taking me one."

But I'm curious how you think about these metrics, especially from a qualitative standpoint. How do you decide this was, you know, the right move? "Let's keep going," or, "Let's iterate." What is, you know, what metrics are you building in? How are you figuring that out? What I'm thinking about, like, that whole mindset there of deciding how to continue.


Ian: Yeah, I feel you on being led by metrics. A lot of times there's just, you know, there's all this push and like, "Okay, what are our OKRs and our KPIs and the monthly metrics?" Especially when you start getting into really big engineering organizations that follow some strict things about, you know, if you're familiar with Agile, Agile will have things about story points and developer velocity and all of these things and all these numbers.


That's fine. I can't second-guess why a certain company picked a certain set of numbers. The thing I've always reinforced is, when these get put in place, I am less concerned with the absolute value of that number - "Like, what is it? We got, you know, X number of these things," or "We got a certain number of projects done."


When you are trying new initiatives and you're trying to optimize this process and say, "Look, let's give our entire legal team access to AI because, you know, we just don't want them spending all this time creating boilerplate that, you know, sure, we have some, you know, a library somewhere that has boilerplate, but then you got to go through it and, like, fix it, like there ought to be help for that."


They can't avoid reading it entirely. But like, let's give them the thing that would have been a couple of hours of just nitpicking. Let's give that a really good run-through and just say, like, "You know what, the legal team can use it right off the bat."

What I want to see is a change in the rate of the metrics and how they're improving. Maybe they've been flat for a while and you figure like you're doing okay, but now they're going up a little bit, or this time spent is going down a little bit.


Possibly the nerdiest thing I'll say today is there's a rate of change in data, and you can think of it as a derivative. But what you see is, this number changes a certain amount. Is it changing faster? Is it improving the way you want to see it?

So over time you've gotten, again, I default to a lot of developer things: you can complete a number of story points within a sprint. Can you get more story points, and is that increasing?


You've decided from, you know, years ago probably, that story points is how you're going to measure developer output. Although I will always caution against using that as a measure of developer quality, but there's just some measure of output - points or the number of tickets.


Are you seeing that start to tick up? Are you seeing the rate at which you're completing things increase?


So you don't have to be a slave to the idea of like, "Well, what I need to go is from 50 to 150 in one quarter this year because AI will triple what we're doing." It's going to take time to learn any new process, to learn any new set of tools.


Can I see a steady increase in how things are doing? Because the other part of this is one of the things we don't measure, and the hidden part behind this, is: are you liking what you're doing? Are you… like, you're crushed by that metric but maybe there's a bunch of things that you're doing that you enjoy more now and you're increasing your speed, and you're finding more satisfaction in it. There's higher-quality output.


Measuring things against defect rates and bugs in code, or, over time, your turnover rate - are you able to keep people who are terrific? So one of the promises of AI, at least in coding, is like, "Well, we can do more with a smaller team." That's fine, but maybe what you can do also is grow the amount of work with the same size. So you're not laying people off, but you're actually completing more work.


So start looking at how those numbers change once you take on the new process. You're not going to see, you know, a tripling in a week because you suddenly have an OpenAI account, but understand that this affects a lot of things that go into just how people work during the day that you'll see dispersed across your metrics.


Kyle: That's such a smart way to think about it.


I think a lot of what we're introducing AI into and what the metrics are that are being used to determine success in a lot of ways is a leading indicator versus a lagging indicator, right?


To your point, you can get more done, or even just, again, this is not necessarily measurable, but "I have more energy and passion to do the things that I want to do," which then is going to lead me to closing more deals and then getting more revenue.


So, you know, you're going to see revenue maybe going up over here, and maybe just a 2% shrinkage in this time used over here, but it contributes, right? It's leading to the end goal of what you're trying to get to. So I think it's a really smart way to think about it.


Harder to necessarily get buy-in or everybody to be thinking that way, so you need a full organizational, you know, push about that. So you have to have the right culture.


I know you're working at an AI-based company, so I'm sure there is some of that built in. But I'm sure it's hard to be thinking about, "Okay, this month I'm going to be doing less time on this, so that's going to give me more revenue in six months" type of thing.


Ian: And this goes back to a little bit of what I was saying before in terms of bubbling up from the bottom. If everything starts with like, "All right, we'll pick the metrics, we'll pin one person as the AI initiative person, and then you've got to go out and hit some numbers," I think you're going to miss the sense that those individuals are getting something from it. They are liking what they're doing. There's, like you said, an increase in energy.


And then you can start seeing that in numbers like, "All right, it only takes this long to prepare for a case. It only took this long to complete some bigger tickets on our list," or, "We got feedback on a marketing campaign much more quickly," or, "I've got a different way to measure this."


And, you know, you start seeing those types of benefits because people will come to you and say, "Here's something that I did that I can do in a really different way and get some better value out of it." Whereas if it's all just top-down…


Again, I've forgotten the quote and who said it - probably Peter Drucker, because everything goes back to him - is something about, you know, "Once the metric becomes a target, it's no longer a good metric." Because everybody's just gaming the system to hit that number.



Kyle: That's so smart.


I'm curious how that translates to everything that you're doing with Qloo and Taste AI because, again, I'm top-level, you know, we've connected and I've gone through Qloo, but I haven't necessarily seen the wheels turning.


But I have to assume there's a little bit of like, "Okay, we have this data, we made this connection, we're changing our marketing strategy, we're maybe adjusting all of our assets" and some of that. And so there's going to be a lag in terms of maybe the revenue.


And so then it's like, "Okay, we're getting a better click-through rate," and you kind of see that.


So I'm curious as to what that process is like with Qloo - how you see the impact kind of trickle from that bottom up even though, again, it could take a little bit of time.


Ian: Yeah, certainly.


So Qloo, to use just an MBA kind of term, is what I would think of as B2B2C. So our customers are often companies that are using the results of our data to do personalization in their own apps.


So large financial institutions that want to connect you when you open up your app for your bank or for your credit card or for whatever and it says like, "Hey, here's an offer," instead of it just being a random, like, repeat, "You just bought a refrigerator. Here's 10% off at Home Depot for a refrigerator," where it's like, "If I bought one, am I really going to go out and buy another one?"


We're personalizing. We've seen these types of transactions, and we think you're the kind of person that would really like, you know, an extra 25% off at Vuori, or here's a, you know, special discount at a hotel.


For us, one of the biggest things is not just the quality of the data but the quality of those connections. So seeing from customers that they say like, "Yeah, people are taking up this interest, and it is having an impact on some behavior."

One of our customers is one of the world's largest sort of dating app systems. So our data is helping to drive whether people match and then continue to have a conversation.


So if we can go from somebody's got an idea about a way to make that better - like, "All right, great, we need to make the match model do a certain model, improve one of our models internally" - they're going to take that on, they're going to work on it, push that out.


The client says like, "Look, this is great. These are matches that we're seeing that are really interesting. Here's, you know…" They'll often share data with us to analyze and say like, "Oh look, this is really, truly improving the quality of the matches, the quality of the recommendations."


So if we can go from an idea to fix - or not fix, but like improve - a model in some domain, to that model being fielded in customer tooling - so, you know, essentially what we sell is an API, but that means you get access to all of this data because we have the knowledge graph.


So to go from an idea to live in the API faster is all the difference in the world.

And we can go from, you know, people saying, "Oh, you know what would be great is if you were paying attention to podcasts like this, so that we can help, you know, understand the world of podcasts for people who are interested in AI and are doing a certain kind of work," great. We can get that out really quickly. We have to figure out what data to grab, how to analyze it, how to work it into our knowledge graph, and then out to a client who can then put in, like, "I'm interested in The Brainiac Blueprint podcast. What other ones would I be interested in?" almost immediately.


That is a huge win. So that is how we track the impact for us. And we have a number of initiatives that we have taken on internally to test that out and use AI and LLM to speed those things up - everything from eval to model types. So it's been a great process seeing people just sort of pick that up and run with it.


Kyle: As a marketer, just like I… It's exciting to hear about these kind of things, and I certainly understand the power of just having a good dataset and being able to make some connections and some correlations for strategic adjustments.

Based on what you said, it sounds like your model, or Qloo's model, is strictly on third-party data acquisition, or do you guys have any of your own datasets that you're kind of building and managing or anything like that?


Ian: We are omnivorous.


We do. We have our own data collection processes. So we do a lot of social listening. We do a lot of information gathering on our own. We do first-party data collection.


So, you know, companies that are… So we'll, you know, get an API deal in a company that does reviews. You can think of book review sites, movie review sites, place review sites. We'll go directly to them, bring that data together.


We'll do our own social listening. We will do third-party datasets. So somebody's got this great dataset on maybe some physical locations and information about buildings that we might not have had before, and we will put all that together.

The interesting part, and what makes our data ours, is building this knowledge graph and understanding what properties are of interest out of that entity.


So if you come to us and just say like, well, YoungBlood is the big new hotness in music. Great. How do we break that down? It's not that we, you know, have data on YoungBlood and his Instagram or Twitter or whatever it's going to be. It's not that we just have like, "Oh great, here's how you find him on Twitter or X, pardon me, or on Instagram." It's saying, "Here are the properties. Here's the type of music he plays. Here are the venues he has played at. Here are characteristics of the audience that he tends to bring in. Here are maybe musical influences."

And those properties give us the connections to other things for us to say, "Based on that, here's a kind of music that you might like."


Because, you know, in the dating apps, we'll often see things like, well, somebody likes live hip-hop performances and someone likes EDM music. Well, what's common there is a physical journey to go somewhere, be part of a physical place where they're playing music. So we're saying, "Ah, what's really interesting is those two people might get along and be a good match," even though only one person put hip-hop and only one person put EDM.


We're getting a match at a much more granular, much more cultural and psychological level that says, "This is the kind of thing that we think means or is important to the person as part of the thing they have expressed interest in." So that's what we bring underlying the data that we bring in from these companies.


Kyle: Very cool. It sounds exactly like, again, not to always go back to marketing, but when you upload a customer list to, you know, like Google or Meta or something like that, it matches those people to their interests and their behaviors and then you can build lookalikes off them.


It's like, "Oh, here's other people that also have these interests and these behaviors so I can show our ads to them." So that makes perfect, perfect sense. It's interesting to see all of the different angles that you can take. It's very, very cool. I'm curious to see where it keeps on going.


For a small business owner, or even a, you know, a medium-sized - anybody that's really trying to get into this - what is your advice for deciding on going to a provider and purchasing either a tool or a dataset versus kind of building and using their own?


Is there a thought process there? Is there a time that you think it's better to do one versus the other?


Ian: So when you say buying a dataset, you mean doing your own analysis or getting… When you're talking about tools, maybe getting involved with like these AI tools and understanding what tool to go out and start working with?


Kyle: Well, I think the AI tool is definitely an interesting topic as well, but I'm thinking about, again, like you talked about the credit card stuff, you know, and how you sync up with that.


So it's like, is it better for me to go out and start somehow figuring out how I can collect that type of information, or is it better to just, "Hey, like go and pay X amount for these people, for this solution, whatever it is"?


And I know I used credit card there. You know, I think that's probably better to purchase because it's not just like wildly available, right? But I'm sure there's some instances where you can just kind of build your own and for the long term it might make a little bit more sense to do that. So I'm curious if you have a thought there.


Ian: Absolutely. We are security-minded and essentially security-first. We deal with no PII. No personally identifiable information comes through. The closest we will get sometimes are things like Instagram handles or X handles, and we anonymize that on our back end. So we can't ever go in and find, "Kyle Lambert liked this, you know, chat on whatever." It's all anonymized for us.


But, you know, at a certain point, you do have to bring in things like YoungBlood. Like we have to say, "All right, I know his social media presences, and those are the things I have to interact with."


Data collection on the scale that we've done it is something that, even with AI, is just a massive, massive undertaking because data is… I've always liked the metaphor of data as oil rather than "data is electricity" because electricity - except for AC/DC - is largely indifferent. Oil, you have to refine. And data comes in and you have to be able to match it. You have to be able to say, "In this…" If I go get datasets from one person and dataset from another person and they both have YoungBlood, they might have different ways that they refer to it. They might have different spellings. They might have misspellings.


So just the normalization and cleaning. So to the extent that you can find clean places that have collected that data, you are going to find, even with the price that they're asking, and sometimes places ask a ludicrous price. I've been at companies where my big job was signing data purchasing deals, and they are crazy amounts.


Even then, you know, you can do a little back-of-the-napkin calculation on what it would take you to collect that data yourself and then triple it. This isn't just like adding a little bit for overhead. It is going to be so much harder and take so much longer.


To a certain extent, to my mind, what you're buying is time. It's not just the data, it's you're buying the time that it would spend you. Because you'll bring it in. You'll be like, "Look at this. We're just sitting on mountains of data." And I was like, "Well, congrats, but the next year of your life is going to be fighting all of the corner cases that are keeping your data from being high-quality."


Kyle: I mean, I 100% agree. A lot of these tools have gotten to a spot where it's easier to learn, but to be able to still take something and build it so that it's useful, like you're saying, it's sometimes just… Yeah, just if you have the budget, just pay for it.


Ian: You're buying time. Think of it that way. You're buying your company's time back.


Kyle: Yeah.

Cool. So I'm curious. So you're 15 years into your career in AI and machine learning. You've learned a lot. Where do you see this kind of going? And I know that's a broad answer, and, you know, in five years it's going to look way different than ten years. But I'm curious, just from your perspective: AI, personalization, Taste AI, cultural intelligence - where do you kind of see that changing or evolving to in the next five years?


Ian: Well, speaking as somebody that's part of Qloo, I think - and we're already doing this - we have to be aware that a lot of the teams that would use the data that we provide or access to the data that we provide are going to be using these types of tools themselves.


So the people who are going to be understanding, it's like, all right, I want information on certain populations, certain groups, how my company is, or how things have performed, and I want to bring in data. So we need to be meeting people at the place that they're going to be at, and I think more and more they're going to be in some form of chat-like interface, or some OpenAI or ChatGPT notion or Claude.


So our ability to get the data there, like our API has, you know, an immense amount of complexity on the back end that we spend all of our time making simple and clear and for the user to understand, for a more technical audience, how to query the API, the parameters that go into the querying, the way to get data back and the filtering and the signal that goes into that.


I would love to see us be able to do that in a much more natural-language kind of way because I think people who are going to want to say, "Look, I'm just looking for these things," are going to be so used to working in that chat format that even, like, jumping back into a light SQL or a light, like, re-query writing - it's like, "All right, so filters equal these WHERE columns equal these," or whatever it is - are going to just default to wanting that kind of behavior.


So even though I don't think the final form of AI and LLMs is the chatbot, like I don't think that's the end of the evolution, I think the frequency with which people see it that way is going to make them comfortable with it, for a while, doing that.


Larger thing, anybody's guess is as good as mine on this stuff. I tend to think, like I mentioned before, I think the hype is going to hit reality here. I don't know. I know there's a lot of debate about whether this bubble is going to burst or slightly deflate or whatnot. I don't know just how down the trend is, but I do think there is going to be some realization that not a lot of the industries are absorbing AI as fast as software coding and tech can.


And, you know, we can get into crazy things about the stock market in terms of like, you know, 40% of the S&P is basically just one, you know, one company - NVIDIA. I'm not going to make any predictions in that, but I do think when the reality hits that it's not going to diffuse out as fast as it does with the tech crowd, there will be a slight diminishing of the shininess of it.


But, you know, that's just going to increase as people find better ways to integrate it, move beyond the chatbot, move beyond some of the basic ways it's used now, get it more integrated into the tools you already are using and where you're at.


I think that's when we're going to realize like, "Oh, it's behind everything."


Kyle: I couldn't agree more. It's just going to be fueling a lot of the, like the apps or the, you know, the browsers or whatever it is that we're using, along with actually interacting with…


I like your statement about using it from like a chat perspective. Again, thinking about it from like a phone model, like, you know, people started building out these applications and the phones got used more because people started using the apps and stuff. So I think if we're able to build a tool that is just natural and people know how to use it - everybody knows how to have a conversation, right? - so it only makes sense that you'll get more adoption for AI as a whole as well as specific tools if you can just speak to it, you know, or type with it. So that makes perfect sense.


Cool. All right, so we've gotten to the rapid fire section. I've got five quick questions for you.


Ian: Let's do it.


Kyle: All right, let's jump in.


Question number one, I ask everybody this. If you could snap your fingers, wake up tomorrow, and there is some part of your business or your life just now fully automated, what would it be?


Ian: Scheduling my daughter, figuring out where she's got to be and when.

I love that she's so interested in so many things, but it means my wife and I have to coordinate our calendars with her calendar and emails from school and emails from the after-school programs. And if I just filter all of that into one tool that just said, "Here you go," I would… I'd follow it like I follow GPS. I'd just open up my calendar and be like, "All right, here's where I have to be now."


Kyle: Perfect. Yeah, that'd be so nice. How old is she?


Ian: She is 11 years old. Just turned 11 before Christmas.



Kyle: There you go. Yeah, just starting to have her friends and her hobbies and all that kind. Yeah, I'm sure that's fun.


All right, question number two. With your work at Qloo, is there one kind of cultural crossover or correlation that you guys have seen that just like really stands out that you found to be super interesting or maybe something you hadn't thought about?


Ian: That's interesting. You know, people use it in tons of different ways.

I will, I'll only half answer you and I'll say what we find the most is that, until we explain a little bit more about what's going on, people are surprised by what doesn't show up.


So you just put things in and they're like, "Wait a minute, why, like, I put in a hotel. Why isn't this showing me all the McDonald's nearby?" I was like, "Right. Is that what you want?"


And it's like, "Well, it's the most popular restaurant in the world," right? It's everybody. I was like, true, but not in the way we think of popular, which is people talking about it and engaging with it.


It's like, yeah, tons of people go to McDonald's, but it's not an aspirational thing where everybody's like on Instagram liking the latest McDonald's reel.


Kyle: Outside of, you know, the 2:00 a.m. munchies, I think you're right. Most people aren't like, "Oh, I need my chicken nuggets right now." So, yeah.

All right. Interesting.


Number three, in terms of your career and everything that you've done with AI and machine learning, I'm curious if you've had one just very large "aha" moment that has kind of like changed how you think or how you operate or anything like that. It's like, "Ah, this makes sense. I'm doing this now from now on," type of thing.


Ian: There were two that I can think of.


One, when I was doing my dissertation, early on, when the whole like vector model process was coming out and talking about things and embedding words into vectors and then doing things… The classic example is that there was, you could put into the system, "king minus man plus woman equals" and it would give you back "queen." I was like, "That's really, really interesting."


But I did that with some entities that were at a company I was working at where, you know, we put in a major company, subtracted the CEO, and then added another company, and we got back the CEO for that other company.

So, like, it understood without telling it - like just, "Here's the name" - it understood, "Oh, you're trying to think of the CEO of this company. I can give that back to you purely without any, you know, refinement and just saying, 'This is based on our data.'" It was like, "Oh, this is going to be a massive change, and this is the most fascinating thing I've seen."


So, you know, that's kind of, like I said, early days word/doc2vec kind of stuff, but I see LLMs doing it, and it just was stunning.


And then, most recently, like, as soon as I started doing most of my work through Claude Code, I am working on code this way from now on. Like I'm not going back to opening up a blank page and starting to type "import pandas as pd," or whatever, you know, your flavor of code happens to be. So, you know, I'm no longer just going to open up a text file and start typing at the top, "import pandas as pd," and then, you know, create my script on my own. I am always using this new tool.


Kyle: Very cool. Very cool.


Question number four. Earlier in the episode you had mentioned that you worked on, it sounded like an app or a system for like fishermen. Are you yourself an angler? And if so, what was the best catch of your career?


Ian: I am a frustrated fly fisherman. I absolutely love it. I have done it for a few years now. I am still terrible at it. So I call it, or I describe it as just walking around in the water annoying fish.


But I was recently at a river here in central Pennsylvania and caught myself a 24-inch brown trout, native brown trout, which just blew me away. It was huge. It fought for a long time. It was great. It was, you know, you're standing in this gorgeous river with the mountains around you. Just imagine exactly what you think from the movies and it was unbelievable.


Kyle: That's awesome. Congrats on that catch. That's probably a good feeling.

Cool. Question number five, last one. I love your background, your studio. I'm curious if you have something like a favorite feature or, you know, if there's a story behind it. Just wanted to kind of ask about it. I think it's a cool little scene you've got back there.


Ian: Oh, well, thanks. So the blocks are actually soundproofing blocks, but I just turned them into a graph because I was like, "Everything I do is data."


But I will say the one that means probably the most - let me do the camera thing as I'm pointing at it - is, it's not terribly visible right now, but I have a blanket that is actually part of the heritage of a group in Kenya called Maasai warriors.

And I managed an international team at one of the companies I was at. I had 100 people in India. I had people all across Africa. And I had such a great team, and we worked so well together.


One of the guys ended up sending me the blanket that he had from his family, and his family is Maasai, and they, you know, make these and they use them for everything. It's like to keep warm. They're like cover-ups. They, you know, things to sit down on, things to wrap up food, carry. It's, you know, in Japan there's a similar thing called furoshiki that has a kind of multi-use. And he sent me one and I absolutely love having it.


Kyle: That is very cool. I'm glad I asked. I thought I was going to be like, "Oh, yeah, I got this cool light back here," but that's a really cool story behind it. I like that.

All right. Well, Ian, we are getting to the end here. I like to just do kind of an open forum. Is there anything that you're super passionate about that you wanted to share with everybody that maybe we didn't get to touch on?


Ian: I think we covered tons of stuff. You know, I absolutely love what I do every day. I have a couple of things I do on the side. I love having conversations with people about how they can use AI.


So, you know, I've got my own little consulting thing called Fierce Highways.ai. But people should just reach out for conversations. It's one of my favorite things is blue-sky thinking and then turning those into rapid prototypes that say like, "You know, prove that it could even work," and just say like, "All right, we can now do it pretty quickly. Here's proof that this is going to work. Now, let's talk about how to make it, you know, scalable and functional."


I love those, and I have those kind of conversations all the time. Even if I don't work on it myself, it's just going through that process of brainstorming and saying like, "Yeah, you think it might not be possible, but I bet you there's a way to make it work."


So I absolutely love that. And then, you know, shout out AI at Work again. This is… You're a Philly fan, and maybe a couple of people in Pennsylvania have heard about this, so if they're in the Pittsburgh area they should always stop by and reach out. But that's about it, Kyle. It's been such a great conversation. I've absolutely enjoyed having the time with you.


Kyle: I agree. I agree. Thank you for joining. You have so many great insights. Lots to dive in there. I hope people enjoy it.


So they could check you out, Ian Cook, on LinkedIn. And then Qloo is qloo.com if people want to check out there. And then you also said fiercehighways.ai is your…


Ian: It's just my own…


Kyle: Yeah.


Ian: Personal, yep, exactly.


Kyle: So people can check you out there. Anything else to plug or share if they want to reach out?


Ian: That'll do it.


Kyle: Awesome. Well, Ian, I really appreciate you joining The Brainiac Blueprint. This has been awesome.


If you don't mind, take a look at the camera and just say, "Stay brilliant, Brainiacs."


Ian: Stay brilliant, Brainiacs.


Kyle: Awesome. Thank you so much, Ian.


Ian: Yeah. Thank you, Kyle.


 
 
 

Comments

bottom of page