Who Do You Trust? Rethinking Data, AI, and Decision Making

Check out the full episode below! Enjoy The Exchange? Don't forget to tune in live Friday at 12 pm EST on the Greenbook LinkedIn and Youtube Channel!

Synthetic data is no longer a niche research concept, it's becoming infrastructure. This episode maps the wave of new products and vendor moves reshaping how companies collect, simulate, and act on customer data, from Qualtrics' synthetic panels to swarm-based agent systems to AI-powered video mining in retail aisles. Lenny Murphy and Tim Lynch, dive into what's actually hard: data ownership, AI team dynamics, cross-functional ROI politics, and what happens when non-human agents start buying things.

Many thanks to our producer, Karley Dartouzos. 

Use code EXCHANGE to get a 15% discount on your general admission IIEX tickets!

IIEX North America

IIEX Europe

IIEX.AI 

Stay Ahead of the Curve! Subscribe to The Exchange Newsletter on LinkedIn Today! 

Transcript

Lenny Murphy: And we're live. Happy Friday, everybody, and a special welcome to the man, the myth, the legend, Mr. Tim Lynch. Tim, welcome. How are you?

Tim Lynch: I love taking over Karen's literal chair for this.

Lenny Murphy: That's right. You know, she's going to be watching this on the plane going, what is he doing? For those who don't know you other than, you know, Karen's husband. Which one else does anybody else need to know, right?

Tim Lynch: That's a full-time job, right? And then the side hustle is, you know, my arc has always been the intersection of technology and behavior. And so a lot of research and insights in the medical field, in the tech field, and now focusing on, you know, how do we bring AI to change how people are behaving?

Lenny Murphy: So that's the definition of change management at this point. Yeah, yeah.

Tim Lynch: Yeah, there's the do it with me and the do it to me change. And they're both going on at the same time.

Lenny Murphy: Boy, do it, do it too. We could take that to lots of different places. I'm sure we will as the conversation progresses. And for the audience, we had promised that we would be focusing this week on the zero human company concept as a lens for some of this, do it, do it, do it with me, do it for me or do it to me. We'll get to that, but there's a bunch of other stuff that we wanted to touch on before we dive into the news. Reminder, IEX North America is right around the corner. So you can use our discount code. So please, please do that. And then followed, we've got IX Europe, and then we have IX AI. So be there, be square. They are going to be exciting events, and the agendas are awesome, and the attendance is going to be awesome, and you literally cannot afford to miss these guys. I'm serious. The Insight Innovation Competition winners were just announced this morning, which we don't have up here as a link, but you can on LinkedIn and that's just an example of the type of disruption and transformation that we're seeing that you're going to see on stage. All right, plugs out of the way. Tim, you're here because Karen is at X4 and she gave us a little sneak peek of some of the announcements there. It's all about synthetic panels for US consumers. They're saying half the cost with a 12x accuracy advantage over general purpose AI, plus a research hub, so to the conversational answer engine. So Qualtrics' big push on launching a whole suite of tools that are based on synthetic data, which we'll get to kind of the providence of that in a minute. And then, obviously, letting all the data that clients have within their accounts be leveraged now in a new way. In some ways, utterly unsurprising, which I mean, no shade of like, of course, Qualtrics is going to do those things. Of course.

Tim Lynch: I mean, what, 30, 50, 70 years, we have been struggling to connect the dots. How do we connect the dots? How do we know? Tom Watson from IBM, if we only knew what we know, right? Like, right. This is not a new problem, but this is a new solution to an old problem that we've always had.

Lenny Murphy: 100% locking more utility and usability of data. And that kind of brings to mind, there were a couple of things this week that I think were interesting on the, that side, let's run through them real quick. And then let's frame this up. But the bigger issue, the, uh, Quantalope released their category twins synthetic off of the client's own brand data. Samuel Cohen at Fairgen did, actually we should have talked about this first, the synthetic data has been this catch-all term. Samuel did a really interesting taxonomy on let's separate these out based on the types and use cases. I really encourage everybody to read it. Samuel, if you're listening, thank you. It's a much needed way to think about this and the way I viewed that was the use case as well as the source kind of defines that taxonomy. And as we see that too, go and step out, Claude, the mirrofish god view. Have you paid attention to mirrofish? Have you seen that? Why don't you explain? Or a fish, because it's a whole other take on the synthetic sample kind of idea.

Tim Lynch: Yeah, I mean, this is, at its core, it's pattern recognition, right? It's pattern recognition to give us some confidence in a prediction that we're trying to make and applying it to predictive, predictive markets. Um, there's a lot of it's gambling here, right? Like you're trying to predict the value of something else and then you're placing, you're literally placing a bet on that. So, um, you know, that, that's a, that's a hard one because it's a very cool application to see how accurate it is. But again, you're, you're, you literally are gambling. And you know, um, fun to see where the tech is going. I don't know that I, uh, would be that risk tolerant when I bring it to my brand, because, you know, I want to play smart bets and, and kind of de-risk most things.

Lenny Murphy: But for an early stage, I kind of hypothesis testing or hovering in this, the, you know, what's interesting about this, if you haven't checked something, your fish is not a research play per se, right? This is a broader play of agents that have their own memories, personalities, and behaviors. Swarm right the way kind of I understand it where these agents that that mimic certain personalities and types are collaborating and working together to solve a problem into your point make a prediction overall the where's that data coming from that's forming those like that a little fuzzy on my assumption is that it's open source you know just everything's available It's open source and social and markets.

Tim Lynch: And, um, there was, um, Oh, I wish I could attribute the quote to it. A person's name just flew out of my head. Um, but like we are, we humans, we are the weakest link. Um, one of the groups that I'm in is talking about, uh, the weakness of humans. We now have agents swarms that can go off and do multi-agent work multitasks for So the question to ask the mirror is, are you ready to articulate what you need, what you need done with clarity so that that can happen? Um, and I did a test yesterday. I'm currently at 45 minutes. I can walk away for 45 minutes and have it do whatever it wants to do. And come back and I'm good. There are people who have it where, you know, their claw bar or their open claw is working for days. I'm nowhere near being able to articulate a need that well.

Lenny Murphy: Well, let's table that for thought for a minute. Um, because I think as we get into this other concept that we want to make sure we talk about, that's important for the audience to kind of understand the scale that is happening here in the applications as we think about business models. I want to drill back into the data piece here, because there was another thing that, as soon as I saw it, I was like, of course. Then it came out that Niantic, the guy at the Pokemon Go company, basically generated a billion real-world visual AI training scans. So all that data, you know, it was, it reminded me of back in the day Zynga with FarmVille, you know, and they made the statement, we're not a game company, we're a data company. So the games are just how we get the data. And I think we have to recognize increasingly that this issue, especially if something is free, we are the product. And, increasingly that's not because they're trying to get us to buy stuff. They just want our data. Um, because it powers things like, you know, mirror fish or all the synthetic or what, uh, you know, what Qualtrics and other companies are doing. And, uh, our friend, uh, Vivek Bhaskaran from question pro, I think you, you know, Vivek from back in the day. And he glommed on, and I don't want to get into the, yes, question for his competitor Qualtrics. Okay, fine. But he brought up a really good point, and other folks were jumping on this bandwagon around data ownership, IP, there's both legal and kind of ethical concerns around this entire concept of, oh, we just have access to all this data, and we're just going to make stuff. Stuff off of that data, and you're going to pay us for the stuff we make off of the data. And in that, his specific piece was, well, no, there's kind of a chain of custody, or chain of ownership, probably a better way to say it, in this world that we are not really dealing with, and particularly in the world of research. And that comes with data provenance, right? Who owns the data? Our operating assumption has always been that the end client owns the data. Whoever's paying to collect it, owns it.

Tim Lynch: It's a work product of what they're paying for.

Lenny Murphy: That's right. And if you're using software to collect it, that software does not have rights to that data, even though it may say so in the terms of service, right? The owner does that. And I remember when Google Surveys launched, back in the day. And that was a big issue. Google was very clear out of the gate. We own this data. We're granting you a limited right. If you use our platform to collect this data, it's our data. You're just basically licensing it. And a lot of brands were like, well, I'm out. No, we're not going to do that. And here we are again. And it's a bigger issue around everybody creating synthetic products. On who owns it. And whether we're talking about Qualtrics, or Qualope, or Mirofish, or Claude, or OpenAI, this is a big issue. And I think it's going to get bigger. So I think it's going to come to a head more than we saw with the advent of GDPR or the California law. I think it has much, much bigger implications. What do you think? You're in the middle of this from the tech infrastructure and change management standpoint. Is this a topic that brands are saying, wait a minute, we need to pay attention to who actually owns the rights to the data in this massively morphine amorphous ecosystem?

Tim Lynch: And on the manufacturing side, it's a sensitive, hot topic around particularly innovation in the manufacturing space. And building something for an end client, they're paying for it. But I'm building the patents on this. I'm building the manufacturing process on this. So there's, there's an entire category of this conversation that's separate, but very related. Um, that has to go on. Um, you know, with the Pokemon thing, it is, um, there was a transaction in kind because. I was battling to get more Pokemon, right? So I wanted to catch them all. That was, that was it.

Lenny Murphy: Did you catch them all Tim?

Tim Lynch: Oh, not even close. But, um, what they were finding and what Niantic did was a lot of the gyms were in public spaces. They picture that in public and that's kind of like, you don't have an expectation. Of privacy when you're in public. Um, there's a whole first amendment auditors that are out doing stuff like that to get YouTube videos. So, you know, like that is their data. It was my picture to prove that I was in public to catch a Pokemon, our transactions done. Now Niantic can go and do whatever they need to do. I don't own that photo anymore.

Lenny Murphy: Well, it's self-driving cars, right? Um, and so we, that world the you know so that people underestimate thinking about the so everything else around Elon Musk right but the idea of him building a massive training set of real-world data right that's part to the operating of his products I would I would agree that's it's necessary right I think we get that the you know well yeah the car needs to learn you know it needs to see everything around it so it doesn't crash and that's a virtuous cycle, that's a useful transaction, right?

Tim Lynch: Yeah, I remember when Google was doing their little egg-shaped car and they were talking about, there was a conversation around the training data for it and mapping and image processing and the engineer from Google had put up a video of one of their cars interacting And so he was like, now we've got to process alligator, dog, woman, or we can just have a rule. Don't hit things. It's much simpler to have a rule that is don't hit things. It's much simpler to have a rule that is don't hit things. And so he was like, now we've got to process alligator, dog, woman, or we can just have a rule. Don't hit things. It's much simpler to have a rule that is don't hit things. Anything other than that is to try and figure out a woman, a broom, an alligator, or a dog.

Lenny Murphy: Is that gonna hit the alligator? That's an expensive repair.

Tim Lynch: You do get a belt and shoes out of it though. So, there you go.

Lenny Murphy: Uh if Ford a man, you know, they're all over that but the uh uh yeah, it's just, you know, I guess we step back, look at macro level. What I was thinking reading this is that the, you know, we, the dimensions of disruption are, seem to just be increasing over, you know, there's a domino effect here. It's across the board. It's just not, these are not isolated things. These are systemic challenges that as we race to use the benefits of synthetic sample, for instance, right, regardless of modeling behavior and personas in all of its different use cases, you know, whether it's research or not, obviously, there's material benefits of doing that. And from a cost and speed perspective, and, you know, competitive advantage, blah, blah, right? But, you know, we do not have well defined guardrails on a lot of this stuff. And we're gonna have to just keep trying to figure it out. And my personal opinion is, as from an industry standpoint, I would rather us figure it out than to be forced upon us from a regulatory standpoint. So somebody's gotta lead the charge and kind of think through that. And to your point around the manufacturing, is there similar, I hear that there's similar angst about it. Are there efforts that you see happening so I have a lot of manufacturing clients. I have a lot of, uh, federal agencies and large manufacturers. The commonality for all of them is it is hard to bring in a cross-functional team that has the time and skills and the ability to do that.

Tim Lynch: And it's, it's, it's space to process this. Um, it's really a resource allocation problem. Um, and, and we're not going to spend the next hour talking about data quality, but it is a data quality problem. Like we're there, there it lies within all of us lies the data hoarder of the folder on my desktop that has all of my stuff. And in an AI world, it's got to learn. What is relevant data and irrelevant data. And the more irrelevant data it has to sort through, the less the quality will be.

Lenny Murphy: No, that's interesting. Cause I, so this, I got, I got a free trial to perplex the computer this week. So, um, uh, and Oh my God.

Tim Lynch: That goes with your, your open claw that goes with you.

Lenny Murphy: I'm not willing to go there, but with open claws, but first I was, OK, do you want to link to all these things? Like, no. I'm not giving you access to my email, to my hard drive, to my file structure. I'm not doing that. However, OK, there's specific things I want to achieve. So I'm creating folders in Google Drive and moving things that are aligned to very specific things. Like, OK, so I'm going to give you access to this folder. See what you do. Right, right. There's, you know, grit data, like, all right, here's the best 15 years of grit data, baby. But it is, it is the same issue. I'm not willing personally to give access to Microsoft's bad stuff. All of my data, all of my personal information, you know, because my hard drive is a mismatch of personal and business and, you know, tax returns. It's all there. It's like, I'm not willing to do that. Um, but I'm willing to do this very specifically because I'm going to get value out of it, or at least theoretically it will, um, do that, have that mindset when we're thinking about kind of commercial applications of third parties and, you know, it's just, it's very hard for client teams.

Tim Lynch: Um, you know, there is that, that traditional dynamic, right? Forming, storming, norming, performing that goes on. Add in this pseudo human, you know, AI layer of a tool that is a computer, but a person, but a robot, but a person. So they're a member of this team. How much data are we going to give it? When do we trust it? Do I get a value ROI? Like is the relationship worth continuing in some form. Um, and then you have, you know, people from the marketing side of the house, the ops side of the house, governance side of the house, contract side of the house, sales side of the house. Um, and there's always somebody on the outside going, did you get ROI ROI yet? Like, okay, we're going to see that in March. And it's like, we're still trying to get to speak the same language.

Lenny Murphy: The priorities. I, I understand exactly. I'm working.

Tim Lynch: I've been in, I've been in finance meetings where they're like, okay, well we're going to bring these three people in for a focus group and you can see the insights team cringe and it's like, what's that from Uh, the princess bride. You keep using that word. I do not think so.

Lenny Murphy: Yeah Interesting times you mentioned applications. Let's have a bunch of new product launches. Let's run. Yeah I'm gonna run through a couple of them then we're gonna stop. With a few that I think are interesting, but some of these are just kind of shout outs. More of the same room talking about listeners. This is the pace of change Content squares launched a new agent to track customer journeys inside AI assistance. And so chat to PT apps, web, mobile, not just digital properties. Hugely important, waiting to see something like that on, okay, we need to now understand what our AI agents are doing, particularly from a purchase standpoint, but not only, browsing, et cetera, et cetera. So there's this whole other category of businesses being set up around that. I think that's really, really interesting. And we're trying to figure out where that goes. Did you see Google Stitch, the Vibe design platform?

Tim Lynch: Well, it's been out for over a year, but it's newly empowered and has new features.

Lenny Murphy: Pretty damn cool. It's a design. Platform guys.

Tim Lynch: So yeah, if you're in the, if you're in the Google ecosystem and this is, you know, a lot of people live in, in houses, right? Some people live in Google houses, some people live in Amazon houses, some people live in Microsoft houses. They all have these tools of rapid prototyping, vibe coding, something from your, your brain to your mouth to, to a screen.

Lenny Murphy: Do you see your wife just commenting? I mean, yeah, I'm in a cab heading to the airport. So check up on you, too. Don't mess up my desk. Mr. Lynch just to make sure Yeah, I think I live on like a trip I live in a compound trailer park, right so it's like But yes, and I'll do all that solutions coming out. It's obvious that Google is trying to own, uh, the workflow, particularly from a creative standpoint, which makes sense since advertising is the core of their business.

Tim Lynch: Um, uh, and they have a, so add in stitch to Pameli, which is another one of their tools, which can then do content creation. So you're nudging the AI to create several creative banners. Hosts like Yeah, okay.

Lenny Murphy: And it's leveling the playing field. So the right small businesses have actually the same tools as you know, a fortune 500 brand without the need to hire an agency. Sorry, agencies. I still look at that as a process is not so concept and execution of the concept, that's the value. But the process, we just don't own it anymore.

Tim Lynch: Well, there is such a need because AI without the artist is slop, right? And we've seen that. We've got AI slop everywhere. And things like Stitch, and Pameli and lovable and all of these other vibe things. There's an artistry there of where you're where you do need your agency. Because yeah, I can create a graphic. But I don't have a background in graphic design to understand white space and subject and shading and color. And so somebody who does and can articulate why those things matter, just has a better prompt.

Lenny Murphy: Yeah, judgment and more and more, I think that's the human element. A couple more things I wanted to get to real quick. They're in here twice this week. They're AI-powered analytical, AskArthur, conversational analytics products. There you go. I'm sure that's based in the NIQ cloud. So synthesis, the same thing we started out talking about. Qualtrics, basis global and answer rocket. AI-enhanced brand tracking. That's interesting that AnswerRocket is doing this because AnswerRocket was fundamentally a data management platform and partnering up. Video mining, launching their BehaviorScout, AI-driven in-store behavioral analytics for shopper research. That's interesting because video mining as the name would say, they've set up video cameras all across things like Kroger and CVS and in the grocery categories. And they've been doing that for years. And now they're using AI to make that more than just observational, but actually actionable and even interactive in a way. So that was really cool. But the one thing that really popped out was that Nielsen discontinuing their ad monitoring and radio monitoring in 137 markets. Boy, what signal do we need to understand that the world of marketing and advertising has changed when Nielsen says, yeah, we're not going to worry about your radio in these markets anymore. They're not particularly relevant. Yeah, what was your take on that? That was really like a, whoa, shit, really, kind of moment for me to bring that.

Tim Lynch: I know, you know, where the younger generation is moving to, but the older generation is still around.

Lenny Murphy: Man.

Tim Lynch: So like, but do they need those signals there? Or since we're all now logged in the TV, they can all get it from our logins of the apps that we're in.

Lenny Murphy: Yeah, and I think I was making a point somewhere earlier this week of the, um, if we're trying to understand, uh, the kind of engage, understand, activate cycle, um, and understand, then monitor the activation component. It's a whole new ecosystem of ad placement. Now, you know, content creators, you know, sponsorships, you know, the, uh, in the app, et cetera, et cetera, that don't have anything to do with traditional ad delivery systems. We built this architecture of programmatic ad delivery. And that's still relevant, but there's no programmatic ad delivery on Substack. YouTube has its own system, in addition to the sponsorship, that people have their own sponsors and do all that. So monitoring all of that It's just a very different game now.

Tim Lynch: Um, right. And how long will it be before that YouTube programmatic power is brought to bear on Gemini?

Lenny Murphy: Yeah, 100%. Oh yeah. The, well, I'm sure you saw the, uh, the, you know, open AI with the ads. Um, and all the, of course they're going to do that. Um, and of course, You know, you're not gonna give them money.

Tim Lynch: They've got to get money somewhere.

Lenny Murphy: Absolutely, and it's gonna be training data 100% How do you know to click through it? The I mean perplexity was very clear navigate but yeah, we're gonna put ads because we're and we're gonna use it to train so the other Like oh the world's changed and that's what was meta Shutting down the metaverse The after What a bajillion dollars?

Tim Lynch: I mean literally Change their name again then because then they rename themselves to meta because of that whole Yes, right.

Lenny Murphy: Yes, and It really was like hundreds of billions of dollars spent on that. So my take was I still think we'll get there Compute bandwidth all of those issues with the real limiting factor plus I You know, so if they had launched the metaverse with AI driven, hyper realistic graphics, I think there would have been a different thing, right? Then we're in ready player one world. Um, but instead it was like, you know, mid two thousands, um, avatar graphics were not, you know, it was like Minecraft and the market had to be ready for it.

Tim Lynch: People have to be ready to interact. That way.

Lenny Murphy: Yeah, right. And we've seen what was the shit? There was another virtual world. I can live. Yes, thank you. None of those things have stuck yet. None of them have stuck yet.

Tim Lynch: They stick to some pockets. Like sure. I mean, life is still around because there's a pocket of people who still Um, maybe, sorry.

Lenny Murphy: I haven't hit the scale. So, yeah, yeah. Um, but it is interesting to, at the, the four methods, put all that money into that. And I'm sure that's not going to go to waste. You know, there will be, you know, there's aspects of that, particularly data that they'll use, but out pivoting, everything's about around Lama. So in AI, right. Interesting stuff.

Tim Lynch: But five, 10 years from now, we may be having the Pokemon conversation all over again because that base, that corpus of tech and data now is not the right time.

Lenny Murphy: So a couple more, uh, the, uh, Kantar plus quilt, um, uh, partnership, the, uh, include a non equity financial commitment to co-develop joint offers. So that's interesting. Evaluate Explorer, a new innovation tool. Now, I know the Quilt team quite well. Friends with Anurag, the CEO. It's an incredibly impressive business that they've built. And for Kantar to come to them and say, can we partner? And I'm sure that's how that went. I don't think it was the other way around. The Kantar Quilt also shows the clear path Cantor realized they could not, they couldn't acquire them. They couldn't pay the valuation. They couldn't build it because they don't have the chops for the bill. So partnering, so can target leverage there.

Tim Lynch: We come together, we do something together. We put skin in the game, but not like you're not taking a piece of my equity. We're both like, Right, right.

Lenny Murphy: Are you seeing more of that playing out in these more of the partnership approach? Playing out overall, too.

Tim Lynch: A lot of partners, which has a weird dance to it, right? Because what happens when a better dance partner comes along in six months? So there's a little bit of that. And so we're going to be in this weird phase of that now on the manufacturing and engineering side, there's far less. There's far more like we are locked in. There are exclusive agreements for a lot of things.

Lenny Murphy: Well, there's so much sunk costs, right? We're going to put your tech into our pants and you know, yeah. It's just interesting. Let's start through a couple more of these real quick. Phoebe, previous Insight Innovation Competition They expanded their conversational AI with emotion intelligence that's embedded into surveys and interviews. So that's all voice. So cool to see that we're seeing that incorporation of voice tech into the traditional research ecosystem driven by AI. So hats off to them. And then this last one, and we'll get into a little more around the zero human stuff. Door dashes dasher tasks. They're calling it the physical intelligence frontier and so back to what we were talking about before but was particularly interesting is that agents? Would be a buyer So agents getting humans through door dash to perform tasks and Right, which was just uh, you know, it's kind of like a WTF moment If their credit card works, that's right. That's right. That's right So, I don't know what was your take on that of just it's very smart for DoorDash like we have people Who want to monetize their time and there's um expand that out Yeah, and we saw this six months plus ago, there's a company blue something out in California.

Tim Lynch: Basically, they have a warehouse, they have a cool chilled warehouse of GPUs. Okay, very cool. And they rent out GPUs. So Lenny wants to train Lenny's model, you can contract with them and use a cluster and do your thing. And they had a genetic moment where all of a sudden, they tripled revenue in a month. Well, and then when they dug in, because like, clearly our salespeople are doing great. They realized it was non-humans that were buying it, coming in, renting the GPUs, conducting it and then getting out and leaving payments. So they have this whole, it's a fun case study on how do you position and brand and market and sell yourself to non-humans now, because non-humans are now a viable market. These things are coming. At the end of the line, somebody's holding a checkbook.

Lenny Murphy: It's the creativity of the capitalist system, right?

Tim Lynch: And it's really smart with DoorDash. They have so many dashes that are all connected, that they have geolocation of able humans ready to do things for them.

Lenny Murphy: 100%, yes. And we should point out, it's not just AI, that's a piece of it. They're also doing store audits and taking pictures, AI training sets, right? So expanding into what humans can do with their time and resources to generate additional revenue. I didn't see it expressly put in there, but I am damn sure that research will be a piece of that. So particularly, you kind of install audits and ethnography and things like that. But even then, it's the agent that may be executing it, not the human. And that brings up the big topic. And when we waste our time, because you and I can go on for a long time, but I know that. So a really interesting article on the marginal cost of entrepreneurship is trending towards zero. And another interesting article around paperclip, which is fully automated, basically business by agents, right, the operating system for business driven by agents. Combine that with our previous conference, you know, piece we touched on, if I have an idea, whether it's a creative or a technical piece, and the infrastructure is there, the systems are there for a person simply to come up with a good idea and without needing to have humans to have systems that do the work of humans to bring something to market. Um, and my personal prediction is that we are going to see more millionaires minted over the course of the next two years, uh, from entrepreneurs that simply have come up with a cool utility or cool idea. It may not be hyperscalable, um, but they get a hundred users paying a hundred bucks a month, you know, um, and, or whatever the case may be. And we were just going to see an explosion of entrepreneurs doing that with virtually no costs other than, you know, compute. So what do you think?

Tim Lynch: The news is interesting because we keep, we're talking about, um, we're taking an economic model of a business where I have staff and resources in a, in a traditional mental model. And saying, well, I can run this now on my Claude bot or open claw. But that's not a zero cost. That's actually a growing cost. So, um, there's going to be, it's less, it's, it's less of a cost now because we've got this, these platforms competing, but at some point they're going to turn up the revenue dollars. Dollar dials and those costs are going to go up. But it's an interesting dynamic of what can be possible to test in the real world market. Is there a market for this? Do, do some, I was going to say somebody, but does something want the value that I can create?

Lenny Murphy: And what does that look like, So I've paid attention to a few of these folks, these, uh, these, the owners, right. They've started these and I wonder about working like dogs, right? I mean, so, uh, just orchestrating and managing the agents, right? So yes, there's all the heavy lift, but the, but the, the management, the judgment and orchestration component is still extensive to your point. The, um, And the point they've had investors come and they're like, I don't need investment dollars, right? You're right. The investment dollars don't necessarily drive hyperscale when you, because I just have to add a new agent kind of thing right onto it. Um, so the economics at all levels are changing across the board.

Tim Lynch: So, uh, right. Well, and you have, you know, computer costs, right? So the more the, um, the investors have an arm in there or a leverage where they're bringing more than money to the table, right? Um, you know, an investor brings X number of dollars. Investor B brings Y number of dollars, but Z number of computers and tokens. That changes the equations. Which we see happening at the macro level.

Lenny Murphy: We're ever seeing, you know, this almost strikes as money laundering, right? To an extent of, oh, you know, NVIDIA is giving OpenAI a billion dollars and OpenAI is giving NVIDIA a billion dollars back.

Tim Lynch: That's right. But I think Steve Martin used to have a funny thing about starting your own bank. And he was like, you know, I have to have the right name. You can't just call it Fred's bank. I'll put it here in my you know, white suit right hand pocket like that whole passing money and round of investments is just right.

Lenny Murphy: But now if you look at it, here's my, you know, it's like the App Store. That was fundamentally a distribution layer. So it's great if you need to be in the App Store to be able to distribute them and then you have a market and, you know, blah, blah, blah. I have not seen yet, maybe you have, NVIDIA, OpenAI, Cloud, whatever, right, or Anthropic, or Google, say, to your point, here's access to tools. Create something. In return, we're going to take an equity position in your company. You're going to give us these tools, but we're going to give you what you need to scale the hell out of it. Um, uh, and effectively be owning a marketplace of, uh, of MCPs or agents or, you know, whatever, uh, for these individual solo entrepreneurs too. Rapidly scale and build the businesses overall. Have you seen something like that that I've just missed?

Tim Lynch: Google does. Um, so if you are. If you are a startup, if you are, forget the term, you must be under two years old. You have to be under a certain head count. I think it's under 10. There's $250,000 Google platform credits. Your computer, your Gemini access build in scale.

Lenny Murphy: Do they take equity though? Is it, do they treat it?

Tim Lynch: I don't think they take an equity stake. It's what they want. Build it, build it in our, build it in our world.

Lenny Murphy: All right. Uh, so it's, uh, so who's, I have not seen a zero human company enter the research space yet. Um, but by God, I've got some ideas. The, uh, what, what odds do you lay? In a year from now, so we do this again, how many zero human companies we may have seen enter space? What do you think?

Tim Lynch: I would, over under, I would think over a hundred.

Lenny Murphy: That's my gut as well.

Tim Lynch: I would think over a hundred. We're, we're getting to the place where the open claw, right? Uh, well, Claude bought to open claw to Claude co-work to Google's workspace to Microsoft's open workspace, like they all have that multi-agent ability. And remember what we talked about earlier about having the artistry of the expert. And you may have been a qual researcher, a quant researcher, an ethnographer, you've got these skills, these technical things are going on and you've been using them, but there's a technical level that they were not at. Now they're there. And when your artistry and that technology come together, there's a differentiator, right? Somebody who has 30 years of financial product research and insights with that technology is a unique thing that everybody else's AI can't have.

Lenny Murphy: Well, we shall see. It's interesting. And maybe we should talk offline about what the Lenny bot and the Tim bot look like, right?

Tim Lynch: And there's Dan Shipper, who leads a team over at Every. He had built a proof app. And the problem was he's Dan Shipper. So we all swarmed his app, crashed his app, which he had coded for, but then couldn't vibe code to scale it fast enough. We're going to have more and more of those stories too, just because as these things get stress tested at scale, as they start layering and leveling up, there's different dynamics that come into play.

Lenny Murphy: If Karen's listening she's saying okay you guys here at 47 minutes but there's a couple reading listings. Let's run through these real quick. I know there's one you and I agree that we really need to call out. So I'll stop on that one The well, there actually were two Cool report a study from fortune about Google AI overviews are 40 40 44 percent more likely to surface negative brand content than chat GPT last week we had a different story that said chat GPT was more likely so I think my takeaway on that was this that AI AI search and AI reviews whether it's Google open or whatever you can't depend that it's gonna be a good thing right and brands better start paying pitch for that.

Tim Lynch: Brands can have a bad day. Any day can be a bad day for your brand. That's the study from the bright edge. It was released in fortune, but fortune didn't conduct the study bright edge for their brand tracking. Um, and it's, it's interesting when it went over to fortune, it was like 40% more. And when you read the study, it's like one of them is 2.1 and one of them is like, okay, gotcha. So you went to percentages? Because yeah, yeah. But that report is really interesting for why it skews. And one of them is based on news. And one of them is based on product reviews. So when you're trying to manage your brand, I have to fight two fights, right?

Lenny Murphy: It is not just your brand. And perception from your marketing. It is the experiential component from your users. And, now that is not isolated to one little review site. It's included.

Tim Lynch: It's codified now.

Lenny Murphy: Uh, I thought it was also, uh, this important, well, there's another one, uh, uh, Oh, it was the same thing. We duplicated the article. I'm looking at my brief and we had it twice. OK, this other one from OpenAI Frontier around SAS licensing combined with the SASter article. So two articles here, Karley. The SASter one I want to focus on, though. Anthropic and OpenAI now capture $40 to $50 billion in annual enterprise spend. AI budgets are growing 81%, while overall IT grows just 3%. Meaning up to the 70% software slowdown is budget flowing to foundation models. So the shift from SAS to AI, uh, for some, I had seen that number quantified. I think we all kind of suspected it, but hadn't seen it that way. What you live in this world.

Tim Lynch: Is that your experience? And we have the ramp data, uh, ramp cards, um, which a lot of not really the large enterprises, but a lot of small businesses are using. Medium and smaller businesses use the RAM cards. They publish their data on what they're seeing AI spend, you know, people basically expensing their tokens. So we knew it was going to be big. Jensen Wong, the CEO of NVIDIA said, "Now his salaries for his employees are on a different scale than the rest of us. That's right. But he said, if I pay an engineer $500,000 a year, and they are not at minimum burning $250,000 in tokens, we're going to have a conversation. So it was said, you know, at the conference in passing, but I had to wait a second. You have an expectation that 50% of your employee's salary is spent in AI usage. Anybody else ready for that kind of expense? My finance department would be freaking out.

Lenny Murphy: Yeah, yeah, yeah. That's a good point about scale, but the transformation thinking I just don't know how many conversations I've had with SAS companies in our space. And it's like a lot of the winds have changed, right? We have to think about this model differently now. So I think that's just one of the examples of that.

Tim Lynch: So right. The orchestration.

Lenny Murphy: That's right. And speaking of orchestration, there was one other article Magnolia made that the creative brief is the last human artifact. I thought this was really interesting on, again, orchestration, human judgment, and we talked about today that the artist, you know, of putting things in context while the execution layer, that's increasingly, we just don't own it. And all of this, I think that's the theme running throughout all of this, today's session, is we don't own processes. So we only, and orchestration of process and creativity, but that the ship has sailed. Is that your take as well?

Tim Lynch: Well, I mean, for some, right. And that's the hard part is, you know, when we're all out in this AI world, it's really for some, you know, it's not for everybody, but it seems to be everywhere. And so between our synthetic, we have our pick of any of the links that I'm sure Karley has already shared. Today on synthetic samples and who's got synthetic and building your lookalikes. Like you can do that now. Now we can test so we can create, we can create samples and test. What goes live, what goes live to these crazy humans.

Lenny Murphy: And what's sellable, what generates revenue. So, all right. So Tim, I had a feeling we were going to set the new record. Um, you can, uh, you can tell, that, of, you know, we set a new record. It's fine. Maybe it's not one that we should have done. But audience, I hope that you appreciate the different perspective we wanted him in. One, because he's, you know, he's family. But two, he's, he's at this outer edge at kind of the macro level of helping organizations change and adopt and seeing these, all of these things shift. And I hope that was a valuable perspective for everybody. We will definitely have you back. Some other points Tim. Thank you. Anything else you want to add before we drop?

Tim Lynch: No, this has been great. It's a lot harder being on this site, usually Right where I don't have to read 40 different articles Yeah, and that was the curated list that was cut down so Alright, have a great weekend audience.

Lenny Murphy: Have a great weekend. Thank you for sticking with us on this epic edition of The Exchange. We will be back again next week with, well, it'll still be epic, but maybe not quite as long. Or who the hell knows, right? Who knows what happens in the next week from a news standpoint. But everyone, take care, be well. We'll talk soon.

Links from the episode:

12x accuracy advantage over general-purpose AI 

quantilope follows with Category Twins 

Samuel Cohen, Fairgen - Synthetic data is doing too much work as a term 

Claude builds a MiroFish God View terminal 

Niantic disclosed that 143 million Pokémon Go players unknowingly generated over 30 billion real-world visual AI training scans  

Vivek Bhaskaran's essay on SaaS data misappropriation 

Contentsquare launches new AI agent and analytics capabilities 

Google Stitch launches as a "vibe design" platform from Google Labs 

NIQ launches beta AI-powered analytical capabilities inside Ask Arthur 

Basis Global + AnswerRocket partner for AI-enhanced brand tracking 

VideoMining launches Behavior Scout™ 

Nielsen discontinues TV ad monitoring in 137 markets and shuts SIGMA radio ad monitoring 

Meta announces shutdown of Horizon Worlds 

Kantar announced a strategic partnership with Quilt.AI 

Phebi expands Emotion Intelligence with conversational AI 

DoorDash introduced Dasher Tasks 

“The marginal cost of entrepreneurship is trending toward zero” 

Paperclip: open-source orchestration for zero-human companies 

OpenAI’s Frontier puts AI agents in a fight SaaS can’t afford to lose 

How Much of the Software Slowdown Is Just Budgets Flowing to Anthropic and OpenAI? 

The Brief Is the Last Human Artifact 

Appendix to “What 81,000 people want from AI" 

Google’s AI overviews are 44% more likely to trash your brand than ChatGPT 

PitchBook's new MCP integration with Perplexity