SPEAKER_01 0:00

Every decision has a signal buried in it. Not everyone finds it in time. At Signals and Stake, this is the podcast for people who sit inside the decisions that shape how companies grow, compete, and survive. Not the lessons learned version, the actual calls made with incomplete information, competing priorities, and real consequences attached to them. Each episode surfaces one of those moments. What was the signal and what was the stake? And then, of course, what happened next. This is not a show about best practices. It's a show about what happens when the stakes are real. Hi, I'm Akonde Davis, VP of Operations at GW Consulting.

SPEAKER_00 0:33

And I'm Roger Wallier, CEO at GNW Consulting, and this is Signals and Stakes.

SPEAKER_01 0:39

So you walk into almost any C-suite conversation right now, and the same theme is on the table: AI infrastructure, platforms, agents, data fabric, models. CapEx is going up, head count being scrutinized. And I think the implicit assumption underneath it all is that leaders who build the infrastructure will win first. Um, but I have been in enough of those rooms to know that the infrastructure is really what's actually the problem.

SPEAKER_00 1:03

And that's a bet that has a ceiling, right? Like it eventually has a cap. Not because like AI isn't real, um, it's it's very real, like it's everywhere. Um, but because the part of the business that actually kind of compounds it and judges it, holds it to accountability, has conviction in it, is not what AI is replacing. AI is that kind of top layer, right? Like it's that when the layer is weak, AI does not work. And when it's strong, it multiplies it drastically.

SPEAKER_01 1:31

Yeah, and that's what I think, you know, we've been seeing this a lot on the op side that enterprises are making infrastructure investment while that human layer is still really fragile, right? The data's not ready, decision rights are unclear, the ownership is unknown, it's ambiguous. And kind of the assumption is well, once the infrastructure is in place, those things will, you know, it's gonna sort itself out. It'll, it'll, it'll figure itself out, but they don't. It just gets faster, it gets harder.

SPEAKER_00 1:57

Yeah. Man, I mean, and honestly, like this is not like an anti-AI conversation, you know. Um, I'm a big proponent of new technology, you know. As an architect, my entire life, I love building stuff, and I think AI has helped immensely. But this is what it this is, is an order of operations. Like, what is the order of investment that conversation that we need to have? And right now, the order is really wrong in most enterprises.

SPEAKER_01 2:23

Well, funny enough, Gardner projects a significant amount of share and generative AI projects are gonna be abandoned uh after the proof of concept. So the infrastructure being built, but the layer that makes it work is not. That's a pretty clear indication, right? If you build it out and you're like, nope, not gonna do that.

SPEAKER_00 2:39

Yeah, I mean, and let's like let's place a bet on it, right? Like let's place a bet, like you know, we can fan duel as this thing, like most enterprises are making. Like, what is the bet? Platform consolidation. What's another bet? Agent framework, optimization, some sort of fancy fancy word, data infrastructure, fine-tuning. All of these things are bets that people are making. And a lot of this is being justified under the assumption that AI is going to really replace the work humans currently do. And the savings will pay for the infrastructure and the costs and you know the the transition.

SPEAKER_01 3:14

Yeah, like an interesting idea. This is gonna pay for itself. Uh, from where I've kind of seen things, the work being targeted for being replaced by AI is really the wrong category. It's oftentimes that the pitch deck assumes AI replaces senior judgment. I Raja, I mean, you know, I I I think AI is amazing. I would not trust AI to make senior level decisions for me.

SPEAKER_00 3:35

I don't think I wouldn't trust AI to give me a recipe. I literally, I literally, I literally asked, I mean, you know, I wasn't sure what to make. And it said you should really make chicken bouillon a's with beef. And I'm like, I don't know what the hell you're making. Like, you know what I mean? Like, what like what was yeah, but well, what was my prompt, right? Like my prompt was like, yeah, I want to make a pasta dish. You know, chicken and beef is fine. It's like, well, you know what you should do? Combine them and you have chicken and beef boulion A's. I'm like, that's not that's not really what I said.

SPEAKER_01 4:04

Yeah. The interesting thing is I've seen these videos where um the most recent example is uh is a guy using uh an AI chatbot, and he says, This is all just you know make-believe, but he says, Hey, I want you to make a deal for this loaf of bread. I I have my budget's five dollars, right? The AI agent talks to the guy who's the same person and says, Okay, we could do $400. So there's some judgment value that's missing, but it's it's really the junior level areas that you can get that throughput. So content production, campaign bills, summarizing reports. That's that's real value, but it's not really the ROI profile that's justified in the CapEx, right? Like it's totally different from what most organizations are having, and they haven't broken those things out.

SPEAKER_00 4:48

Yeah, right. And you know, ROI is one of those terms that have been around for a while for a reason, right? Like return on investment, obviously, everyone knows what it is. And it it's not, it doesn't show up at the level that was projected. And that's, I think, the bigger thing is that the cost of AI or agentec technologies is exponentially greater. Hence the reason, you know, companies are making these decisions. Well, let's just, you know, we don't need these jobs, we don't need people to monitor it. Um, and it's usually the spend. Well, better models, cleaner data pipelines, more agents. The gap isn't infrastructure in any way. It's it's who did you let go that could have told that agent how to do their job better and then make it more streamlined. I think that human layer that's underneath is the one that research now shows that a lot of people are regretting, right? Like there's a reason Anthropic, you know, uh the company who owns Claude is hiring a bunch of engineers and a bunch of, you know, I think you mentioned earlier, like GPT, like uh Chat GPT or OpenAI is, you know, hiring BDRs and SDRs. Why is that, right? Like, why why are the companies, and I think personally, this is my conspiracy theory, right? That this is my tinfoil hat on. They just put all that information out there so you would fire everyone, and now they're hiring everyone. You know, like that was that was their version of getting employees. Oh, one, two, switcheroo. Yeah, yeah, exactly.

SPEAKER_01 6:12

Yeah, I mean, I think there's something valued valuable there because again, Adobe's 2026 digital trend report shows that 68% of organizations can't really demonstrate ROI in their AI investments. Uh, other research from MIT and BCG consistently shows the majority of enterprise AI initiatives failed to deliver meaningful financial impact. And I mean, how could they if it's not accounting for that junior level work? Um, I think what makes it uncomfortable to say out loud is it kind of sounds like okay, we're criticizing the AI investment, but we're not. We're criticizing how it's sequenced the infrastructure bet that got placed before that foundation was actually ready. And now people, I don't know, Raja, if retrofitting is the best word, but retrofitting that human layer.

SPEAKER_00 6:56

Yeah, I mean they're hiring them back, right? Like there was this big kind of thing in the market that said AI is gonna take your jobs. AI took a bunch of jobs. The you know, you you saw it on LinkedIn, you saw it on feeds, you saw it everywhere. And now all of a sudden everyone is realizing like it can't do that job without someone monitoring. And that is substantially harder than building it, right? Like it's substantially harder making that decision now. So the the right order of things has always been who is going to teach, you know, it's it's like a training, it's like you're getting an intern. Who is going to teach that intern? Well, this is the first case and scenario where someone said, you know what, all the people that are gonna trade train the interns, we're gonna let you guys go. And the interns, you guys work here 24 hours, seven days a week without food, water, and and you got it, but they have no source data. That source data are the human-led conversations. That that is where human in the loop is going to be pivotal.

SPEAKER_01 7:54

Yeah, and and once, you know, once those agents are live and running, they're doing the things, they're acting on what they know, what's been baked in, the logic. Um, it's not really like, oh, I can come in here and like twist some knobs and everything is updated. This is a redesign that has to happen. You can't stop the train once it's moving. And that's kind of where where things are are at.

SPEAKER_00 8:16

Yeah, I mean, and that's the amplification problem we talk about. It's going to elevate and escalate bad decisions. When agents, AI, I feel like we've said AI more than on this episode than any other episode that there is. Uh, but when it lands on like a weak human layer, it doesn't fix the weakness. It doesn't, and if there's no one there to guide it, it's not going to make adjustments to it. You know, there once again, like someone still has to swipe right, someone still has to tell it what to do, and wrong decisions execute faster. Why? Because there's less of a barrier to it. You know, unclear ownership is going to create faster ambiguity. Bad judgment is going to scale. The technology kind of exposes which leader is paying close attention to it and which leader is not.

SPEAKER_01 9:03

Yeah, and it doesn't resolve it, right? Too, because so for the sake of not saying AI, uh, you know, 75% of orgs cite the top uh artificial intelligence challenge as data integration. And naturally, um, those are human layer problems, right? That that can't be addressed with just spending more money on the infrastructure. It's a category problem. And, you know, to kind of tie into that, the reason I think it gets mixed at the executive level is because every enterprise has access to the same models, platforms, vendors. Uh, you know, we've talked to organizations that have the exact same tech stacks and are getting completely different results. And, you know, obviously we've been in those conversations a lot because it's more expensive to fix the problem than it is to set it up right the first time. The difference is never gonna be the tool, it's the ownership and the quality of that human layer uh assessing it.

SPEAKER_00 9:51

And that's the moat, right? Like when we talk about a moat from a castle, it's not a platform, it's the clarity of who owns it, what's swimming in it, who's gonna lower the drawbridge. If no one's there to lower the drawbridge, then you know AI is just going to either fall in the moat or run into the door, depending on which side is which side. Like the conviction behind it.

SPEAKER_01 10:11

Falling into the moat just sounds so funny. Just a bunch of soldiers just yeah, just falling straight.

SPEAKER_00 10:16

But if you're sold, but on the you know, flip side of the coin, if you're the soldiers inside it, where are you going? You're just gonna eventually just break the door down and then fall into the moat. Like there is no kind of place for that. So the accountability is like something that goes like you know, sideways or both ways, but those are the things competitors aren't gonna be able to copy by signing the same contract. The leaders who are going to treat AI as a substitute for the human layer are racing toward this involvement of them in a cohesive platform, in a cohesive system, a process everywhere where they do things better. And the leaders who are treating AI as more of like a replacement or a multiplier, you're it is going to multiply all the bad decisions. And it's going to build something that is not going to, you know, have a commodity, it's not going to have structure, it's not going to have any legs to stand on. And once again, I think we talked about this in the in the last episode is that the gap of failure you will never catch up from. Yeah. Unless you just pretty much just throw in the towel and restart again, which is even worse. Like, you know, you make your decision at that point.

SPEAKER_01 11:21

If you were to replace the the terminology AI with um, you know, what we were into coffee, an espresso machine, or if you were to replace it with a fancy kitchen or a fancy car, the same person who's not equipped to operate the equipment, to drive the car, to get the most out of it, is not going to reflect well on the product, right? Like if you put a bad race driver in a really great race car, it's going to look like the race car sucks. The same is true with AI, you know. If you put um no investment into the human element, but you put all the investment into the AI, then you're losing out on leadership, decision rights, and the operating model that's really gonna make it work. So that ROI gap is not a surprise, it's arithmetic. The bad driver in the awesome car is gonna have a terrible lap time. A great driver in a crappy car is going to have a great lap time. And that's not because the technology, that's because the people, that human layer that's going into it.

unknown 12:13

Yeah.

SPEAKER_01 12:13

Who's steering it? So, with that being said, Raja, what is this telling us about where the market is right now? What is the signal that we need to be paying attention to?

SPEAKER_00 12:22

Well, I think the we need to recognize and rebalance, right? We need to invest in the human layer with kind of the same seriousness that we invested in the platform layer. Yeah. Because they're those are the ones that that is the investments that's going to make the infrastructure actually work. It's not going to be an agent that's going to be do what that is just going to run a process. Someone is telling it what to do. And if that someone doesn't know what they're doing, or if there's another human layer prior to that, like it is just going to amplify it. And if no one is telling it what to do, it's going to be even worse. So investing in that human layer layer is at the most, it is pivotal. Like you're seeing it with the companies like Anthropic, like ChatGPT that are hiring more humans. Why? You know, why are they hiring them? It's not because they they they are literally the inventors of the tool. Like we're buying their tools, they're the inventors of the tool that are investing in that human layer. Not because they have better tools or they created it, because that underlining layer, it's finally going to match what they're building on top. And once again, I'll go back to my conspiracy theory. My conspiracy theory with my tinfoil hat is they purposely put all of this information on the market so they can just scoop up all of this human-led resources for themselves. That's my that's my tinfoil hat thing.

SPEAKER_01 13:44

Well, you know, and and maybe they didn't even because if I was to say they put it on the market, you know, I'd have to kind of dig. I think everyone else put it on the market. I think everyone else was like, oh yeah, this is what's gonna happen. And they kind of were just like, sure. So the the the technology is not really the signal. The gap around the return on your investment is the signal. Is that is that fair to say?

SPEAKER_00 14:05

Yeah, and I and I think that's correct. And just remember, like when we talk about return on investment, we're talking about a very old concept because at the end of the day, you're still investing in technology. What we're saying is put the people that know what your company is going to do, how to scale it, how to run it, the processes, the efficiencies that you've built as your foundational layer so it can teach that that additional technology, agentic technology better. I guess a better way uh the they can teach it what it should be doing rather than what it assumes that it should be doing. And that is ultimately the signal.

SPEAKER_01 14:42

Yeah. Or or you know, if you if you're letting all these people go, the things are gonna change in six months, right? The data has to be iterated on as well, like the that layer. Now, if that's the signal, what are what what's at stake? What are the things that are at stake if it's ignored?

SPEAKER_00 14:55

You know, I think it's the compounding gap, right? It's the compounding gap of getting it right over and over and over again. You do something wrong, you fix it in the in you know in the in the olden days, you know, circa 2024. I know, two years ago. Yeah, two years ago. You know, you did something wrong, you fixed it, you learned, you move on. You do something wrong now, you hit a button, someone an agent does it 50,000 times or something, like you know, some some exponential amount. And I've seen and we've seen what it looks like in 18 months. That's why 18 months now, after Claude GPT, the dust of AI is settling, and people are going back to hiring human humans again. You know, why is that? And it's funny, just FYI just saying humans in general, like people, right? Like there, there's a reason that they're hiring people. Yeah, yeah. Why? Because they they got rid of that layer that really made things process and work more efficiently, right? We're running faster on what was already broken. Now we see the errors of it. Now everyone is saying, okay, well, we need something, someone to come train these models, which by the way, we've been saying all along someone needs to swipe right, someone needs to train these models. And that ownership question just got harder to answer. And that's where I think where you're gonna see and kind of in the future is that people leaders who are making those decisions are going to be the clear owners of that of those technologies. And that's where you're that's where you're gonna separate yourself from your competitors. It's not in who has the cooler agent, who is the human or the person that is powering that agent in a better light. That is gonna be the differentiator.

SPEAKER_01 16:32

Exactly. The vendor's not gonna be the one who answers for it, not the ops team, um, you know, not the not the intern, right? Not not even the AI agent, but the C-suite. They're gonna answer for it because they're the ones who approve the investment, they're the ones who believe they could deliver on the ROI, and it was never possible. Now, uh, as part of our takeaway, we have the signal of the stake, and obviously sort of the closing thought, right? The question I would sit with is this when you're looking at your investment in AI and your investment in your leadership team or development of operating models, what is that ratio? Is it wildly out of balance? And if it is, just know that a different type of bill is going to be coming due. And that's gonna be the bill that you're you're gonna have to hold accountability for when your AI is not returning that ROI.

unknown 17:17

Yeah.

SPEAKER_00 17:17

And mine's a little bit simpler, right? Like if your AI, and that's just so funny, just in general, say the if everything that you've invested in, let's say from an autonomous perspective or an agency perspective stopped working tomorrow or started doing things that were that came to light that it shouldn't, who is the person that's going to make the call of whether to stop it, change it, or modify it? Do you know who that person is? Can you contact that person? And if the answer is no, you did not build a moat, you built a dependency, and that is and that is going to set you back.

SPEAKER_01 17:55

I think that's spot on Raja. And again, thank you so much for listening. That is Signals and Stakes. We appreciate you tuning in. Be sure to check out next week's episode where we dive into more market signals and what's at stake.