The Forward Slash /Prove It or Park It: AI's ROI Reckoning

Ninety percent of large companies say they are using AI. About six percent are getting five percent more EBIT out of it. Todd James, former Chief Data and Technology Officer at 84.51° and head of AI at Kroger, now CEO of Aurora Insights, joins James Carman to explain why that gap has almost nothing to do with the technology, and everything to do with whether anyone is accountable for the return.

At a glance

Key takeaways

• Adoption is nearly universal. Returns are not. About ninety percent of large companies now say they use AI, and roughly forty percent report seeing some value. But only about six percent are getting five percent more EBIT contribution from their AI investments, and those companies are compounding an advantage the rest cannot easily close.
• The failure is a business failure, not a technology one. Todd has sat in rooms where a CIO walked a CEO through a multi-year AI plan with no value or performance expectation attached to it, qualitative or quantitative. He cannot think of another capital investment in his career that would survive that.
• Start with the P&L, not the use case workshop. The pattern he sees fail is a room full of people generating interesting use cases disconnected from the P&L. The pattern he sees work is the reverse: understand which economic drivers AI can influence for each core area of the business, then find the high density or high value workflows sitting on those levers.
• The value pool is judgment. Process improvement handled the fifties through the seventies. Automation took the deterministic yes-or-no decisions. What is left is people, customers and employees, making judgment calls. That is where the remaining value lives.
• Watch the leaky bucket. Deploying a copilot and assuming savings without measurement means you never learn where the freed-up time went. It might be going to a real problem. It might be masking a different one the company should be fixing. It might be solitaire. Without tracking, all three look identical.
• Moving too slowly is a real risk, not a safe one. Todd frames it from sea time: act recklessly and you get into trouble fast, but hold back too long in a treacherous environment and that is equally dangerous. Most boards are attentive to the first risk and blind to the second.
• Know what you never hand to an outside model. Accounts payable is not where anyone differentiates, so plug into whatever pattern the frontier labs commoditize. A pharmaceutical company's process for developing new products, or a retailer's process for understanding its customer, is a different conversation, and the IP exposure goes beyond whether your data is being trained on.
• AI can produce an answer. It cannot produce confidence. Account balances and password resets belong in a bot. "Will I have enough money that I am not a burden on my children" does not, even though the math behind the answer is exactly the kind of thing machines are good at.

Host and guest

Frequently asked questions

How many companies are actually getting value from AI?

Adoption among large enterprises has climbed to roughly ninety percent, up from about half in 2023, and around forty percent say they are seeing some value. But Todd draws a hard line between "some value" and enterprise value that transforms the business. On that stricter measure, only about six percent of large companies are getting five percent more EBIT contribution from AI, and those companies are seeing close to twice the revenue growth of their peers once you normalize for industry, size, and geography.

Why do so many AI pilots never scale?

Todd points to four recurring conditions: the pilot came out of technology with no business sponsor, the use cases were generated in a room disconnected from the P&L, nothing was built for scale, and no one was accountable for a return. The result is an organization that has been at it for a few years and whose results are still, in his word, ho hum.

What does "start with the P&L" mean in practice?

Work backward from the economics rather than forward from the technology. For each core area of the P&L, identify which economic drivers AI can plausibly influence. Then find the high density workflows, or the high value and high risk ones, that sit on those levers. That is where you build. Todd describes this as the approach that has produced value for the last decade, before generative AI and after it.

What should a company never expose to an outside model?

Anything that constitutes competitive differentiation. Todd's test is whether the process is where you win. Accounts payable is not, so if the frontier labs commoditize that pattern, use it. A pharmaceutical company's new product development process, a retailer's customer understanding, a company's personalization advantage: those he would be very wary of exposing, and he notes the IP risk exists even if you assume complete goodwill and integrity from everyone involved.

Should companies replace their call centers with bots?

His answer is that it depends on what the customer came for. Status of an account, a forgotten password, whether a plan supports an option: get those answered fast, preferably without a phone call. But a customer asking whether they will have enough money in retirement to avoid being a burden on their children is not asking for an answer. Todd's framing is that they want confidence, and he does not see AI delivering that. He also warns against putting friction in front of the escape hatch to a human.

What do the companies that are succeeding have in common?

Four things, in Todd's account. The senior leadership team is genuinely engaged, to the point where the head of supply chain or financial products talks about AI as part of a growth plan rather than as an AI strategy. They move at pace. Leaders start with the economics. And they build for scale, with an operations mindset that makes solutions reusable, monitorable, and controllable at volume.

Full transcript

An origin story that does not lead to Fortune 25 retail

James Carman: Welcome to the Forward Slash Podcast, where we lean into the future of IT by inviting fellow thought leaders, innovators, and problem solvers to slash through its complexity. I'm your host, James Carman, and today we're talking to Todd James. Todd James studied math and computer science, very much like me, another crazy person, at the Coast Guard Academy, which is not the usual origin story for someone who ends up running AI for a Fortune 25 retailer, but that's where he landed. Chief Data and Technology Officer at 84.51° and head of AI at Kroger. He now runs Aurora Insights and his thesis is blunt. Most AI efforts stall not because the tech disappoints, but because leadership can't scale past the first pilot. Todd, welcome to the show.

Todd James: Hey, it's great to be here. Nice talking to you today, James. Thanks for having me.

James Carman: All right, so I've got to get into that thesis, but first I want to hear about the Coast Guard Academy. I don't know that I've ever met anyone that went to the Coast Guard Academy. How did you end up doing that?

Todd James: Yeah, it was interesting. I didn't know much about the Coast Guard Academy growing up, but I knew I needed money for college. And I look back, I think my dad's guidance was get a scholarship or join the military, and with the academies I found a way to do both. So it was coming down to the Navy, and the more I learned about the Coast Guard and its mission, I went into it. It's smaller than West Point and the Naval Academy. It's only about a thousand cadets and it's up in New London, Connecticut. But it opened up a really, really exciting career and some interesting sea stories that came with it. We used to say it was not a great place to be, but a great place to be from.

James Carman: I can only imagine. I see those videos of the ships out there and the big waves crashing over them, and that's where they send you all to go save those people. That's just crazy. I couldn't do that. We'll have to hear about some of those stories sometime.

Ninety percent are using AI. Six percent are getting paid for it.

James Carman: Okay, so the thesis. AI efforts are stalling. It's not a technology problem, it's a leadership problem. They can't scale past the first pilot. Tell me more about that thesis.

Todd James: I think the data supports it. There was some data that came out towards the tail end of last year and early this year, from several sources, and it kind of coincided. We've seen a big material increase in the number of companies that say they're using AI. You go back to 2023, it was right around fifty percent, maybe fifty-five. At the end of last year, about ninety percent of companies. And you're like, hey, that's fantastic, this AI journey, we're finally there. Ninety percent of companies are using it.

Todd James: But when you get beneath the numbers, you see a pretty interesting trend. And this is based on data biased to large enterprises. As you get down into mid-market and below, the data gets even a little bit tougher. About forty percent are saying we're seeing some value. But the real challenge is that some value is different from enterprise value that actually allows you to transform the business. And when we look across large companies, only about six percent of companies are getting five percent more EBIT contribution through their AI investments. And those that are are seeing material benefits. When you normalize out things like industry and company size and where they are in the world, you start to realize there's a pretty material lift in total shareholder return for those companies that are leading. One source said just south of two times higher revenue growth. So you're seeing an increasing and compounding advantage for those companies that are getting ahead, but it's just a small percentage.

Todd James: And the unfortunate thing is that by not moving here, companies are putting themselves at pretty big risk of being disrupted from the outside. That has an impact on companies, it has an impact on shareholders or owners, and it has an impact directly on employees' lives, which can be pretty disruptive.

The J curve, and the dip most companies are sitting in

James Carman: We met at Cincy AI Week and had a great conversation. I gave a keynote there and I talked about this, that lack of ROI. You're right, some people are seeing some value. But the keynote was really kind of a reassuring message. It was about that J curve where you're going to see maybe even a decline in productivity as you go down through that dip in the J, and that's because you're making those complementary investments, the things you need to iron out in order for this stuff to take off. And then as you make those investments and get to the other side, that's where you're going to see it. I think a lot of folks are just squarely in that dip.

Todd James: Yeah, and there's an argument that you can be optimistic about it. But we've been doing artificial intelligence on an enterprise scale now for over a decade. I know people say you go back to the fifties and Turing, but when machine learning came and we started to really industrialize some of that, as well as some of the advanced stats, we've been doing this for a while.

Todd James: I used to describe it as: when you start your AI journey, yes, you start small. You start with a pilot and the results are kind of ho hum. And then if you stay at it, they should be exciting, and they should be meaningful, and they should be nonlinear. The challenge we're seeing is that a lot of companies have been doing this for a few years and it's still ho hum.

Pilots without sponsors, roadmaps without numbers

Todd James: What you're seeing when you recognize it is pilots coming out of technology without business sponsorship. You see people coming together in a room and coming up with the most interesting use cases, disassociated from the P&L, which we should talk about more in depth. You find that they're not building for scale as they look to deploy. And as a result, these organizations aren't really getting value from it.

Todd James: I've sat through discussions with CEOs where they bring their CIO on and they start talking about their AI plan. There is no mention, qualitatively or quantitatively, across a multi-year plan, of value or performance improvement expectations. And I'm trying to think, and I see your face here, James, you see it in the market too. I'm trying to think across my career of any capital investment we've ever made in business where you don't have clear line of sight and accountability for ownership of those outcomes.

Todd James: So I do think part of the big challenge here is we're pushing technology for technology's sake and we've gotten away from some core business fundamentals. If you're going to make an investment, you should see a return and someone should be held accountable for that return. For some reason in AI it's a lot of experimentation and an ability to go back and tell your board, tell your owners, tell the fund if you're private equity, that we're doing AI. There's so much pressure to do that. But it's not necessarily connected to a real value plan.

Boards, earnings calls, and the five meter knife fight

James Carman: And I'm sure a lot of the work you're doing at Aurora Insights is helping coach folks and guide them through those challenges. Do you feel like the market is taking a turn, that people are getting fed up with the experimentation and starting to see the light of, maybe we should make some money off this stuff we're investing in?

Todd James: There are two vectors I'll look at this from. We'll come back to basic human dynamics, but let's start from a business angle. From a business angle, yes, you're seeing boards put more pressure on the business from a competitive and a return perspective. Although a small number of boards, when you talk to them, can actually articulate what the plan is to get there.

Todd James: I've spent time over the last year with the capital markets community, some of the research groups. The question I've been working with them on is how do you translate a CEO's earnings call claims around AI into understanding whether there's actual proof around it? What you're seeing on these earnings calls is increasing pressure. The investment community is looking beyond "they talked about AI" to "where can I see it in the P&L, how is it showing up, and how do I have confidence in their ability to deliver?" So across public markets, we're seeing an increasing level of pressure coming from the investors, translating into the fiduciary accountability of the boards and them putting pressure there. I'm also seeing and hearing individual CEOs who will call up and say, look, I've been investing in this, but I'm not seeing value. So I do think there's a shift. But depending on where you're playing in the market, it differs based on the type of the company, the size of the company, the industry. There's still, as I would call it, a lot of arbitrage across the market around the sense of urgency and the preparedness to move.

Todd James: The other thing that's interesting: I was in a meeting in New York City two months ago with a group of CEOs at a large fund and we were talking about AI. Not universally, but you could see a legitimate concern and uncertainty. One of the people after the discussion translated it into fear. They've seen fear. I think there is a fear of the unknown. You hear people have invested and they're not getting a return. Your company needs it, but you don't necessarily understand what it is. So do I lean in on this, or am I looking just to make it to the next round?

Todd James: It reminds me, right after I got out of the military I worked as a consultant and one of my clients was the Army. I remember a gentleman telling me, as I was trying to talk strategy with him, "Look, Todd, you can't look down range when you're in the five meter knife fight." I think a lot of these CEOs right now are in that five meter knife fight, and it's a big risk to move into this area, which is creating drag. But we all know in business, if you don't address a problem, if you don't take advantage of a capability, it doesn't go away and it's going to impact you one way or the other. The hesitancy to act is as big a risk, even though we're paying more attention to the hesitancy to act too fast and the risks associated with that. And I think it's going to catch a lot of these companies up.

Is the capital spending a lab problem or everyone's problem?

James Carman: Something interesting as you were talking about the recent earnings calls. A lot of these companies are making big capital investments and a lot of those stocks took it on the chin a little bit, because the investors are like, wait a minute, you're giving all our money away to these data centers. Is that a lab-only phenomenon, or are we going to see capital investments like those big chunks coming out of returns for other stocks outside of the labs?

Todd James: Well, we've seen a lot of investment. You've seen a little bit of a slowdown in the market with some of the larger technology stocks, though I'm always careful with the market, because a slowdown today could be a buying opportunity tomorrow.

Todd James: Two things come to mind for me from experience. One is that this is a fundamental transformation, similar to what we saw at the turn of the century when people started using electric power to recreate how they think about manufactured goods and assembly lines. You go from factories built near streams with a single gear apparatus driving everything.

James Carman: The pulley systems and the axles and everything.

Todd James: Right, and now with electricity we can do a lot of things on cheaper land, on a flat plot. If you view it that way, there's a lot more runway from an investment window perspective around what's happening with these labs and the upward potential for continued advances and consumption. What's the upper bound on more intelligence? I know it's prediction, right, it's providing recommendations and predictions of the future, but let's just call it intelligence. What's the upper bound on that? I think that's a question we all have a hard time answering, because theoretically it's nothing.

Todd James: Now the flip side, and you probably remember this: remember all the fiber build out? Once again, a major shift. We were going from localized network computers to opening up the internet, e-commerce starting to drive more traffic, and where are we going to put it? So we need to build bigger pipes. And if you were in a city that had a lot of these telecoms, you remember the absolute reckoning that happened, the number of companies that disappeared, the amount of workforce that was freed up for other opportunities, which is a nice way of saying it was hard to watch your neighbors go through that back in the day.

James Carman: It's a very nice way to put it.

Todd James: So I think we've got to balance that. But right now I am personally bullish on this over time, not the day-to-day cycles or year-to-year cycles. This is fundamentally transformational.

Todd James: The other thing we need to look at is, don't just look at the models at the frontier labs. The opportunities beneath that are significant. Think about a run-of-the-mill high-end GPU, about a hundred miles of copper in it. So you start with the minerals, you look at the rare earths, you have that capacity that needs to build out to support it. You have the power for the data centers, you have the data centers themselves, you have the infrastructure companies fueling the models. You need to look across the entire value chain. In the near term, if you were to ask me, there's probably more opportunity in the infrastructure that still needs to be built out to catch up. But then again, I'm not your investment advisor. I worked at Fidelity for fifteen years. Call them, it's a great company.

What 84.51° actually does, and the flywheel behind it

James Carman: One unique experience you have that a lot of folks don't is, as you said, we've been doing AI for a long time and you've been in it. You worked for 84.51°. For folks who don't know, I'll give my layman's definition: they take customer data and do customer analytics. They're an offshoot of Kroger and they do that customer analytics work using artificial intelligence and machine learning for Kroger specifically. That's the gist of it. You may have a deeper insight.

Todd James: That's a great way to put it. If you go to 84.51°'s website, they are a retail media, insights, and data science company fueled by the Kroger data asset. It takes data from Kroger based on a customer data asset and data associated with the operations, and it monetizes it. It has an insights business. It helps companies like your major brands understand the path to purchase. It has a retail media business. When you're looking at retail margins, they're about three to four percent, grocery retail about one to two. Retail media, depending on how well you execute, can be like forty to seventy percent. So these are higher margin businesses. And then they have some other things, a venture capital arm that uses some of the data science. But basically it's taking that data and traffic data from Kroger and returning profit and more traffic back to Kroger. That's how the flywheel works.

Todd James: And you're seeing it across retail, and to be honest that's probably a longer discussion. These new profit opportunities across companies. I've seen it in legacy media and have had some discussions there, where you have a rich data asset and a business model that's being commoditized, and you have an opportunity to create new profit streams that operate at a much greater scale and a much greater profit margin, which enables you to do new things. I think companies need to start thinking that way as they look at artificial intelligence. Not just how can I improve today, but how does an enriched data asset and an analytic capability allow me to open new profit streams that are adjacent to my current business, but also complementary, and create a new economic structure for my company?

Where generative AI has actually paid, and where it has not

James Carman: Your background is the unique aspect for me. You worked for 84.51° and you were head of AI for Kroger. But the AI you were doing is fundamentally different from what people are trying to do today with generative AI and language models. The models you were using for the analytics you were doing were purpose built for that data. They were trained, you can measure whether they're doing a good job. With these language models, they're auto-regressive, they spit out the next token, okay, that looks roughly right. And then they're trying to generalize this language generation to all these different problems where you can't really measure whether it's doing the thing exactly like you want it to.

Todd James: It's interesting, I was thinking about that for whatever reason this morning before the call. I look back across what I'm seeing in the market now, but from leading AI at places like Fidelity and then at Kroger. What I can tell you is that with generative AI, the places where I saw it delivering meaningful EBITDA-type improvement were when it was used in combination with some other form of predictive or optimization science. A few exceptions with some of the paired programming, the copilot programming. I've seen examples and I've taken budget bets on those that proved to be accurate, where you're dampening some of your growth projections for what you would need in the future around engineering. Around the programming I've seen some benefits, when you're able to manage them and do them in a highly measurable shop like IT.

Todd James: But one of the biggest mistakes I keep seeing, and I see it in companies and I also see it in private equity funds, in those operating groups sometimes, is the idea that we're going to deploy chatbot and copilot and we're going to get benefits. And you're like, well, where are you going to deploy it? How are you going to measure it? "Well, it'll make finance faster." So how much faster? You're making an investment without a budget bet.

Todd James: Yes, you're making individual people's jobs easier, and I think that's fantastic. But if you're going to do that, call the moment. It's like, I'm buying them Excel so they don't have to use green graph paper. Right? Cool. Call the moment. But if you're doing it around value, you need to be able to measure it. You need to be able to look and say where the targeted opportunities are. And you usually need to make some investment to structure it so that it's more useful than what you're just getting from a frontier lab. And over time, if it's valuable, you probably want to find a model that's a bit cheaper.

The leaky bucket

Todd James: But this idea of democratization, that I'm going to get savings without any kind of tracking and measurement, or I'm going to get increased revenue and benefits, where have we ever received value without measuring it? I've heard it called the leaky bucket. I made an investment here, but I'm not seeing it. Where's the value going? I'm sure there's value, there's more efficiency. Well, maybe they're doing something else. Maybe they're covering up another problem that the company should be focused on fixing, and because they're more efficient they're able to spend more time there. Maybe they're playing on their phone.

Todd James: I want to be optimistic, but it doesn't matter which of those it is. Whether they're playing solitaire, which I hear, I play too much on my phone in the evening, I've got to stop that, or they're actually going and fixing a real world problem or developing something, you still have to know how much time's freed up so that you're putting the right strategic focus behind where you should be reallocating it.

James Carman: Because if you free people up and all they're doing is going and doing non-value-added stuff with all that free time, that doesn't help anything.

Todd James: It doesn't help, and I would even say, what if they're doing value-added work on something that is becoming increasingly inefficient, but you're not seeing that problem? You're not seeing the benefit because they're going over to fix this other fire. Well, you want to know where they're going and you want to know how much of their time is there, so that you can actually say, hey, we should fix this fire. That's a big risk too that I've seen.

Purpose-built or commoditized: where the value pool sits

James Carman: Right now all the businesses are just kind of trying to general purpose leverage this AI thing, specifically generative AI. I've called it the papaw paradox. My grandfather would have known about generative AI before a CEO did in the boardroom. We had ChatGPT in our hands, it was on our phones and people were playing around with it, tell me a joke, those sorts of things. It was exposed to the public first before it made its way into business. And so it's catching on, it's catching fire, and now we're like, okay, we've got to apply this to business. What I'm swirling around here is, do you think we'll get back to what would be considered traditional AI, where we're realizing okay, these general purpose things are great, but I really need some predictability. I need this to be purpose-built for the thing I'm looking to do, because all of this non-determinism is just janky.

Todd James: I think there are two things we'll see. People have been doing AI, using AI. I've heard one term, AI-ing it. And you go back to the stats and you're not seeing the value. I'll talk purpose-built, and then I'll also talk about commoditized.

Todd James: For companies where a commercial solution isn't a really good fit, or where it's proprietary, you're going to see organizations lean in and build competitively differentiating solutions. And what I would say is, the good ones are going to look at their P&L. They're going to understand, for each core area of the P&L within their business, what economic drivers AI can actually influence. And then they're going to find the high density workflows, or the high value and high risk workflows, that are impacted by those levers. And they're going to build out solutions. That's what the companies that are seeing value have been doing for the last ten years. Without generative AI, with generative AI, with a commoditized capability of generative AI through the frontier labs. That has been the approach to value, and I don't see that changing. As pressure goes on, we'll see some of that discipline return, or we'll see the companies that don't have that discipline no longer be a factor, and the ones that do take space in the market.

Todd James: Because at the end of the day it's this simple. Since the fifties we've been doing process optimization and improvement based on Deming and Juran, lean, six sigma. In the seventies we started automating those things that were yes-no, deterministic decisions. And what was left? What was left was individuals, customers or employees, making judgment decisions. That's the value pool. Making the customer's journey more seamless so that they can have a bigger impact, or making your associates more efficient and reducing friction from their workflow so that they can have a bigger impact and grow as you scale. Those are the two pools, and companies have to look around and see where they're at. Where are the high density workflows, where do we have data to go after it? When they see that, they realize it's a pretty meaningful opportunity to drive more value.

Forward deployed engineers, and what you never let them touch

Todd James: So that's thing one. Thing two: there was a six and a half billion dollar investment by Anthropic and OpenAI separately earlier in the year around forward deployed engineers and consultants. There's a lot of hypothesis about why that's happening. But clearly, if you get a critical mass of effectively deployed forward engineers and consultants with expertise in how to solve business problems with your frontier lab models, you should be able to drive up utilization of those models pretty significantly. But you should also be able to accelerate the pace of change in the market around AI enablement of process flows, both on the customer and the associate side.

Todd James: You could see a scenario where, if you were one of these frontier labs and you have the forward deployed engineers, you start crosswalking private equity firms and improving all the finance processes and become finance as a service. I'm not saying the whole finance group, but you could start to create non-competitively-differentiated patterns where the intelligence begins to get serviced, it's much more accessible. I think we're going to start seeing that on more of a commoditized basis across the market as well.

James Carman: The forward deployed engineer absolutely scratches an itch that the market has right now, of not understanding the art of the possible and having people that can come in, look at the processes and say, how about we rethink this? Let's take a step back. Let's approach it in an AI-native fashion. As you were pointing out, like the electrification era. We had factories which had pulleys with steam engines attached to them. And all we did was say, okay, let's get rid of the steam engine, let's put an electric engine on that. And we didn't change anything else until later, when we figured out, let's put electric motors on all the machines and then we can move them around wherever we want. It's that idea of, let's rethink this and reorganize how we work.

Todd James: I think it's an accelerator if deployed effectively. And there are others, right? You've got other consultants, other firms looking to do it. But I think that investment is a signal you shouldn't ignore and it could be a real accelerator.

Todd James: The other thing you've got to keep in mind is that with acceleration, you increase the pace of your absorption of risk. If you're a CEO or a CIO in a company right now thinking about this, you should be really excited, because it could help your organization get expertise and capability you may not have. But you'd also better be thinking about what you own and what you don't let them touch. There are IP considerations associated with how these models work that go beyond, are they sharing my data? For competitive proprietary processes, that's probably not your best path if you're looking at the viability of your business in perpetuity.

James Carman: And I'm not saying they did it, but Anthropic partnered with Figma. They had someone from Anthropic on the board of Figma, and then all of a sudden Anthropic has Claude Design come out, and Figma's stock goes, you know. So there are murmurings.

Todd James: There are stories out there. What I would say though is, even if you look at how the models work and you assume goodwill and integrity, there's still IP risk that you've got to be careful about. If you're leaning in and don't know what you're doing there, you may be fast and better today and competitively compromised tomorrow.

James Carman: And there's also, it's compromise, but it's also, how do you differentiate if everybody's using the same models to make the same decisions over and over? There's no IP.

Todd James: That's it. Take accounts payable. Most companies aren't differentiating on accounts payable. So if that pattern starts to get really well understood within the frontier model landscape, I'd plug into it. I'd use it. I'd take it. But your process for developing new products at a pharmaceutical company, your process for understanding your customer at a retailer, your process for creating better personalized experiences than your competitor, those are the kinds of things I would be very wary of exposing to that ecosystem.

Todd James: Here's the good news. If you're a CEO or a CIO, whether people are acting with good intent or malintent, however you want to look at it, the risks are there and they're yours to own. So how you walk through those, how you assess them and the decisions you make, doesn't really matter what's happening in the background. The decision set's the same for you.

Which industries are pulling ahead

James Carman: With the work you're doing at Aurora, you're not in one industry per se, you work across industries. Are you seeing any patterns from an industry standpoint? Where are some industries excelling and really leaning in and getting a lot of bang for the buck, and others where they're not?

Todd James: I'd say it's tough to do it by industry. We'll start with industry, then I'll get into the patterns that I think probably matter more, because within industry, you go down far enough in the market and everything I'm about to say starts to break.

Todd James: From an industry perspective, I'm seeing it in areas where data is either highly valuable or you have high velocity, high volume data. Historically the highest monetary value per unit of data has been in healthcare. So there's an economic return coming from that. In retail, when I moved over to Kroger, I'd heard about it, I knew about it, but until you get there you don't understand it. The velocity and the volume of data in retail is incredible. Using my Coast Guard analogy, when you've got a lot of water going at a fast pace against the rudder, it's easier to turn the ship. It's the same with data.

Todd James: The other area is financial services. They generally have such big margins that they can do a lot of investment, and a lot of strategic investment. That's an industry where you have a little bit of an opportunity to say, I want to drive value now, but I also want to figure out how I can change the industry entirely in three years. You see a little bit more of that flexibility. Those are probably the three industries. The fourth one you would look at is obviously what's happening in defense, where the nature of AI and these new technologies is changing pretty rapidly.

Four things the companies that are winning have in common

Todd James: Now here's the thing though. For me, the pattern that I see not work is, let's figure out how to deploy frontier lab models without a broader lens on full AI capabilities and without business cases. Back to that example: you see a roadmap, it's on a great slide, it's got multiple years, and there's absolutely no value associated with it, or even operating improvements. I'll give you credit if you say we haven't figured out the dollars yet, but we know it impacts this lever and this is our big lever in the manufacturing environment. Okay, probably worth a try. But when you're not even doing that, that's a pattern I see across portfolio companies and funds, and I'm also seeing it in the market, especially as you go down to the mid-market where there's pressure to act but not as much capability to act on your own.

Todd James: The companies that are really making this work, I see four things. One is the senior leadership team is engaged. The companies I work with now, the C-suite's engaged, because if it's not, you're not really transforming your company. You're doing point solution projects. Hire a company to come in, put a point solution in place, and realize a point solution's worth of value. Where you have the C-suite involved, they're rethinking their strategies, they're rethinking how the business is structured, they're rethinking their financial and economic structure. For me it's not, do I have an AI strategy. It's when your head of supply chain or your head of financial products stands up at a board meeting and talks about AI, but talks about it as part of their growth plan and what they're going to achieve and how they're going to hit their metrics. Those are the companies that are really excelling, because it's embedded in what they do and they realize it's just another tool.

Todd James: Two, pace. They're moving fast. That risk of waiting too long can have a big implication on companies. We learned early in the Coast Guard, if you act too aggressively you're going to get yourself in trouble. But in a treacherous sea environment, if you don't act fast enough, that's just as dangerous. The companies that have figured that out are doing better.

Todd James: Three, leaders start with the economics, and we've talked about how you look at the P&L first.

Todd James: And then four, you build for scale. I love when an operations group gets involved, because they understand that these solutions need to scale, whereas I think some of the analytical and mathematical background isn't as oriented to scale. But when you take that mindset of, how do I industrialize this capability, make it reusable, make it monitorable, make it so that I can drive this in high volume and velocity in a way that's controlled and valuable and measured across the organization, those are the organizations that are seeing value. Because an individual algorithm is going to impact a decision point. Transformation happens when you impact all the decision points. That's what we're trying to do, activate them or better inform them through data. That's the aspiration.

What happens to us when the shallow work disappears

James Carman: One of the things happening with AI is that in the past a lot of us, knowledge workers, would have some deep work time. I'm actually reading the book Deep Work right now, so I'll use his phrase. But it would be interleaved with more shallow work, more mechanical things you don't really have to put your cycles on. AI is taking away some of that shallow work, especially in software engineering. You're not clickety-clacking on a keyboard. So you're in that more cerebral land more of the time, and people don't realize that's very exhausting. So how is that affecting what we do outside of work, in our free time? How do we busy our mind?

Todd James: It's interesting. I do think it's going to change the nature of how we spend our time. It's funny, probably years ago I had a discussion with someone in a role with a high cognitive load. They were a very cerebral person, and you just assume they go home and they're reading literature. The response I got was, by the time I get home I am so exhausted mentally that I like activities like lawn mowing, gardening, exercising, sitting back watching a game and talking to friends.

Todd James: I find that in my own life too. On a day where I'm out and having a lot of conversations with people, nothing is better for me than to come home and read a book and get into literature. My reading's not generally easy. I read philosophy, I read source materials around history. That's my thing. Cool or weird, either way, that's what I do. But on a day where it's really hard and you're working on concepts, it's the gym, it's the game, it's yard work, it's rearranging the garage. It is something where you just don't have to think.

Todd James: I actually think it could be good, because we've become so accustomed to coming home and closing our doors. Maybe the idea of getting out there because we've had such a high cognitive load day and interacting with people would be a good thing. I talked to someone recently and she was telling me how her generation now, because of all the high cognitive load, they all like to knit, and they're getting together to do that. Not something we would have talked about twenty years ago as a trend across a group of people.

James Carman: It's getting some outlet of more mechanical, non-deep activities where you don't have to use your brain. I do think that is a human cost we're not really reckoning with on a broad scale. As we say, we're giving people an opportunity to do more high value work, and that usually means thinking, cerebral type stuff, and taking the mechanical, the rote things out of our day-to-day. I do think that's going to have a human toll and we're going to have to adapt over time.

Todd James: Yeah, but it could be good, and I'll leave you with this. What if we spent more time having pizza on lawn chairs in the front yard with our neighbors, volunteering more at church, showing up to help out at the charitable pharmacy, taking walks with friends around the neighborhood? I actually think that if we do this right, and it goes back to those knitting circles, community at arm's length is something that could be a very good thing for individuals and for society, especially as you see how we've devolved into people pulling into their garages, closing the garage door, and they're out for the night and really don't know their neighbors in a lot of cases.

James Carman: I think we would be better for it. I think it could be a nice renaissance period for humanity. You come home, you never see your neighbors, until you do something that makes them mad, and then you see them. For me it's a matter of a readjustment period and rethinking how we do things. Because I'm one of those tinkerers. I go home and I code. I like to do that. But if I'm doing that all day and then I go home and do that at night, and the way we're coding now, that's just a lot of mind work.

Todd James: And I would say too, if you're only doing one thing, you get stale. You need intellectual diversity.

James Carman: Heck yeah. I've got to hear about these history books. I hated history in high school, I didn't want to do it, but as I get older I'm digging it and I want to read about the history. It just blows me away. What happened? Why wasn't I interested when this is what I was supposed to be doing?

Ship It or Skip It: leaving the enterprise after thirty years

James Carman: All right. So the next segment on our show, this is kind of our take on hot or not, we call it ship it or skip it. We alluded to this earlier. You recently left about thirty years in the enterprise, whether consulting, Fidelity, or Kroger. Now you're out there on your own, hanging out your shingle and doing your own thing. How is that? Is it something you would say to people, man, I'm having such a blast, absolutely do it if you can, this is so much fun? Or is it, yeah, I don't know about that?

Todd James: If it's ship it or skip it, I would say for me it was a ship it. I have wanted, since I left the Coast Guard, to be an entrepreneur, and I've been fortunate on two fronts. One, I married a woman whose father was an entrepreneur and fully understood the ups and downs, what it's like when you have to cut off cable to pay the electricity so you can hit payroll. Her father went on to be successful in construction and roofing. But she understood what that meant and was always giving me a realistic picture that it's not all roses.

Todd James: The other thing is that every time I was about ready to do it, I got a great opportunity. What I would say is, make sure it's something you really have your heart in. I did it realizing that you can talk about wanting to do something your whole career and never do it. What kind of person is that? So I made the jump when I was in a very good role at Kroger, just because I knew at some point you stop talking and start doing. I'm energized to do it. I'm excited about doing it. It's something I've wanted to do for a while and I'm committed.

Todd James: What I would say is, it's not all roses. It's ups and downs. It's tough. It's lonely at times. You go from running large teams, thousands of people, to all of a sudden you're the individual who has to walk in and figure out how to fix your printer.

James Carman: You're the IT guy too.

Todd James: And business development's different. Even when you're doing relationship-based and word of mouth work, you weren't doing a lot of that inside a corporate setting. So it's a change. I would talk to a lot of people and do a lot of exploration before I did it. There's a financial aspect you should probably consider that will vary your stress levels as you're going through it. But for me, I'm learning a lot, I'm energized, every day it's something new. I see the direct impact of the work. And I'm very passionate about it. So I'm glad where I'm at.

Todd James: But there are days. The best example I had, one day Trish, my wife, came in and my head's on the desk, which is uncharacteristic for me. And she's like, what's wrong? And I'm like, I'm selling to me. You get in there and you realize you're talking to busy people and they're all excited and they get off the call and they're like, let's meet again next week, and then their admin calls and says he's not free for four weeks. And I remember that, because that was my life. But it's different. So I'll leave it there. I would say, know yourself, explore it. You only get one life, so make it meaningful and make it interesting. But make sure this is what defines interesting in your life before you do it, because there will be ups and there will be downs.

James Carman: Make sure the rewarding parts are rewarding enough to get you through those head-on-your-desk days.

Using AI without becoming vanilla

James Carman: What about AI and going it alone these days? You've got AI as that copilot, that thinking partner. Does it make it easier?

Todd James: If you do it right. Around some base activities you can really get easier.

James Carman: Fixing your printer?

Todd James: It did help. But even things like, you're always researching and you're always learning. So research bots, quality research bots that ensure provenance, then surface things up in a way that you can review and figure out what's interesting, and then go to the source material to validate, as opposed to spending all night googling like we were ten years ago. Those things are helpful. It's helpful to iterate.

Todd James: What I generally try to do on the things that are really important, because you can get sucked in, it can actually become a taskmaster if you're not careful. You keep going down and you'll get into sitting there for an hour and realize, why did I do that? My original idea was good. AI can start having some patterns that drive you up a wall. So what I try to do is I'll use pen and paper on the things that are important. I'll put time into thought and discernment. And then I will use it for validation. When I say validation, it's not, hey, what do you think? It's, here's the market, or here's a particular opportunity, here's what we should be checking, what are you seeing out there, are these theses valid, are these theses novel, why or why not, give me evidence, give me proof. And then some final questions like, would you have approached it differently?

Todd James: But if you just start with a blank slate, what you end up looking like is everyone else. Even some of the words. I used to love the word "reframe." You can't use it anymore, because every time I get an email from someone with it, and you'll get three or four a day, it's AI generated. Or patterns. If you want to list things out, you can't do patterns of three anymore, like X and Y and Z. It's got to be X and Y, or A, B, C, and D. But if it's X and Y and Z, you know it's AI.

Todd James: What's frustrating is that people used to write and use some of this. But unfortunately, if you're just using AI for the things that really matter, you're going to be the ice cream store that looks like vanilla ice cream. And vanilla is great, especially Graeter's, as we all know. But you go into an ice cream store and all of us want to have exciting flavors. If you're just using AI, you're going to struggle to have exciting flavors, using the ice cream analogy. For those of you not in Cincinnati, Graeter's is the best ice cream in the world.

James Carman: I absolutely cannot agree more. I think when Oprah did that, I'm sure they saw an uptick. Oprah said on her show one time that it was the greatest and they went bonkers. Their sales went through the roof. They had to ship people all over the country.

Todd James: It's really good, but there is no diet in the country that allows you even a spoonful of that delicious ice cream.

Should you replace the call center with bots?

James Carman: What about replacing, a lot of folks are doing this, completely shutting down their call center and just having bots take over. What do you think about that phenomenon? Is that a good thing from a customer experience standpoint?

Todd James: I think it depends on what you're doing. I was heavily involved in AI and call centers at Fidelity. One, it's all about customer expectations. The most important thing with your multi-channel, omni-channel, which we've been talking about for years, is meeting the customer where they're at. AI provides an opportunity, around a lot of interactions, to meet a customer where they want to be and where they're at. And if you target certain audiences and certain populations for certain types of problems, that's probably all they ever want, especially for tech products.

Todd James: Where you have to be careful is that meeting the customer where they're at may in some cases go beyond a chatbot. So I'll look at financial services. Status of my accounts. I forgot my password. Does my 401k plan support this option? I don't need to talk to a person. Get it to me. Get it quickly, preferably without even having to pick up the phone, doing it over chat. That's great.

Todd James: "Hey, I want to make sure that I have enough money that I'm not going to be a burden on my children in retirement." Probably not an AI-enabled answer. In fact AI provides a great capability to do the advanced mathematics, like millions of Monte Carlo runs, to come up with the right scenarios. But the person empathizing across becomes even more important for some of those interactions. So, you said I had to do ship it or skip it: it depends on who you are. You've got to stay anchored in business fundamentals, and the most important thing, meet your customer where they want to be met and use the approach that works.

James Carman: Increasingly a lot of the population is at the point where they don't want to talk to someone. But you can't do the old IVR approach, press one to do this, press two to do this, and go six levels deep. If you can get them where they need to go quickly with AI, which you can, I agree, that's a better experience.

Todd James: If you want an answer for a lot of stuff, AI is fine. But a lot of people don't always just want an answer. Sometimes they want confidence. And I don't see AI delivering confidence. When I say that, it's someone looking across and saying, I get it. You saw what happened with your parents. You don't want that to happen to you. That's hard for AI to replace.

James Carman: I think those escape hatches, a lot of companies put friction in front of them to get out of the AI-driven experience. And it's like, just give me a human being, please. Sometimes I'm just at that point and you're like, whoa, but you could look here, or you could do this. Nope.

Todd James: You're pushing the button, the pound or the zero, or yelling, representative, representative.

AI's PR problem, and the data center fight

James Carman: The AI world right now is kind of framing itself, and there's some doom and gloom, there's going to be no jobs for anyone. It feels like AI's got a PR problem. How are you feeling about how the AI industry is representing itself to the country, to the world right now? Are they doing a good job, or could they do better?

Todd James: To be honest, I think there's opportunity there. I think there's low trust. If you look at the core, the frontier labs, I think there's concern, I think there's relatively low trust. Some of that's earned, some of it's unfair. But if you look at general sentiment, that's probably a pretty safe statement to make.

Todd James: When you look at individual applications and how companies are talking about it, think of some of the exciting innovations coming in healthcare around being able to detect patterns and get earlier to cancer prevention. Hey, we all like that. We like taking less time to build our grocery lists. So there are cases where it's going well.

Todd James: One area where there is tremendous need, from a geopolitical and future viability of our way of life standpoint, and where the tech community has done a horrible job, is data centers. You see a lot of different states pushing back on data center builds. There is an essential nature for us as a country, from an economic standpoint and an ability to sustain our way of life, to be able to advance these capabilities. And that story is not getting across at the company level. It's not getting across at the societal level in a way that's probably necessary for where it needs to be today.

Todd James: And by the way, with that needs to go some changed behaviors. The data center companies could have come in and said, look, I'm not going to drain your water, I'm using closed loop cooling. We'll use cameras that don't shine lights. I'll plant those quick, skinny trees that grow so you don't see it from your neighboring farm. And we're going to subsidize your electricity bills because we understand what this takes. They could have done these things.

James Carman: They could have given a what's-in-it-for-me for the communities a little better.

Todd James: A little bit better, before becoming the bad guys and gals.

James Carman: It's like the boogeyman now and everybody's got this knee-jerk reaction, anti data center. And some of this stuff, it doesn't matter whether it's true or not, it gets out there and takes on a life of its own. Like you said, the closed loop cooling, they don't do evaporative cooling. We're not taking everybody's water.

Todd James: Yep. They're data companies. They can measure. They can report on it. They can offset it with benefits. There's a lot they could be doing. And you don't need bright lights. You're sitting there three farms away and you don't want to see light. I'm not sure they need bright lights other than really localized for safety. Forward operating bases aren't lighting up everything around them. There are other technologies that allow you to have cameras that work in the dark.

James Carman: Right, they don't look like a football stadium. I'm sure they could afford a few night vision goggles for their guards.

Lightning round

James Carman: Now we move on to our final segment of the show, what we call our lightning round. This is the meat and potatoes. We've gone through the appetizer, we've talked about these trivialities, AI and the silly stuff. Now we're getting to the real conversation. These are the hard hitting questions. There are right and wrong answers. We have a complex algorithm that we use to score it, that takes a few months to run, and we'll get you your answer eventually. This is not explainable AI, the algorithm.

Todd James: I'm nervous, man. This is either the day where I realize or lose my dream. So let's get into it.

James Carman: If you were to get a superpower, invisibility or super strength, which one are you leaning towards?

Todd James: I'd like invisibility.

James Carman: Have you ever tasted soap?

Todd James: I grew up in a German Catholic family in Cincinnati, and when I used bad words like all my friends, yes, we tasted soap. Ivory soap.

James Carman: Rightly so. But we had to go get Graeter's to get the taste out of our mouth. That's why Graeter's was invented, I think. That was actually the impetus. When it comes to carbs, you're going for bread, pasta, rice, potatoes. What's your favorite carb?

Todd James: Sourdough bread. Sparingly, but sourdough bread.

James Carman: Black beans or refried beans when you're making a burrito?

Todd James: Well, you can have black refried beans, and that's how you should do it. A little bit of lime juice, maybe a little bit of chipotle pepper sauce in there, a little salt and pepper. That's the answer.

James Carman: Are you a foodie? Do you Instagram your food?

Todd James: No.

James Carman: And this one's an interesting one, we've been asking this for a while now of every guest, if I remember to. Do you like the smell of gasoline?

Todd James: No, I don't. I don't mind it. But if Bed Bath and Beyond comes up with one of those little bathroom air fresheners that you plug in that's gasoline, maybe I'll look at it.

James Carman: Scale of one to ten, how good are you at trivia? I feel like you'd be pretty good at trivia.

Todd James: I'm pretty good at trivia, so give me an eight.

James Carman: How many cups of coffee are you drinking per day?

Todd James: Four. Decaf I don't have to count, because it's not bad for you. I love coffee. There is a simple pleasure of a good coffee. Two in the morning, two in the early afternoon kind of works out.

James Carman: What was your favorite childhood television program? I don't think that's a security question typically, but some of these are security questions, so I try to steer away.

Todd James: I like Tom and Jerry. That was a good one.

James Carman: And they had the little spin-offs. I haven't thought about that cartoon in a long time. What's your favorite type of tea?

Todd James: Ooh, that's tough. Turkish.

James Carman: Turkish, okay.

Todd James: How it's prepared, the fact that it can be eleven o'clock at night and you're drinking it with them, the whole ceremony. It's not much of a ceremony, but the whole way they make it and serve it, and how it's the drink for socialization. I think that's cool. It tastes good too.

James Carman: Now, aren't they the ones that do the coffee where they swirl it around the little pot in the hot coals and it boils in the pot?

Todd James: I didn't see that, but it was with the tea. From what I was able to see, you'd pour a little bit of the tea, then you'd pour hot water on top of it, and there was a way to do it. And it was usually in a glass, not in a cup. It was fun. It's better there than trying to replicate it here, but that is by far my favorite.

James Carman: Nice. I've never been to Turkey. How many hours of sleep do you typically need in a night?

Todd James: I have to get at least five. I'm usually in the six to seven.

James Carman: Six to seven. Not bad. Respectable. Okay, well, that does it for the lightning round. Feel good about this one?

Todd James: I do.

Where to find Todd, and two closing thoughts

James Carman: So if folks want to get in touch with you, what's the best way for them to reach out?

Todd James: Two ways. One, find me on LinkedIn, that's a great way to do it. The other is my company website, aurorainsightsllc.com. I've got a lot of my published content there, and a little bit about what we do as an organization.

James Carman: Any closing thoughts, any parting words?

Todd James: First of all, it's been a pleasure. I've enjoyed talking to you, James, and thanks for having me on the pod. I'll leave the audience with two big thoughts. One is that the days of making AI investments where you don't see and achieve a meaningful return are over.

Todd James: And the second, this is actually an example I wrote about in the book. As I mentioned, in the Coast Guard when you're at sea, you learn pretty early in your career that if you're too aggressive and reckless you're going to get into big trouble quick. But what you also learn, and it's not quite as acute a problem in business, is that if you are too reserved in your decision making, you run the risk of waiting too long, and that can be equally as treacherous and deadly for sailors as it can for businesses. So if you're in a position where you're not sure what to do or how it's going to impact you, the safe strategy is not to wait and see how it plays out, but to educate yourself, engage, and start to lean in. And the good news is, once you start leaning in, there's nothing that prevents you from catching up and differentiating yourself in the market based on it.

James Carman: I feel like your second insight, that whole don't go too slow, don't go too fast, I feel like that's a book waiting to be written. Like, life in the trough.

Todd James: Well, that's a chapter I wrote in The AI Shift. It's funny how the lessons early in your career, even in different environments, start to help you shape the patterns that you're seeing today.

James Carman: I just like the books. What was it, Turn the Ship Around? That was one of my favorite books. I like maritime books for some reason. All right. If you'd like to get in touch with us here at The Forward Slash, drop us a line at theforwardslash@callibrity.com. See you next time.

Ship It or Skip It

Where to find Todd

Todd is on LinkedIn, and his firm's work, along with much of his published writing, is at aurorainsightsllc.com.

Got a topic for us?