1. Podcasts
  2. Talking AI
  3. Why Most Companies Aren’t Getting ROI From AI
Talking AI

17 Nov 2026 · 48m · 23 chapters

Episode artwork

Why Most Companies Aren’t Getting ROI From AI

In short

Why companies fail to get AI ROI, arguing the bottleneck is not tools but organizational change, correct use-case selection, end-user experience, workflow redesign, data scope, and ROI measurement that accounts for adoption and full costs.

Guests

Brandon Powell, CEO of Hatchworks AI and author of The AI ROI Engine; Matt Paige is the host and also Powell’s CEO.

Key claims

AI is an enabler, not a strategy; “check-the-box” adoption (ChatGPT/Copilot) is table stakes. ROI requires changing how work is structured for AI (agents orchestrate tasks; humans supervise). Most enterprises won’t unlock value without organizational redesign (McKinsey cited: ~6% have made org changes). Adoption reality (coverage × usage) and true cost (change management, governance, technical debt) are often ignored.

Notable examples

Hatchworks engineers reportedly stopped writing much code as agentic building took over; a call-center agent example shows better UX via predictive resolution instead of cost-cutting voice bots; GenROI uses AI-guided discovery to prioritize use cases beyond leadership-only workshops; “pilot purgatory” fails when KPIs aren’t defined upfront.

Written by AI. May contain mistakes.

Chapters

0:00

The Importance of Organizational Change for AI ROI

Learn why organizational structure is critical to achieving ROI from AI investments.

“You can have all the best tools, all the best processes, all those things.”
0:25

Exploring AI ROI Challenges

Discussion on the discrepancy between AI usage and its returns in businesses.

“Most companies can tell you how they're using AI, but far fewer can tell you what it's actually returning to the business.”
0:46

Guest Introduction: Brandon Powell

Meet Brandon Powell, CEO of Hatchworks AI and author of The AI ROI Engine.

“So we're going to have a unique opportunity to dig into what we've seen firsthand, including some of the lessons learned along the way.”
1:01

Hatchworks AI's Pivot to AI

Insight into Hatchworks AI's transition to focus on AI and its challenges.

“It's funny, Brandon, I was just looking back.”
1:30

The Decision to Go All In on AI

Discussing the challenges and decisions made during the pivot to AI.

“I do want to start here though, because it was a big moment when we made this pivot, our own pivot to AI.”
1:54

Challenges of AI Implementation

Exploring the structural changes required for successful AI integration.

“Like what went into that decision and what did that look like for us going through it in real life?”
5:01

Lessons from Internal AI Transformation

Learning from the internal changes necessary for effective AI use.

“And the funny thing is, like, obviously we did the rebrand and everything, but under the hood, like we had to change literally how we operate.”
7:48

Reassessing Engineering Roles in AI

Examining how AI impacts engineering roles and productivity.

“We're not, I won't say we're lazy, but we're efficient.”
10:09

AI Spending: Investment vs. Cost

Understanding different spending categories for AI and their implications.

“We can just do so much more now that we have AI as a tool that we can leverage.”
11:19

Five Key Areas of AI Investment

Breaking down five major areas where companies spend on AI and their impacts.

“I think how you approach them and how you consider them relative to it being an investment, a cost center or something that's returning back to the business.”
Show all 23 chapters
14:00

The Importance of Business Transformation

Learn why transforming business processes is essential for effective AI integration.

“But at the end of the day, it's easier than ever to do integrations.”
14:59

AI as an Enabler of Strategy

Understand why AI should be seen as an enabler rather than a standalone strategy.

“to do because of just the way you're trying to fly your plane.”
16:11

Implementing AI in Core Business Processes

Discover practical ways to leverage AI for critical business processes.

“reason and think and can take action and do things.”
20:03

Prioritizing AI Use Cases for Success

Learn how to align AI use cases with business outcomes to ensure success.

“Like Gen ROI was something that we're like, okay, we've done this a lot of times.”
22:39

Six Common AI Implementation Failures

Identify and understand the six key failure modes in AI implementations.

“time it's more than one of these and there's more than these, right.”
28:00

Measuring AI Success: Key Performance Indicators

Understanding the importance of one key performance indicator for AI ROI.

“You should have decided up front how you were going to measure.”
29:10

Defining AI Native Companies

Exploring the characteristics of AI-native organizations and their operational models.

“And I feel like this is another thing that gets thrown around a lot, but what, how do you define that?”
30:39

The Shift from Human-Centric to AI-Centric Work

Discussing the transition from human-centered work to AI-driven processes.

“It's providing the decisions that were made.”
31:49

Smaller Teams and Increased Efficiency

Analyzing how smaller teams with AI assistance can outperform larger teams.

“And it's interestingly, you talked about the middle management layer and that kind of old, going back to the Roman armies, like not much has changed.”
33:56

Leveraging Time Savings for Productivity

Learning how to maximize time savings through AI for better productivity.

“and they were impressive, like what they did by themselves against a team that maybe had three or four or five people in there.”
35:09

Understanding the ROI Engine in AI

Exploring the three levers of ROI in AI and how to calculate potential value.

“So I think it's extremely interesting because you could have two seemingly same use cases and you can get drastically different returns based on each of them.”
40:37

Challenges in Calculating AI ROI

Identifying common pitfalls in ROI calculations and the importance of true costs.

“So if you pay to Hatchworks, it's gonna cost this.”
42:07

The Future of AI in Business

Explore the evolving role of AI agents and small teams in the workforce.

“What's your view on the future of where AI takes us being so close to it?”
A man walking through an airport, asking his phone a question

Ask about this episode

Get the gist, explore an idea or ask about something said.

Claude

On Talking AI, what are the key points of this episode?

Searching your listening

Transcript

Automatic transcript. May contain errors.

0:00You can have all the best tools, all the best processes, all those things. But if you don't change the structure of your team and more importantly, your organization to work for AI, you will never get the ROI.

0:14Matt Paige: Welcome to the Talking AI Podcast, where we talk AI with both experts in the field and early adopters. I'm your host, Matt Paige, and we're here to demystify AI for you so you can get some value from it. Let's talk some AI.

0:29Matt Paige: Most companies can tell you how they're using AI, but far fewer can tell you what it's actually returning to the business. And with the amount of money that most businesses are investing in AI right now, that's a pretty big problem. So what's separating the companies that actually get a return from everyone else? And that's what we're exploring today. And joining me today is Brandon Powell, CEO of Hatchworks AI and the author of the new book, The AI ROI Engine. And he also happens to be my CEO. So we're going to have a unique opportunity to dig into what we've seen firsthand, including some of the lessons learned along the way.

1:01Matt Paige: And we're going to get into why so many AI initiatives fail, what's working for companies that are getting real results from what we're actually seeing out there in the real world, how business leaders can start seeing a return on their own investments. It's funny, Brandon, I was just looking back. You were the first guest on what used to be the Built Right podcast back in 2023. Good to have you back on. Yeah. Thanks for having me. I must've not been very good on that one since it's been three years. That's back on. Yeah. Yeah. Maybe that's it. I don't know. Thanks for having me. I do want to start here though, because it was a big moment when we made this pivot, our own pivot to AI.

1:40Matt Paige: And AI has always been something that we've done. It's always been a service of ours. But we hit a very deliberate pivot to go all in on AI. And at the time we did it, it was not clear that AI was going to turn into what it is today. But let's start there. Like what went into that decision and what did that look like for us going through it in real life? Yeah, look, I think about that moment back in 2024, early 2024. Really 2023, we were having that discussion. We started the discussion. Yeah, ChatGPT had been out about nine months or so, and we had all been watching ChatGPT. And then Claude came along and we were watching this technology evolve faster than any technology that we had ever seen, especially from an adoption perspective.

2:27And at the time, we had a very successful software development business. We were building a lot of really cool products for companies. but you realize that when you watch that product and it being language and coding being language that this thing was gonna change. Now, I don't believe that any of us knew how fast we were about to see the change to how software is being built as well as what's being built. But when I look at the way that we pivoted the business, we took a major risk and that first year was really tough and we made the change, we changed our branding, we had to change pretty much everything about the company.

3:05We had to change who our salespeople were because we had to sell outcomes. Clients weren't just going to buy technology. They wanted a business outcome, which we're going to talk a lot about today. We had to change our CTO. We had to get someone that understood that what's this new way of working going to be. You helped us rebrand everything. We had to rebuild all of our services. And honestly, that first year in 24, we sold Nothing, not much because the market wasn't ready. But now look at us, we can't even handle, we have so much work coming in from where we're playing. And so looking back, it looks like a no brainer.

3:39But when we were in the decision, I was very torn. Do we just go play the cost game and go open up offshore? So now we have onshore, nearshore and offshore so we can play the cost arbitrage game. Or do we just get so good at AI that it doesn't matter where the people are because we're going to be able to do more with less and it equalizes everything. And that was the bet that we played, that we should get as good as there is at leveraging AI so that we can run these smaller teams or even one-person teams, which we're seeing a lot of today, and produce the same outcome or better than you would get from some large team.

4:17Matt Paige: Most companies don't have an AI problem. They have a prioritization problem. There are hundreds of different places you could be using AI, but the hard part is figuring out which ones are actually worth investing in, which ones you're ready for, and what should actually come first. That's why we built Gen ROI at Hatchworks AI. Instead of basing your AI strategy on a few workshops or whatever happens to be in the room, Gen ROI uses AI-guided discovery to gather insights from across your entire organization. It helps us surface the themes, patterns, and connections that are easy to miss. Then our strategists work with your leadership team to turn those insights into a prioritized AI roadmap, a way to business value, feasibility, readiness, and what actually makes sense for your organization.

4:53Matt Paige: So if you're investing in AI, but still trying to figure out where it can actually create the most value, I highly recommend checking out GenROI. Go to hatchworks.com backslash GenROI, or just hit the link in the show notes. And the funny thing is, like, obviously we did the rebrand and everything, but under the hood, like we had to change literally how we operate. And I think we would be not true to ourselves. We'd be frauds if we go to a client telling them to do this transformation, if we didn't do it ourselves. I think that was one of the more painful things, especially with engineering being so core to what we did, but any lessons from that?

5:29Matt Paige: And was there like a specific moment when you realized, okay, things have to change? Yeah. So at the end of the day, we'd been using the technology and I would say the inflection point for us was actually, I would say later last year, around November, December of last year, and then January of this year with Claude Opus 4.6, when building came really fully agentic. And so I had this one moment that I will never forget. And I talk about it in the book. I do these five by ones, which just means it's me and five people. And usually it's like, it's people that are doing the work. So a lot of individual contributors, just so that I can talk to people who are actually doing the work, right?

6:10So most of the time as leaders, we get our information from our direct reports, but think of this as my way of getting customer feedback from our people that are on the ground working. And so it's a January five by one. I'm on there with, I would say, five of our best engineers, three or four principal level been really on the frontier of leveraging AI and how we need to do software development with AI. And there was like this awkward silence. And I've known some of these guys for years. And I was, they were telling me, Brandon, we don't really write code anymore. And I was like, that's interesting because these are software developers.

6:47You don't write code anymore. So what happens when I knew they were using agents, I just didn't know that we had turned the corner where literally there was little, very little code being written. And so I'm thinking to myself, they're like, yeah, we just kick off these agents. The agents run, we come back and we look at them. And at that point we had made some changes to our team size, but I'm thinking to myself, what do these guys do while these agents are running? And so I was like, so I asked it out loud. Cause it's a safe place. I'm like, so what do you guys do while you've kicked off these agents?

7:20You've given it a task that you needed to accomplish. And there was this silence. And I was like, come on, guys, you guys can tell me no one's in trouble here. What do you guys do? And they're like, we play video games. You play video games? Really? So what's happening here? And that was an aha moment for me, because I realized if you don't change the structure of your teams, they can produce the same outcome like in minutes that they were producing in days before. And humans are humans, right? We're not, I won't say we're lazy, but we're efficient. And so if I'm going to be able to complete my work now in minutes instead of days, then how do you capture the rest of that time?

8:02And that was an aha moment for me that you can have all the best tools, all the best processes, all those things. But if you don't change the structure of your team and more importantly, your organization to work for AI, you will never get the ROI because that was my RQ that like, why do we need more than one or maybe two full stack engineers on a project? They're kicking off agents. There's a QA and you have these pieces. We have been running sometimes four developers plus QA in a pod. And so that was the aha moment for me where this thing has gone from, all right, let's use AI as a productivity tool to AI can go and do the task and the outcome or the goal that I give it.

8:46And then my job is to be more of the orchestrator and kind of the conductor over the top of this thing. And so that was a big moment for me. And it's funny now looking back, but I also believe that was a moment where we got even more focused on what is the right model that you need to actually be able to deliver these outcomes for customers. Yeah.

9:07Matt Paige: And it's such an interesting thing because in our type of business and in a lot of businesses, like engineering was the bottleneck. That's always been the bottleneck. You can't find enough engineers. And the funny thing is it's shifted, right? Because now engineers could just do so much more. And part of it was figuring out, okay, what do I do now that I can spin up an agent to do one thing? Well, you can do many other things in parallel as well. But what was interesting is, and we talk about this, like engineering is almost like that canary in the coal mine, because this is not specific to just engineering.

9:39Matt Paige: Everything that an engineer can do and is enabled with AI, the same thing can be done across any domain in a business, product, design. And you start to see this emergence. It's almost like a contagion within a team where they just become like agentic and they're just like, oh, wow, I can do things now. Because I think there's this perception that AI can do my work and I won't have to do anything. But I just do not think it's part of our human nature to just not progress and do more. I think that's where we're seeing now. We can just do so much more now that we have AI as a tool that we can leverage.

10:20Yeah, completely. And I think that what I'm seeing is individual contributors as well as leaders that are those doers and they get it done. They're more productive than I've ever seen ever in my career. But I would also say the counterside is I think they're also tired because we're able to do so much more. But it still takes a lot of brain capacity to be able to remember like all these things that you've got spun up that are happening. And of course, people that have ADHD like me, it's even worse with our 70 different screens. All these different things running.

10:56Matt Paige: But yeah, I totally agree with you. In the book, you break down five different places where companies are spending money on AI. But the interesting point is that the spending on AI doesn't necessarily mean you're creating any values. Like break those down for me, the five different areas where we see folks spending and not that any one of them is bad. I think how you approach them and how you consider them relative to it being an investment, a cost center or something that's returning back to the business. Yeah, look, there's probably more than these five, but this is just five buckets that I think of the world in.

11:33Right. And one of them is licensing. Everyone's selling AI applications and agents for every single thing you can imagine. and pilots and you can license those. And so obviously license costs is something that's there. And then tied to license, but I separated is consumption. So this is things like models, like tokens. It can be APIs. Consumption, and that's a hot topic and we could do a whole podcast on consumption, but I think things are going to have a perspective there around consumption. But at the moment, it's a big deal. I think I saw Accenture this week capped everyone at$29 on spending.

12:10I don't agree with that approach, but I'm just saying that, yeah,$29. I would say our best engineers are using hundreds, if not some of them, thousands of dollars in tokens every month. And anyways, that's a different story for a different day, but the consumption is a big thing that's top of mind for a lot of customers. And then a lot of customers we work with, they want to build something. They want to leverage their data. They want to build an agentic platform. They want to build something custom. I love that stuff because it's always is tied to a business problem, but they are spending a lot of money on making AI custom to their business to solve the problems that they have.

12:48And I think, yes, that build versus buy is up. It's becoming more and more complicated, but it's also becoming so much easier to build that more customers are defaulting to build versus let's just make that thing work that I can buy. It covers 80 % of what I want it to do, but it's not going to get me all the way there. And so more and more customers are building and not just buying. And then there's this whole idea of how do you enable your infrastructure? So things like investing in your data and your data structure and your cleanliness of your data, the infrastructure that this rides on. And obviously there's a huge boom with infrastructure, seeing a lot of clients buy their own hardware.

13:26They're going back and saying, I could put this in my cloud or I could, if I really want to get back to controlling things, I'm going to put it in my own. I'm going to build my own cloud, private cloud, or I'm going to use my own data centers. And so infrastructure is continuing to be there. And of course, integrations in terms of with MCP and stuff like that, the answer is no longer always, let's just get all the data in a data lake or a data warehouse and let's spend the next six to 12 months doing it. Integrations have never been easier in terms of connecting things. I would also say they've never been riskier in terms of what's happening in some of the MCP stuff.

14:01But at the end of the day, it's easier than ever to do integrations. But that fifth bucket, which I think is the one that I see the lightest amount of focus on, but it's actually the most important one. And that's to actually transform your business. And that means redesign. I don't love the word transformation, but I'll just use it here. I mean, redesigning a process from the ground up. Link slate it. Reimagine it. Don't just add AI to what you're doing. Thinking about your change management. That's part of transformation. it's now the biggest single problem in every business is changing how people work.

14:37That's the problem we had. We made a pivot two and a half years ago. We still work every day trying to make sure we're changing how people work. You can't do that on six months and then all of a sudden expect that you're gonna get a return. You're not. You've got to stay at it every single day for years to really change how people work. So those are the five buckets. I think the transformation one is the hardest one to do because of just the way you're trying to fly your plane. And that requires you to redesign the plane while you're flying it, which can be very difficult. Yeah.

15:10Matt Paige: And I think so many business leaders, it's either a directive from the board or they're giving the directive that we need AI, we got to do AI. And a lot of them feel like, okay, we've adopted ChatGPT or Quad or what it is, check the box, we have AI. And then you have the other side of it where folks are viewing AI as a strategy, but it's not a strategy. It's an enabler of strategy. I always go back to Seth Godin's purple cow analogy. Right now, everybody's purple. Everybody's using AI. So that's not going to differentiate you in and of yourself, but it can enable differentiation. I think that speaks to that last bucket, that transform one.

15:48Matt Paige: And that's frankly, it's the really hard work. Like when we go into a client, like that's the hardest part because you truly do need to think about like, okay, how do we do work today? And it's not just let's slap AI onto it and make it faster. It's like we have this new thing that's never existed before, which effectively is metered intelligence, you know, in the form of tokens that's probabilistic in nature and can reason and think and can take action and do things. Should we even be doing things the same way? So I don't know. Are there any experiences you've had either positive or negative or like getting companies to kind of figure out where that transformation is?

16:31Matt Paige: I feel like that's half the battle is figuring out what should I do versus knowing I need to do something. yeah and i agree ai by itself is not a strategy and a lot of companies now their board says what's your ai roadmap and they intermingle that with business strategy and ai is absolutely an enabler of your strategy it's not the only enabler you have a lot of other things part of your strategy that are going to go there and it's definitely i'll use the simple analogy of a hammer looking for a nail. And what happens is they focus on this roadmap on, okay, we're going to implement these use cases, or we're going to get this tool and everyone's going to start using it and we're going to get it there.

17:15But then there's this whole realm around what's the actual value that you're creating for what you're doing. How does it tie to the P &L? That's what I talk about. I'll ask that simple question. Where in your P &L does this initiative that you have part of your AI strategy, where does it change your business? Doesn't, they can't tell you. And a lot of times companies with AI will maybe they're risk averse or whatever, and they'll go try to go, they'll just, I call it dabbling or playing. They'll build pilots or they'll implement it in processes that are not critical to the actual business outcome.

17:50Every one of us has a part of our business that is critical. I'll give example in our, we're in the people business indirectly, right? We have to have great people to do the work. Therefore, one of the most critical processes for us is how do we find those people and bring them into the organization? That is a critical process to us. So how can we leverage AI in that process? So just think about whatever business that is. A lot of businesses have really complicated supply chain, and that impacts lots of parts of the P &O. And we'll be like, why are we, that's a key part of your strategy to differentiate is your supply chain and how you don't have stockouts and how you're able to always have enough inventory and these things, but they tend to dance around that.

18:32Do it, figuring out where AI could help you in that core part of your business, like supply chain is a game changer when it comes to the bottom line. It's a game changer when it comes to your strategies. At the highest level, I agree. And at the same time, like because everyone's using Claude, that doesn't mean that you transform your business. That's just like measuring adoption, just like token usage. Oh, we use so many tokens and okay, that's cool. what's that tied to from a transformation of a business outcome?

19:00Matt Paige: Yeah, it's just table stakes now. And I'll give another example for us internally and you can go deeper on it, but like how we do strategy and discovery is like one of the core things we do because we have to figure out, okay, how can we actually help a client in traditional discovery? I've done it before. You've done it before. It's very painful, right? Because you got to go figure out who, what stakeholders are you meeting with? You got to schedule the time. You got to do the interview. You got to do the analysis of the interview. And we built GenROI, which is essentially leveraging AI to help with that process.

19:32Matt Paige: So instead of interviewing 10 people in the business, 10 executives, we can interview in number of people in the group. It could be 100, it could be 50, it could be 20, and it can be at all layers of the organization. And then you can leverage AI to take all of those insights to connect all the dots. And then you have the deep, more meaningful in-person conversations with the necessary folks. It's just an example of us using AI to enable a core part of our business versus like that in and of itself being the strategy. Yeah. Like Gen ROI was something that we're like, okay, we've done this a lot of times.

20:10How do we help the client prioritize the right use cases? That's why it's called Gen ROI because we want them to generate ROI by picking the right use cases. And too many times the workshop involves just going and meeting with the leadership team. That's great. I love meeting with leadership teams face-to-face. The problem is that in most organizations, just like the example I gave you where I was clueless that my top developers were not coding, and we're not even a huge organization. So the bigger the organization, the more work happens down below, the more ideas are happening down below. So GenRY is a way of also capturing input, not just from the leadership team, but from everyone in the organization or as many people as you want, because a lot of it's leveraging AI to actually collect that data.

20:59And we're seeing more and more that the client has a roadmap. In fact, they have five roadmaps, like one for each organization. And now it's even more important because there is no shortage of ideas of what AI can do. But how do you prioritize those based on business outcome, return? Everything needs to be looked at through that lens, because that's where you stop this chasing shiny objects and every good leadership team is going to have some tension between what the CMO thinks for the top use cases versus what the COO thinks. Everyone's going to have a lean towards AI use cases in their area.

21:35And so being able to come together and blend all that into one roadmap that's singular across the organization and we all align that, oh yeah, that is that those are the top three use cases for us as a business, regardless of if I'm the CMO, the COO, the CFO, it doesn't matter. And I think that we've seen a lot of value just because if you don't, and we'll talk about one of the biggest failures here in a second, but if you don't pick the right use cases, it's over before it even started because you're never going to get the ROI on choosing the wrong use case.

22:08Matt Paige: Let's go into that. So there are in the book, like we've done over a hundred, hundreds of implementations related to AI all the way from strategy to building unique things to integrating things. And you've, you talk about these six failure modes in the book from all of that learning. You hinted at one right there, but go through the six. I think these are critical for folks to realize. And I think once people actually hear them, they're going to be like, yep, yep. That, that one, I agree with that one. Yeah. And look, most of the time it's more than one of these and there's more than these, right.

22:42But these are most of the time One of these can actually think the whole thing, right? So I mentioned the wrong use case. You've got to put the effort in to pick the right use case. Now, I didn't say go and think about it for six weeks and do all this stuff. You've got to move quickly, but you've got to prioritize. Matt, we say this all the time, but building's not hard, but should we build it? That's the question. And if you don't think about, okay, it's not like, can we do this? Yes, we can do it, but should we do it? Is this the right one that we should do? So picking the right use case is critical.

23:15The number two thing is like they ignore the end user experience, okay? I've grown up building products. That's been my whole career. And user experience has always been at the center of it. But for some reason, with agentic AI and classical AI now, a lot of people just forget it. And let me give you an example. We'll work with a customer and they will say, we want to implement agents in our call centers. I say, that's cool. That's a great use case. And then they'll say, talking to humans is one of the most important things we have in our business. We pride ourselves on exceptional customer service, human to human.

23:53And then I'll say, why are we implementing an agent? Why are we going to put an agent in here? So they have to now talk to the agent. Agents are great. And they'll say, because we want to cut costs or we want to drive costs out of the business. And I'll say, okay, that's a completely different user experience. So what about this? what if we leverage predictive AI to figure out why they're calling you in the first place? And we actually try to proactively resolve the issue because you have all the data there. So they don't call you in the beginning in the first place. Credible, right? User experience.

24:27Or when they do call, and Johnny always calls on the third week of the month about his bill, as soon as your customer service rep answers the phone, we already tell them why they're calling most likely and how they can fix the problem. So another metric here is if you just reduce your call time by 80%, you would get to the same place from a cost perspective, but you would have happier customers who didn't have to talk to some voice agent. I'm not saying voice agents are bad. I'm just giving an example of that slight different lens on customer experience completely changes the outcome for the business.

25:02the third one is, and this is like every business, AI is a tool. If you just roll out ChatGPT, Copilot, it's so easy, right? It's just included in your 365 or Gemini. You can do that and everyone can be using it. But if you don't ever change the operating model or like to go change the change of the behavior in your business, it's a tool that you're paying for, just like every other tool you have, but you're never going to actually get an ROI on it. And so then the fourth thing, and this is the one that, this is the one we see probably the most, and that is the client does not reimagine or redesign the workflow.

25:43Most workflows in every business were made some businesses 30 years ago, some businesses 20 years ago. AI did not exist. or if it did exist, we weren't using agents. We weren't doing agentic AI. And so now you've got to really think if I started this business from scratch, what would AI do versus what would humans do? And how do I elevate my humans? And so these broken processes get automated. It just creates more brokenness. And this is such a simple thing, but it never gets fixed. Every client's, nope, we can't do it that way. This is how we've always done it. We just need to apply AI to this process.

26:22because we're checking a box. And that's a bad outcome in the end. Number five, the missing data. Most people understand that are messy data. I want to caveat this with messy data. You only need data for the use cases that you prioritized. Okay. Lots of consulting companies like ours like to come in and say, oh, you need to transform all this data into a data lake and you need to clean it. And we need to spend the next 12 months doing that because that's what you need to do first. That's a terrible idea because you're not getting any use cases to market while you're waiting for your data to be cleaned.

26:59We call it taking a bite of the sandwich. Each use case is a bite of the sandwich. So we only need the right meat and cheese and bread for that one use case. And that data does need to be clean and it does need to be great to get reliable outputs. But you don't have to bite the whole sandwich. You can do it a bite at a time, especially when it comes to your data. And then the final thing is around measuring ROI. Okay. Measuring. So example, they'll roll out something and let's use the customer service example I gave before. What was your call hold time before without AI? Exactly. And what would your call hold time with AI predicting why Matt is calling?

27:40Okay. Did we, do we have an improvement? You have to measure apples to apples. Otherwise you don't know if it's actually being successful. So we had so many clients. They're in pilot purgatory is what we call it. They've got all these pilots. And I'm like, why didn't you either launch it or kill it? What are you doing? And they're like, we're not sure if it's working. You should have decided up front how you were going to measure. You only need one KPI. You don't need 10. You can pick one really important KPI, measure the old way of doing it versus the new way of doing it. And so these are the six.

28:12There's probably more than this. But if you miss on any of these six, it's not going to go well when it comes to getting an actual ROI. And like I said before, most of the time, it's not even one of these. It's usually multiple that we see. Yeah.

28:24Matt Paige: And what I love about the six too, is these are first principle things like these aren't necessarily new and emergent since AI. It's just things that you need to focus on to have success. And you talked about voice agents. I will give you a good AI voice agent use case. We cloned Brandon's voice and created the audio book version of the AI ROI engine, which saved him hours and hours of reading and something, frankly, like you probably would have never gotten to just because you have no time in the day. And it actually did, it did a really good job. The only thing Brandon had to do is when you had to read, like record yourself for like 15 minutes or something.

29:0445.

29:05Matt Paige: Oh, 40, but it was, it was pretty impressive. But we talked a little bit about the AI native or, uh, and company. And I feel like this is another thing that gets thrown around a lot, but what, how do you define that? What does that look like? Because I think a lot of new startups and companies that are emerging today, they're just doing this by default because they're small, they have constraints, they don't have money, and they're working in such a different operating model than every other legacy business out there. So I guess what is the difference between the two and how does an existing business begin to make that transformation?

29:43Yeah. So when I think about an AI native company, the core concept is that we design work around AI first and then figure out what the human should be doing. What's the most important? Today in most businesses that I've seen, all the work is designed around the humans and it's done like that. And like you said, every company in the last year or two out of YC Combinator, they do this from the beginning. And one of the core things that we've seen this in different businesses like Block and stuff like that, they don't ever hire a lot of the middle of the pyramid, right? We've always been working in these pyramid organizations since the beginning.

30:21It's been going around since the Roman armies, right? Where you have this pyramid structure with the leader and a leadership team and middle management, and then all these individual contributors. And I think the core of what we're seeing with a lot of these AI native organizations, They never hire the middle. The middle is AI. The middle is an intelligence layer that everyone's working from. It's providing the context. It's providing the decisions that were made. And it's really hard for an existing business to make that type of change, to go literally from designing work around humans to designing work around AI when you've already got all these humans.

30:59And at the same time, we like to talk to organizations about it because it's the North Star. Think about this. If you're in an industry and someone starts a business and they're going to leverage AI in about half the labor that you do, you're going to have a problem at some point. That is a threat to your business. And so that's why I like to talk to companies about the AI know-to-date organization. The bigger you are, the harder it is going to be for you to get there ultimately. But you've got to take, again, steps towards getting there. How can I put more work on agents? How can we make sure our humans are in these positions where they're overseeing what agents are doing versus doing work that they shouldn't be doing to begin with.

31:38And I ultimately think over the next couple of years, you are going to see some massive companies disrupted by companies that started in the last couple of years that are now AI native from the very beginning.

31:51Matt Paige: Yeah. And it's interestingly, you talked about the middle management layer and that kind of old, going back to the Roman armies, like not much has changed. And what was the purpose of it? It was to disseminate information up and down the ranks. And now you have AI that if you give it that context, it has that detail. And if you think about a lot of the things we do, you don't necessarily need to do them. So you hear a lot about these intelligence layers and things that people are building. I know we build those for a lot of clients as well. But I think the interesting thing too is like this concept of a smaller team.

32:26Matt Paige: The typical team we have going to a client is remarkably smaller than it used to be. And it's a beautiful thing because if you think of large teams, they have built-in complexity because there's all these different lines of communication. It just adds complexity. And we not only have these smaller teams, but now have this forward deployed engineering model. Curious your take there and kind of that evolution of moving to the smaller team mode. And I believe this is going to be the future of how companies operate too, kind of this smaller pod model. Yeah, look, I think for a long time, certain teams performed really well based on how they were able to work together with each other.

33:08And a lot of teams didn't perform well because of that, the interpersonal and the way they work together. And now in these really small pods where agents are doing a lot of the work, anytime you reduce the number of people, you can always move faster. Just we hear the saying, like, I could just do it faster by myself. We're seeing a lot of that where it's just one product owner, one or two engineers, and they're delivering an outcome. And they're actually, depending on how agentic, they are able to move faster than having to worry about blending together your code and my code and someone else's code.

33:39They're able to orchestrate the entire thing while they're thinking about it. And I think that we're going to see individual contributors. We're seeing one person teams outperform entire teams. and we're seeing it. We just did a hackathon, an awesome hackathon where a couple of the teams were just one FDE and they were impressive, like what they did by themselves against a team that maybe had three or four or five people in there. And I think that if you, that coding is just like you said earlier, that's just like the canary in the coal mine. That's like the first one. We're going to see this in finance.

34:15Why do we need all these finance analysts churning through spreadsheets? agents can do that work. Do you still need a CFO? Absolutely. Do you need, still need someone checking and making sure that stuff is right? Yeah. But like, why do you need a human building invoices to send out or doing accounts payable? And I'm not saying that you get rid of all those jobs. I'm just saying that team size of what I, even every organization looks like, it's going to change as agents. I call it agents are starting to move into the front seat and do more of the work and the persons in the backseat, making sure that they're driving and doing the things correctly.

34:49Matt Paige: Well, I think part of that is kind of transitions into the three different levers in the ROI engine. But if you get this time savings, you've got to fill that new time savings. Otherwise, you're still in the same spot you have. You got to increase that productivity. But talk about the three levers, and then we'll get into the actual calculation of the ROI engine. So I think it's extremely interesting because you could have two seemingly same use cases and you can get drastically different returns based on each of them. Yeah. Like you said before, from a time savings perspective, it's just like what I saw with our development teams.

35:28They were saving time, but how do you capture that time? How do you capture that and turn that into something financial? So either they need to be producing more or they need to work less hours that they're being paid for, right? And so you've got to think about time savings is only great depending on what you do with what you saved and being able to make that into a business outcome. And really when it comes to AI, there's three levers that I think about. One is profit, but that's not just like making money from like more revenue. That's also cost reduction. That is a lot of companies we work with, they're PE backed and so they're going to grow and they're going to need to hire more people.

36:06So cost reduction doesn't mean getting rid of those people. It just means as we grow, we can hire less of a certain function. That's another way of doing cost reduction and still creating a more bottom line profitable business. So almost everyone thinks about profit. So that's a pretty easy one. The second lever is risk reduction. Some of the most impactful AI use cases over the last 10 years have been around fraud, around compliance stuff. Humans just aren't very good sometimes at risk. And so having something that can go through thousands or millions of transactions or looking at things and looking at patterns that could be happening and then raising that to a human to decide, is that something we should get into or not?

36:47Risk reduction is a huge thing. And AI is really good at looking at that and creating a really a business differentiator, because if you can, every, every business has a certain amount of risk. And if you can reduce that risk, you can grow and have a more successful business. And then the third one that is the hardest for companies to wrap their head around, but is a game changer when it comes to beating your competition, is decision velocity. Back to what we just talked about with these AI native businesses, their decision velocity is so fast because they're not pushing things down 10 layers of an organization.

37:25they've got an intelligence layer that's always got the right context of where the leadership team is and so i just asked the intelligence layer what was our x what was that what do i need to you know what's tomorrow and i'm getting something immediately and a business that can make decisions quickly is is going to be successful i look at what we've done in hatchworks and we go up against big companies now like accenture and slalom and others and we in the past you wouldn't have had a chance. But now like we were taking business and part of that is because we can make decisions much faster. Our velocity of the way we make decisions is much faster than a massive organization, especially one that hasn't adopted AI like we have.

38:09Matt Paige: Yeah, totally. And so this leads into the actual formula itself. And I think this is beautifully simple in and of itself. We talk Talk through how you think about calculating ROI in this approach. And I know it's hard to follow along for some of these equations. We'll share the book website as well, where you can see all this stuff printed out. But when I think about ROI, you've got your potential value, okay? And your potential value could be like how much money you're going to save. What's the risk that you're going to, you got to put a money to it, right? It makes us this money. It loses us this.

38:44on risk reduction, it costs us$2 million a year because we miss things during risk. Or if we make these decisions that much faster, like hiring or something like that, you got to put a number to it. And then you multiply that times the adoption reality. Okay. The adoption reality and that adoption reality, this is what trips most customers up. Okay. Because adoption reality is the reality, And most people just assume everyone's going to adopt it. And that is not going to happen. It is not going to happen. As much as you want it to happen, it's not going to happen. So what you got to think about is what's the adoption rate when I'm thinking about the adoption reality?

39:25And what is the workflow coverage of what I'm putting AI on? Let me give you an example. If you have a 10-step workflow and AI is helping you with eight steps because two steps humans need to do. We need their judgment. We need their thinking. So now I'm automating 80 % of a workflow. So that's my workflow coverage. But then on those eight steps, how many of my humans are actually going to use it? Okay. And if you say a hundred percent, it's not right. And so then you've got to multiply that. So 80 % of my coverage, 50 % of my people will actually use it. And then that gets you to an adoption reality percent, right?

40:06And so you've got to take that adoption reality and multiply it times your potential value. Because if I have a potential value of a million dollars in savings, but my adoption reality is 40%, what does that turn into? Now I'm already at 400 ,000, 400 ,000 already before I've even taken out my true cost, okay? So once I do those, my true cost is what is it going to cost me to implement? and people also get tricked up here on true cost because it's not just what it's gonna cost you to build it. So if you pay to Hatchworks, it's gonna cost this. What is it gonna cost me to run? What are the tokens?

40:43Everyone usually gets all that. But what they forget in true cost, change management. Good luck calculating that. What is it gonna take to actually get the adoption I need? And I need to budget for that. I need to budget for governance. No one likes to talk about governance. It's really important, especially in some of these industries. The technical debt, you will have failure. You will have outages. You will have a system that hallucinates at some point. So what is the technical debt that you should be tying there as well as for the incidents that are going to happen? And so what I'm trying to do is help companies with this calculation get all of the metrics on the table so they can be honest about what their potential ROI really is.

41:26Because most of this is all already there up front. They make poor judgments on what they actually think adoption will be, or they don't really take into account the full cost of what they're going to do.

41:39Matt Paige: Yeah, it makes so much sense. I think we just overlook that every time when we're planning and thinking about the return on an initiative or a use case that we're doing. And in the book, you go into 90 days to getting the number. And I'm going to save that for folks to go and check out the books. Cause there's actually an approach and a process that steps you through going, going through this process, getting to that ROI number and ideally getting to a positive ROI or you don't do it. But where I want to end it is there's so much up in the air with AI. What's your view on the future of where AI takes us being so close to it?

42:17Matt Paige: We're in it every day. Big loaded question to end it. Yeah, I wish I could project all that. I mean, look, in general, I think the easy things that you see already, right? AI agents are going to be a permanent part of the workforce, right? And small teams, which we've talked a lot about in this call, they're also going to be a big part of the workforce, right? That's going to be the way that business is done. I think we're going to see more and more hundred million, a billion dollar businesses with five people, 10 people. I don't think that's going to change. That is the trajectory that we're going on.

42:57I think that organizational structures, that takes, it took 2000 years. We've been building them this way for 2000 years. So that's why I keep challenging companies because that's their, that's going to be their biggest unlock. I think I saw some data last week from McKinsey that only 6 % of enterprises have actually made organizational changes. Okay. So they're like, it's impossible to fully unlock the value of AI without that. And that when you make work changes, that's going to change your job responsibilities. Right. And that's where people get scared because they're like, I don't, I've done this job for a long time.

43:34And those job responsibilities are going to change just like we've seen in software engineers. They can now wear multiple hats. They can now do three different jobs because AI agents are doing a lot of the work and that's going to happen in, in, in really every single space for what we're doing. I believe that the human is really important. I'm, I consider myself an AI humanist. I think that there's a lot of things that humans do really that are really important, like selling things now. It's so hard to know what to believe anymore. The one-to-one relationship of a human selling to another human is really important.

44:07I still think in customer service and some of those things, people, there's things you just want an answer for. And there's other things that are like super critical and you want to talk to a human. And so I believe that the human responsibility and oversight is going to be really important. And I think the biggest prediction that I continue to kind of, I think is going to become big. I think open source is going to be one of the biggest winners out of this entire thing. and that more and more companies are not necessarily going to see a good economics from these closed source pay token type of models.

44:44So yeah, I'm going against what the Anthropics and what the OpenAIs are doing at scale because I'm seeing a lot of companies where, look, let's just go and take an open source model, train it on my data, install it in, if I'm in Azure, AWS, or even in my own on-prem, I think that is going to be the future as well as you might use closed source for certain things. We just see a lot of tokens being wasted, especially on agents, where that's now 80 % of all token use is agents. And those agents don't need the most complex frontier model to do most of the work that they're doing. And so I'm pretty bullish on the fact that we'll see more open source.

45:23We'll see a lot more of multi-model where, I think of it as I come from mobile, least cost routing type of work where certain things go certain ways and certain go other ways. And I think we're just scratching the surface of that. And so moving away of, oh, let's roll out claw to everyone in our business and that's all we do. I just, I don't think that's the future of where things are going.

45:45Matt Paige: Yeah, that's a great spot to wrap it up. And by the way, you talk about open source and fine tuning models. Hatchworks can help with that too to get the last minute sell in there. So Brandon, where can people go find the AI ROI engine? I believe it's available on Amazon and we have the website as well. where do we want to direct people? Yeah. So definitely go to Amazon, just search on the AI ROI engine. You can download the paperback or the Kindle version. We've also built Madness team, built a great website for the book where you can get all the artifacts, like including like some of the calculations that are hard to understand because I was verbally talking through them, but go to www.theairoiengin.com and you'll have lots of great content there that you can download.

46:26that's all from the book, the different artifacts and stuff like that. And if you want to reach out to me directly, I'm on Instagram at Brandon C. Powell AI, as well as on LinkedIn at Brandon Christopher Powell. I would love to connect, would love to get your feedback on the book. I hope that this is just the first version. And as we continue to learn more and more from our amazing clients, what works, what doesn't work, I hope to be able to include more in that into the book and into the ecosystem so that we can continue to help customers get the best ROI they can on all this AI investment they're making.

46:58Matt Paige: Good stuff. Brandon, thanks for jumping on and talking some AI. Thanks for having me back after three years. I know three. We'll have to do it sooner than three years for the third round. All right. I appreciate it. Thanks for listening to the Talking AI Podcast. If you enjoyed the show, give us a follow or subscribe on your favorite podcast podcast. And don't forget to leave us a review. We love this. For more info on Talking AI, visit TalkingAIPodcast.com. Quick break in the pod. I keep hearing the same pattern with companies I talk to. Quad's helping employees move faster, but in many companies, the business itself hasn't changed.

47:34Matt Paige: The value's still trapped in isolated chats and experiments. And that execution gap is why forward-deployed engineers have become one of AI's most talked about deployment models. They embed with your team instead of advising from the outside. It's also why the FDE model is now central to every client engagement we lead at Hatchworks AI. As an official Anthropic partner, we embed Anthropic-certified FDEs to identify high-value business problems, build and deploy the solution, and put governance and security around it, then transfer the capability back to your team. If your cloud rollout is still mostly individual usage, check out how Hatchworks AI FDEs work at hatchworks.com slash cloud dash FDE.

48:07Matt Paige: You can also find it in the show notes. Now back to the show.

From the publisher

In January, Brandon Powell sat down with five of HatchWorks AI’s best engineers and learned they had stopped writing code. Agents did the work. So what did the engineers do while the agents ran? After an awkward silence, one of them told him: “We play video games.”

In this episode of Talking AI, Matt Paige sits down with Brandon Powell, CEO of HatchWorks AI and author of the new book The AI ROI Engine. Brandon was the first guest on the Built Right podcast back in 2023. He returns to share what HatchWorks AI learned pivoting the whole company to AI, and what hundreds of client implementations reveal about why most AI spend never turns into a return.

His core argument: better tools don’t create ROI on their own. If you don’t change the structure of your teams and your organization, the time AI saves simply disappears.

In this episode, you’ll hear about:

  • Why HatchWorks AI went all in on AI, and why the first year was so hard
  • The five-by-one meeting that changed how Brandon thinks about team structure
  • Why engineering is the canary in the coal mine for every other function
  • The five places companies spend on AI, and the one that matters most
  • Why AI is an enabler of strategy, not a strategy
  • How GenROI brings input from the whole organization into AI prioritization
  • The six failure modes that sink AI initiatives, from the wrong use case to pilot purgatory
  • What makes a company AI native, and why those companies never hire the middle
  • Small pods, forward deployed engineers, and one-person teams
  • The three levers of AI ROI: profit, risk reduction, and decision velocity
  • The AI ROI formula: potential value times adoption reality, minus true cost
  • Brandon’s predictions on agents, org design, and open source models

Key moments:

  • 0:00 Intro: what is AI actually returning to the business?
  • 0:57 Why HatchWorks AI went all in on AI
  • 4:30 “We don’t really write code anymore”
  • 8:13 Engineering is the canary in the coal mine
  • 10:13 The five places companies spend on AI
  • 14:38 AI is not a strategy
  • 18:45 GenROI and AI-enabled discovery
  • 21:58 The six failure modes of AI initiatives
  • 28:29 Cloning Brandon’s voice for the audiobook
  • 29:16 What makes a company AI native
  • 32:09 Smaller teams and forward deployed engineering
  • 35:19 The three levers of AI ROI
  • 38:50 The AI ROI formula
  • 42:28 The future: agents, org design, and open source
  • 46:52 Where to find The AI ROI Engine

Get the book: theairoiengine.com

Mentioned in this episode:

GenROI by HatchWorks AI

Most companies don't have an AI problem. They have a prioritization problem. There are hundreds of places you could use AI, and the hard part is knowing which ones are worth the investment, which ones you're ready for, and what should come first. GenROI uses AI-guided discovery to gather insights from across your whole organization, not just the few people who make it into a workshop. Our strategists then work with your leadership team to turn those insights into a prioritized AI roadmap weighted by business value, feasibility, and readiness. Learn more at https://hatchworks.com/genroi

More from Talking AI

All episodes