Fireside chat · 1:35 to 2:00 PM · Thursday, May 14, 2026
All right, just very quickly, this afternoon we have planned for you some very interesting interactions, a couple of firesides, as well as a keynote from Cathy Hackl. So I'd encourage you all to stay in the room. For the first fireside, we're going to have Phil Fersht, our CEO, as well as Mohamad Ali, who heads IBM consulting at IBM, join us here. And later this afternoon, I will be talking to Julie Durham, who is the CIO of Optum Insight and Applied AI at UnitedHealth. So with that said, Phil, Mohamad, please. Thank you.
Terrific. So you guys absorbing enough information for one day? Yeah, well there's more to come here. So we've been hearing a lot in the last 36 hours about how hard this is, the change we're having to go through, how we have to really think through why are we doing this and then build back in terms of a strategy with an outcome. So it's great to have someone here who's going to talk a bit about what they're doing with clients who are aligning revenue with AI investments. Mohamad Ali, who leads essentially the services arm of IBM, but some things you may not know about Mohamad is he was the CEO of one of my competitors, IDC. Ah yes, we don't talk about that anymore. When I first reached out to you, I think, yes, you work at IDC. Who's this guy? And IDG Communications and Carbonite. Carbonite, yeah. So, and personally, it's great to see someone who's, you come from a strong software technology background, actually leading a services organization, and then bringing in some talent like Neil Dhar and some other people who are like...
Yes, who actually know how it works.
Yeah. So it would be good to hear a bit from you, Mohamad, on, maybe we can start a bit with this aligning AI with revenue and growth, because I think that's the conversation we need to turn to in the last afternoon of this summit.
Yeah, no, absolutely. So I have a question for you guys before we get started. How many of you are primarily using AI for productivity today? Okay. And how many of you feel like you are successful at that? Okay, that's great. And how many of you are sort of moving to the next step of using it to drive new business models and increase revenues? Okay, so that's actually a really good sampling right there. So as Phil said, you know, my 30 years has mostly been in the software industry. After grad school, I actually started a company called Neural Applications Corporation, which, big name for what it sounds like, which was a neural network company. We actually did neural networks, fuzzy logic, genetic algorithms, everything you could possibly want. And eventually we sold the company and then I joined IBM and I got a chance to work on the software business at IBM.
And so, as Phil said, you know, I didn't start out, like I'm not, my background is not consulting. And when I rejoined IBM about three years ago, the idea was, and this is like right after ChatGPT hit, right? Like right after November 2022. The idea was that a lot of these labor-based businesses are going to change to become human plus digital labor. And what's digital labor but a bunch of software? And so about three years ago we started this initiative to create a digital labor-based consulting business. And so you know, Phil's point here about productivity to revenues, we now have three years under our belt at this, and we've done a lot of work on the productivity side. We got a chance to do it with an AI-first company at scale, and I'll tell you about it in a sec, and a company that had been around for like 100 years, and that's IBM.
So the AI-first company was Riyadh Air. Anybody here ever heard of this airline called Riyadh Air? Okay. So they were founded a little over three years ago with the idea of competing on the premium level against like Emirates. So they immediately placed an order for 200 planes, making them basically the size of Emirates overnight, and they needed a software stack. And they actually wanted us to build an AI-first software stack. They did not want to go use traditional airline software. And initially we thought this thing is doomed for failure. But three years ago we started building what we call IBM Consulting Advantage, which is our AI platform to do services projects, and we did it. And three years later they're flying. You can get a commercial flight from Riyadh to London today. And it all runs on this AI-first stack. Their HR department is AI-first. Their finance department is AI-first. They actually have a way smaller organization than their competitors.
And then we also got the chance to re-engineer a 100-year-old company, IBM. And what we did was we decomposed into 490 workflows, we picked the top 70, and we went after a $25 billion spend. So this is all about productivity, productivity, productivity. And so with all that background, one of the things that we quickly realized, like halfway, two-thirds into this, is that there are actually tremendous opportunities to turn this into revenue generation. So a lot of the work that we're doing now is helping clients create new business models that didn't exist before, that they couldn't do without AI. But now with AI, they can create a business model that drives revenue, or take their existing business model and reconfigure it to drive revenue. So we could talk, Phil, much more about the revenue side, but I kind of wanted to lead up to how we got here.
Yeah. Well, I mean, Riyadh Air sounds very interesting because they didn't have any legacy to start with. Right. It's built from ground up, which might be the story of the world's economy as we look forward as we're gonna get a lot of these small to medium sized businesses springing up because they can just be made more competitive.
Yeah, I mean it's actually amazing, 'cause Tony Douglas, who's the CEO, and he is a very, anybody ever hear Tony Douglas speak? He's this like tall British well-spoken James Bond kind of guy, like super impressive, right? And so one day he says to me that, listen, what I actually want to build here is an e-commerce company that happens to have 200 aircrafts, right? And that's how he thought about it. And I actually, and if you think about it, here's a major company that was created from scratch, you know, in 36 months.
And so think about that disruption to almost every industry that we've got. Yeah. Yeah. I think there was a company called Lemonade on the insurance side. It's been around for about a decade now, but they basically started off with a minimal number of employees. And I think they're very competitive now in the insurance industry for P&C. And you never talk to anyone. Right. Because their cost structure is so low, right? They can afford to provide incredible service. That's right. That's right. So we talk about Services-as-Software all the time.
Where did you hear that term?
It's two years old now, actually. It's going on. We're now having to come up with something new. So what have you actually industrialized at IBM that proves delivery is becoming software-led and repeatable?
Yeah. So, you know, when we built ICA 1.0, we just actually launched 2.0. It was an environment where we can industrialize digital workers. And so the way we did it was we built sort of a middle layer, put all the AI stacks underneath, any LLM you want, any builder you want, whatever tools you want underneath, and we created these frameworks for building digital workers on top. And what a digital worker is, is any bit of software that calls an LLM. As long as it's non-deterministic, it's a digital worker. So we did that, and we ended up with like 30,000 of these things, and then we had to figure out a way to curate them down to ones that were really valuable. And we call this the hire-to-retire for the digital workers. So people are always building them, they're hiring them, we have to figure out how to retire them, we starve them of tokens if they're not useful. And so we ended up with about 4,000 of these things.
And now we have deployed these in our factory, right? So you can think about our factory as Bangalore, Romania, Louisiana, like all these big locations. And so if you are like building a Ford truck, you know, you could look down in the plant and you could see these robots and these human beings building the F-150, right? So I've got a dashboard just like that now. I could look down into the factory and see that there are 800 human beings and 300 digital workers, building these projects. However, these digital workers are human built and they're human in the loop. We don't like let them do stuff without human in the loop. So for, so we got these 4,000 things fully industrialized and when I say industrialized I mean think about it as manufacturing, right? And that's our business today. And if you look from 2024 to 2025, our, you know, our margins expanded 20%, 20%, right? And that's because of how we're delivering this stuff.
What's 2.0? So 2.0, we are going to have this machine build these digital workers and we're going to have these digital workers becoming more autonomous. Now why didn't we do this in 1.0? Because we just didn't know what they were going to do. Now we have a lot of experience with it. So just an example, we have these 28 digital workers that do threat investigations and in March they did 70,000 threat investigations for 112 clients using 9 billion tokens. And every one of them, we had human-in-the-loop verify. And now we're realizing we don't need the human-in-the-loop to verify, they're actually that good now, right? So now that we have a couple years of data of running this thing, we're getting comfortable with setting the guardrails so we can automate this. And that's gonna take the consulting business from what it is today into a whole new era. And ICA 2.0, we just launched this week.
So we've spoken an awful lot about getting to that point where you feel comfortable moving past the proof of concept. What works and what isn't working with the clients that you've been engaging with?
So I mean, I'll tell you what doesn't work. What doesn't work is buying a whole bunch of Gen AI licenses and giving it to 50,000 people and say, go be productive. I mean, you guys know that, right? So I think you're all past that, right? I mean, like many of you in this room are representative of the clients that we serve, and many of you are way more sophisticated than that today. You've realized that, you know, the hard part of this is re-engineering your processes to be human plus digital. And in some ways, that's what we did at IBM three years ago when we took the whole company and we decomposed into 490 workflows and people thought we were crazy because there was no AI when we were doing that. You know, we didn't use AI to do that. We did it manually. And then we took these 70 workflows and we re-engineered it.
So, Phil, to your point, what I'm discovering is the hardest part of this is not what LLM you pick. And people are so, like, they spend so much time trying to figure out which LLM to pick. Like, why? Right? I mean, they're all gonna be commoditized anyway. I mean, who's gonna pay $25 for a million tokens? All that's gonna be a commodity. And then the builders, everybody's like, well, do I use Claude Code or a Codex or OpenAI Codex or whatever, right? I mean, those things are gonna get commoditized. And the hard, hard part is the process re-engineering. And so we've been focusing on building tools for this process of re-engineering. And then the super hard part is then convincing management, right, which is what Arvind's dealing with one of his clients. He built the thing, it works, and now they don't actually want to change how they run their process, right, because they're afraid their people get upset. So I would say the process re-engineering and then the change management are the hard parts.
Right, right, okay. And where you're achieving real revenue growth and real creativity with clients, I know you've been sharing a bunch of stories. I don't know which ones I can talk about legally or not, but maybe you could share one.
So I've got two. So we have this one client, and they're a components manufacturer. So if you're going to build a data center, you would go buy, like, you know, several million dollars worth of these $2 parts from them, right? And so they get these long lists of components that are required by the contractor for this data center, and somebody has to sit there and match it up to their catalog, right? And sometimes it could take days to do this matching. So they hired us and they said, hey, can you build some agents that do this matching? And that's an easy job for agents, right? So we did it and then they demoed it to the CEO. And so the CEO says, hey, wait a minute. If you can mine my catalog and do this, can you mine my competitor's catalog? Because they're all public. They're all online. And if you find a part where there's scarcity and I'm the only one with the part, then I want to charge a premium. So now they've implemented this and they have revenue growth out of it. So that's one example, right?
Another example is building an entire new business that just didn't, like, that didn't make any sense before Gen AI. And now you can do that, right? So if, is anybody here at IBM Think, at IBM Think? Okay, all right, great. We got two people. So, you know, one of the things that on my keynote, you know, I started out with Providence Health. There are 120,000 people, 51 hospitals in the Northeast. And two years ago, this woman, Carol, came to Think and said, hey, I want to use AI to hire nurses faster. I want to use AI to hire people faster. I want to run 20,000 people. I'm always hiring tens of thousands. But nurses are a huge problem. And so we said, great. What platform do you run on? She says, Oracle for talent acquisition. We said, okay, we're going to build a bunch of agents, and we're going to stick them on top of Oracle and see what happens. And so today, this thing is running at scale. She's hiring, on average, 12 days faster, right? Think about, like, a nurse 12 days faster.
So then the second thing that we showed was how all these 4,000 digital workers run on AWS, including secure facilities like GovCloud. And then the last thing we showed, and this is to your point here about new business models, right? So Pearson, has anybody known Pearson, the learning company textbooks? Okay, so they have a product called Credly, and it provides like skills badges to people. Anybody know about Credly? Okay, so we use Credly at IBM and I have to take like, you know, like business kind of guidelines training like we all do and you get a badge. You know, I have a security background, so for fun I go and I take the security test and I get a badge, right? And so we all get these badges.
And so one day I was with Dave Treat, who's a CTO, and I said, hey, you know, have you guys been thinking about this? He goes, oh, yeah, we've been thinking about this because we're going to have to credentialize digital workers as well. Now, remember, this is a market that didn't exist three years ago. And he's like, how are you going to do it, Dave? And he had this whole theory on how, well, you can't just, you know, give a digital worker a test because they can memorize everything. And it doesn't work. So what you need to do is this whole five-step framework and you actually have to create workflow problems that they've never seen and you have to test them on the workflow. And with a human being, this could take forever, right? So you go to multiple choice. But with a digital worker, if they built digital workers that are testing other digital workers, they could do it really fast. So you could provide these complex workflows that the other digital worker hasn't seen before and it has to pass the test.
So we said, great, let's go build that. So we took, our thing is called IBM Consulting Advantage, this AI platform. So we built something for him called Pearson Advantage. And on Pearson Advantage, we help him build some digital workers that actually test our digital workers to credentialize it, right? So I have a digital worker that's a cloud architect. So now that digital worker has to talk to his digital worker to be tested and then get a cloud essentials badge. And then guess what? It shows up in our IBM HR, Credly HR system, right? That is a business that did not exist before. That is an incremental business to their existing business. That's revenue growth, right?
So maybe everyone's been hearing about client zero all the time. But maybe you could just explain it in a way that everyone can really understand and how this is credible to clients.
Yeah. You know, I would say client zero is sort of what made this real for all of us. Have you guys heard the IBM client zero story? Nobody? Not even one? Oh, thank God. He's one of ours. That's right. You and IBM are, of course you do. So, you know, when I came back almost three years ago, Arvind was already on this thing. Because November 2022, OpenAI announced ChatGPT, and it just takes off. And IBM's been in the AI business for a while. In 1957, it was IBM, AT&T, somebody from Dartmouth and MIT that had that famous conference where modern AI was born. And, you know, IBM did, like, the Watson and the Jeopardy and the this and the that. And this startup comes out of nowhere and brings this LLM in a consumable form to the masses, and it just takes off. And, like, where are we?
And so Arvind's thought here was that if we didn't, we're not going to be the ones to bring this technology to the market, even though we've been working on LLMs for a long time, we need to be the best users of it. So that was the beginning of client zero. And he said, we are going to be the first client for this stuff. And that's when, and this is actually before I joined, he started, he said, we're going to decompose the entire company into workflows, 490 workflows. And then we're going to pick 70 of these workflows that represent this $25 billion spend. And we're going to build digital workers for them. And so initially we built 220 digital workers at the time on Watson X because that's what we had. And so on average, you know, each workflow had four digital workers and we re-engineered it.
And over a three-year period, we've made a lot of progress. The $25 billion like-for-like is now about $20 billion, right? And so probably the most significant of this is our HR budget in 2022 went from whatever it was to 40% lower in 2025. You can even see it in our cash flows, right? So that $5 billion that I mentioned of savings, the cash flow went from $9 billion to $13 billion. We've reinvested a bunch of it. And so that's the client zero story at a corporate level. So what does that mean to us in consulting? So in consulting, you know, we did a bunch of that work. And there's all these assets that were created, all these digital workers that were created. So when we built Consulting Advantage as the platform, one of the first things that we did was we went and we harvested all these digital assets. So now when we go to a client and the client says, hey, I need you to change my HR process and make it agentic, I don't have to go build it from scratch. I take those assets. They live in Consulting Advantage. I deploy it onto whatever AI platform they've got, and then we can go fast. And this is why that part of our business is growing so fast, our consulting business.
Yeah, I mean, taking a heritage company, public company, there's no hiding here. You had to do this in plain sight of Wall Street to see how you did it. Yeah, that's right.
Yeah, I mean, Phil, you're right. There was no hiding because when Arvind announced this, if you guys track the IBM stock, has been flat for 10 years, right? And so the analyst looked at this and he go, well, we don't believe it. You have to show us every quarter. So we went and we got this guy named Nick Fehring, who's our controller, and we said, Nick, you actually have to report this in the 10-K and 10-Q each every quarter. And many of you have probably, many of you are companies, you know this, right? They'll announce a big cost reduction and it's kind of smoke and mirrors, right? So like you take the costs out there and you put it back in here and you don't actually get it to the bottom line. So Nick's job was to make sure we actually got it to the bottom line because he now had to report it every quarter. And if you look at our quarterly results, right, you could actually add it up. So the EBITDA expanded, so that's one piece of it. The R&D expanded from 9% to 12%, and the SG&A changed as well. You add those three numbers up, and you get the $4.5 billion of savings, right? And so it had to be SEC reportable.
Right. Yeah. So one more question, and then I'm going to go out to the audience. We've got Anthropic and OpenAI, and I think Gemini about to announce as well, making big noises that they're getting into execution. Anthropic said they're going to go after the mid-market. OpenAI, just basically they're going to go after everybody. Is this going to change how you show up?
I mean, that is not the thing that's going to change how we show up. We're showing up differently because we have to, right? So like, you know, I told you the three clients I visited, they're all giant companies. And when I go see them, and you know, these are one below the CEOs and so forth, I no longer take a PowerPoint. I go on my laptop, I turn it on, and I say, hey, this is how I've deployed 4,000 digital workers. Look at them in my factory. This is how I run my business. And almost immediately they go, oh my God, we have to have that, right? It's real. And so that's how we're showing up now. And just today, we announced these Forward Deployed Units, not Forward Deployed Engineers, right?
If you think about it, like one of the clients, and I bet you the financial services folks in here have these. You have like five 25-year-olds from Anthropic as FDEs in your shop helping you build stuff, right? And yeah, you could build stuff, but are you really going to be able to deploy that in your processes? So our FDUs (Forward Deployed Units) are six people instead of 30, and they can re-engineer your processes, build the digital workers, deploy them in 30 days. That's an FDU, right? So we're showing up this way just because we have to show up this way. But I think the Anthropic and the OpenAI announcement that they're creating services business is a validation that actually the hard part of this is re-engineering the processes, right? And so, you know, maybe accidentally we went there immediately three years ago, but that's turning out to be, I think, our strength.
I think our traditional competitors are kind of in trouble, right, because we have this model, the Anthropic and OpenAI services organizations are going to have this model, and those that don't have the toolings and the capabilities to actually re-engineer the process with AI and deploy it are, you know, are going to hit a wall. I mean, our competitors who are doing deals where they're taking over 7,000 people, you probably know who that is, right? At a time, like, what are they going to do with all these people? Like, they're going to run into a wall. And so I don't know which one of our traditional SI competitors are going to hit that wall first, but somebody will. And then these new types of businesses, whether it's the two you mentioned, ours, you know, some of our competitors are doing a reasonable job as well. Those are going to be ones that, you know, that excel here.
Terrific. Thank you. Another question over here? How are you going about with IP?
I mean, I go back to the other example. Because you have these digital agents, bots, whatever you want to call them. Do you use your own LLM? Yeah. No, we don't. We use anybody's LLM. Sure.
Yeah.
Regardless, the point is that in the future you can essentially say, you know what, I've packaged this. Yeah.
All risk, compliance, da, da, da, da.
You create an IP marketplace. Yeah. And then all the companies said, hey, we thought this was exclusive. This was our design, our risk controls, everything. And now you put that in the marketplace. Yeah. So how do you…
Yeah, so we have an answer. Let me tell you what it is. So there are three layers to this thing, right? So you can think about it as the core platform, the IBM Consulting Advantage core, right? This is the thing that everything runs through. It does the observability, I could see my factory, I can, you know, I use something called Arize Phoenix to track every single digital worker and what they say to another digital worker or human being. All that's core, we built that, right? I spent hundreds of millions of dollars building that. I hired an engineering team because that's like what I know how to do, and that engineering team has built that, right? I have 600 engineers that this is all they do, it's like a software business. And so if a client wants that, that's a license fee, right? But many of them don't want that. They've built their own, right? Many of you here have built your own. Like back in New York, you have Eliza, right? So you've got your platform. So I don't need to deploy my platform. But if somebody needs it, we can do it. Like three of our clients that I mentioned here already didn't want to go build their own. So that's that piece.
The second piece is what you call the agent marketplace. So I've pre-built a bunch of agents without any client IP, right? I did it myself. And so if you want those, like the Ask HR agents, there are probably 200 of those agents that do different things. Like there's one agent that'll create an employment letter that says I'm employed here. Like that agent, I built it, right? If you want to license it, fine, but that's really not my core business. And then the third is services, traditional services, but high value services, like these FDU services where you've got, let's say it's Eliza, you've got Eliza and then on top of Eliza you want to re-engineer your workflow. So we've got some tools to do it. We will use our own tools to re-engineer the workflow into Human Plus Digital. We will build the digital workers and we will show you like on the screen. Let's say, do you want to deploy it on, you know, OpenAI, you want to deploy it in Anthropic, CrewAI, a LangFlow, whatever you want to deploy it on, and we will deploy it, right? And so when we do that services work, that IP is your IP. Does that make sense? Yeah. Okay. Sure.
Gentleman over there. Thank you so much for your time.
Sure. Quick question. Thank you. Thank you. So from your vantage point, one quick question, because you're dealing with a lot of organizations, those who are in various stages of their journey in AI adoption, right? So some are probably stuck at AI strategy right now. They don't even know what they are doing. Some probably at governance and operation stage as well. So how do you really manage that? And then what's really the key messages for that from your perspective?
And this is the different stages, right? Different stages. And what company are you with? I'm with Cognizant. Okay. Okay, so like Cognizant, the company that does SAP implementations?
Yes, we actually are an AI and automation platform, so we work with a lot of organizations as well. I'm chief AI officer, so in that way, my role actually deal with a lot of organizations there who are really into this journey as well, so we are seeing that problem firsthand.
Yeah, I mean you're absolutely right. Organizations are at different phases, right? I mean, there are some organizations that really want to do something, and they have the sponsorship of the CEO, but they don't have, you know, they haven't pre-built their own platforms and all that. And so, you know, those are probably the easiest to work with, because for us, we just come in, you have the sponsorship of the CEO, so you could do the change management. There isn't a whole bunch of stuff that you have to pre-use, and we just come in with all our tools, put it in there, and we can get stuff running in weeks, right?
At the far other end is a company that's pre-built this whole AI platform, have got this religion that I only wanna use this AI, or it's like this whole religious battle on the technology stack. Then you're not dealing with the CEO, you're dealing with a director or VP level, and you have to re-engineer this entire process. Those are the hardest projects. When those comes, I almost tell the team, don't bother taking them, right? I think to accomplish what we wanna do, like IT services companies used to operate in the CIO land and now all my conversations are with CEO or one below, like CFOs and so forth or heads of HR 'cause we're re-engineering this entire process. That's not what IBM Consulting used to be, right? And so it is, you're right, I mean, there's this whole flow and you have to apply different techniques to different stages. And there are probably some projects we shouldn't do.
Yeah. Thank you. Terrific. I think we have to sum up that we're out well over time.
If you don't have an account, Register here |
With the exception of our Horizons reports, most of our research is available for free on our website. Sign up for a free account and start realizing the power of insights now.
Our premium subscription gives enterprise clients access to our complete library of proprietary research, direct access to our industry analysts, and other benefits.
Contact us at [email protected] for more information on premium access.
If you are looking for help getting in touch with someone from HFS, please click the chat button to the bottom right of your screen to start a conversation with a member of our team.
