Transcription
I'm Tim Huang, and welcome to Mixture of Experts, a recording live from IBM Think 2026 in Boston. Each week, Moe brings together a group of brilliant minds working at the cutting edge of technology to discuss, debate, and guide you through the week's news in artificial intelligence.
On this week's episode, we've got Ambi Ganison, AI transformation leader at IBM, Hillary Hunter, GM of IBM Power and CTO of IBM infrastructure. And joining us for the very first time is Tim Crawford, founder and CIO strategic advisor for AVOA.
So, three big stories today. We're going to talk about the recently released IBV CEO study. We're going to talk a little bit about this really interesting data point about AI dominating the VC landscape. But because we are at Think 2026, I figured we should just take a moment to talk a little bit about everything that's going on, the trends that we're seeing. Hillary, maybe I'll start with you. Like coming out of this year, you know, I was coming down the elevator and I was like, there's just so many things to see. Uh, what stands out to you as the most important from this year?
Responding a little bit to some of the feedback clients have been giving me from their reactions to things. I think there's a lot of excitement about seeing IBM integrate AI in a way that really drives productivity end to end. If you go back even a couple months ago, we were talking about AI and the software developer life cycle and AI in IT operations and we were kind of talking about it on a domain-specific way. Um, many clients have been giving me feedback that the vision that they're seeing here is the pieces coming together that we're talking about complete life cycles, completeness of outcomes, not individual siloed applications of AI. And I think I agree with them that you're seeing that on the show floor, you're seeing that in the keynotes and in the announcements that were made.
Yeah, I think that maturity is so interesting. I guess I mean maybe I'll turn it over to you. I mean, uh, Bob is everywhere. I see like every, you know, the character is everywhere. The placards are on every single countertop. The character is, you know, ubiquitous. Um, it does feel like, you know, maybe what's happening with Bob is is part of this maturity, part of this idea that you're going to kind of see these agents become much more of a full end-to-end kind of life cycle. Do you want to talk a little bit about that?
Yeah. No. Um, we had a phenomenal announcement with Bob earlier this week, right? Um, we are seeing some landmark improvements in terms of how those coding agents are changing over time. You and I have chatted about some of these in the past episodes, right? Uh, we've seen an overall step-function gain in terms of how these coding agents are maturing. But the critical piece is, you know, we're not treating and we're not thinking of Bob as just a coding agent, right? So yes, we're leveraging Bob for, um, code generation, for developers to go build applications, for modernization, right? Yes to all of those, but it has become such a powerful instrument internally, right? So for my consultants, for my teams, we are leveraging that heavily for us to do it across the stack, right? So all the way from going and defining, um, you know, even building PowerPoint decks to going and, you know, manipulating Excel files, like all the day-to-day work that we do, we're infusing those into the mix there, right? So it's, um, it's a fantastic capability. It's sort of a super tool that sort of unlocks.
Yeah, you're like literally touching like for everything. It seems like.
Like I can't stop playing with it.
Yeah, that's great. Um, Tim, you know, before we started recording, you were talking a little bit about how you feel like almost like we're now moving into a phase of kind of lessons learned, right? We're like a little bit maybe exiting the hype phase and moving into like, okay, what's going to be the long-term vision as technology? Uh, do you want to talk a little bit about kind of like the lessons that you see kind of emerging again, as someone who's been in the space for a really long time?
Sure. Thanks, Tim. And thanks for having me here. I think one of the things that that we have to be realistic about is AI is not new. We've had a chance to start to work with the technology and now what we're starting to do is starting to figure out where can we apply it to have the best return for our organization. But at the same time, as part of that maturity, we're also starting to think about the business impact of it, which means we have to think more cohesively across the organization. So no longer is it just simply we're looking at one department or another department or one function or another function. We actually are starting to think about this cohesive process and how it impacts how we change how our business operates. And so those are some of the things that I'm starting to see finally happen as part of that maturity. It's taken a while to get here.
A while being a couple of years.
I mean, that's like forever in AI time, right?
It's forever in AI time. But if you look at the last couple of iterations of innovation, I mean, we are moving at lightning speed.
Yeah.
And that's going to keep going.
That's right. And Hillary, do you want to build on that a little bit? Cuz I think like, you know, in some ways it's easy to kind of talk about process and like how these tools will shape process. You know, one of the big announcements it sounds like looking at some of the headlines was IBM Concert and thinking a little bit about how kind of automated infrastructure looks, but that's like a tool, but it's also kind of like a different way of thinking about the business process. Do you want to talk about that relationship? Because I think it's one of the most interesting things that's happened.
Yeah, it it absolutely is. And it's, it's interesting because even again, months ago, we would have used the term AI ops and just using that phrase now puts you in a particular box that's no longer appropriate because we're having much, much more expansive conversations, right? Understanding of a landscape and understanding how well-maintained it is from a security and vulnerability perspective. Uh, looking at your assets from the perspective of being able to implement more complex solutions. We have clients that honestly, because of the ability to create automation and scripts, are able to manage more complex environments, which includes also doing things like high availability techniques, DR flipping over to the cloud, coming back on premises. Historically, that would have taken a lot of specialized skill with the combination, I would say, of infrastructure as code, the ability to generate infrastructure as code using AI, and the ability to use AI then to monitor those environments and understand that everything is working as expected. There's a completely new skill set that can be put in anyone's hands, and that becomes transformative to have visibility from a Concert perspective of what's going on on your estate, to have the ability to manage an even higher quality estate because AI can help you get there, um, and then to be able to optimize it on an ongoing basis using IBM automation products, for example. It's a tremendous amount of new capability that is suddenly in everyone's hands.
Yeah. Yeah. And I feel like this is like, it's happening almost at every layer from like the user layer to the org layer to like the whole business. You know, you find like using some of these AI assistants, there's like almost too much to keep up with. And so you need a whole another set of tools like just to monitor. It almost seems like.
Well, if I can share a phrase, I heard this earlier from someone who was attending the conference. They said, "The extent of executive AI literacy and personal use will drive that organization's AI speed." And I love that because it really is a very personal thing for leaders to be up to speed on everything that's possible coming along in that journey. It's not just developers writing new application code. This is a pervasive and broad conversation across the full organization.
Yeah, definitely. Yeah. If you want to jump in.
Yeah. And that's actually something that that I'm seeing as well in my research is that when you have that executive buy-in, it changes the dynamic for the AI project or the impact that AI is having with that particular business process demonstrably. And so the literacy plays a big role in that, but also the conversation that we're having about AI, because now it's more of a balanced view of, okay, I need to have the trust and understanding what it's doing and how it's doing it and ensure that I'm embracing the opportunity, but at the same time, I understand the risks associated too. Right? That's what builds trust over time. And so it's a combination of who you're partnering with, the technology you're using, but then also making sure that your team top to bottom, as you said, Tim, understands how this is playing within their company and within whatever it is that might be. It might be customer engagement, it might be sales automation, it might be manufacturing and supply chain. There are a number of different examples of how this is actually playing out for companies.
Yeah, absolutely. And I do want to talk a little bit about risk before we move from this segment because it does feel like the other big theme of this you're just walking around is the security question around all this. Uh, Ami, I see you nodding, so maybe I'll just bring you into the conversation because you know way more about it than I do, but it does feel like, you know, the surface of security is also changing, like how do you monitor, how do you pay attention, what do you even need to pay attention to, feels like that is evolving. And I don't know, do you want to talk a little bit about that? While we've been talking about kind of all the good, I think there's also the risk, right?
No, you, you, you constantly need to balance both of those aspects, right? So yes, and then I totally agree with, uh, what Tim was mentioning. So as executives are getting educated on it, right? So they're getting educated not just on the superpowers of AI, but it's equally important for them to get educated on what are the ways to go and apply the right types of guardrails. So you're deploying the AI responsibly, right? So it's a security element, it's a compliance element, right? So both of those come into picture. So how do you mitigate those risks? Um, if you're, and we've chatted about some of this previously as well, right? You go and put some sort of a chatbot, and we've seen these happen in the past. Um, if you go and put a public-facing chatbot, right, there should be some basic guardrails in place when it's out open to the public to make sure that that's not going and spewing something that's completely inappropriate, right? So there are things of those nature that, um, you know, there's a framework, there is a proven methodology to go institute some of those upfront. And, you know, like I always say, governance is, it should never be an afterthought, right? It's, um, it's very compelling to go run at 1,000 miles per hour, but that doesn't mean that when you run at 1,000 miles per hour, you forget the critical component of introducing the guardrails.
And right.
It doesn't take long to go and institute those.
Yeah. That's great.
Yeah. Do you want to jump in, double-click on that real quick? Because I think the governance piece is really important. I talked to a lot of different companies, both end-user organizations as well as vendors, and one of the things that is woefully lacking in the industry, but IBM talks about a lot, is governance. And I can't emphasize it enough. If you are only looking at the opportunity that AI brings without understanding the risks and the governance models that need to be put in place, you are asking for trouble because you have to understand how to create the appropriate balance within your organization. It's not always from nefarious purposes. It's, it might not be a bad actor that's going to infiltrate your organization or or leverage a rogue agent. It could be a rogue agent that had a very specific purpose.
Sure.
But it went sideways and you didn't realize it until after you've consumed a phenomenal amount of resources, tokens,
And you've done damage in terms of how the data has been played out.
Mhm.
So it's important to ensure that you have that balance in right from the get-go, as you said, of both opportunity but understanding the governance pieces that go with it to ensure that you're not in, you're not bringing in undue risk into the equation.
Absolutely. Hillary, you want to jump in?
Yeah, just to jump in and build on that. I think there's a lot to be learned personally from the cloud era. I remember when I was working full-time in cloud in like the 2018 to 2020 timeframe, we would do roundtables as CISOs, and and I always joked, it was my favorite party question to ask a CISO, is your cloud environment more or less secure than what you had on premises? And and back in those times,
80, 90% would say cloud is worse, right? Cloud is worse. And by the time we got to kind of 2021, 22, 23, and today, almost every CISO will say their cloud environment is more secure. Why? Because the full picture came together of infrastructure as code, automation of compliance, security boundaries, no humans touching a firewall, automated by code in the provisioning, things like that. And I think there's a tremendous amount to be learned here. Those that can get out ahead of AI and govern it with structure, govern it with technology, and use the available tools to solve those problems can move much more quickly and get to confidence that the AI is safe. Just like in the cloud era, we got to confidence that the cloud can also be a very safe place to do work.
That's right. But that's actually, well, actually go ahead.
But but it is more than just the tools itself because during that same period, we learned something about ourselves that we were making a lot of assumptions when we'd make that statement that our data center is more secure than the cloud. There were a number of assumptions that went into that state.
Absolutely.
And so by learning more about how we operate and in some cases having a heart-to-heart with ourselves and being honest, comparing and contrasting the alternatives, you start to realize that, okay, I need to look at this differently. And that's why I think that maturity is really coming to help us as we move in through this agentic era.
Yeah. It reminds me of a conversation of an organization that was very frustrated that their risk department wouldn't let them use a PowerPoint generation tool for fear of hallucination until they explained the rate of human errors in PowerPoints. Right?
So I think the next thing I want to talk about is, so the IBM Institute for Business Value CEO study is out for this year. So they basically poll 2,000 CEOs. Um, and I think one of the most eyepopping stats out of it is, let me just read it here, is that 64% of CEOs say they're comfortable making major strategic decisions based on AI-generated input. And I think what is so interesting is basically that like, I think not very long ago, we were like, don't touch this stuff. It's full of hallucinations. It's never going to be reliable. Like you should stay away, stay away, stay away. And I guess the question here is like, have we crossed this threshold now where, you know, reliability is actually now something that is widely shared? Like a majority of CEOs now say, yeah, I'm going to make a big decision, it's going to be informed by AI, that's that's okay. And if anything, the error rate might be higher with the humans I rely on. I guess, Ami, like do you want to talk a little bit about that? Like, is that an important threshold?
Yeah, so it's, I think there's there should be a nuance in this, right?
Sure.
Um, generated decisions and inputs for making decisions is not new. That has existed for ages, right? Like you, we've had traditional machine learning models for a while. Like even 10, 12 years back, we've had models for, you know, risk analytics, you've had models for, you know, inventory optimization, you know, which products are you going to go and shelf at any point in time. So those things have existed for a while, and some of those decisions were made with the help of AI.
Right.
The, the critical piece, and you know, like we've always maintained this as, um, as a perspective that, you know, you always, it's not just about generating these decisions or the recommendations for the decisions, but you always accompany it with, you know, the, the explanability and the, the traceability of those decisions. So I think we are seeing some of those manifest into the agentic AI realm as well. And so when CEOs are talking about trust in AI, right? Um, we are sort of repeating the same mantra for the agentic AI world, right?
Yeah.
Um, so, you know, bottom line, the, the notion itself is not new. Like the, the leveraging of AI to make decisions and, you know, using that to recommend decisions to be made has existed for a while. I think we're taking some of the approaches that we have done in the past, and then the CEOs and the, the senior leadership who are now looking at it, you know, more from the agentic AI realm are now saying, okay, we have, and we see the right approaches emerging in order for us to go make those decisions. Yes, they may have thought about hallucinations and some of the risk. Sure. But, you know, when you have agents in production, right? Like, as a rule now, we always make sure that, you know, when you have agents in production, you always show the traceability, right? Like, okay, this agent went in and then kicked off another agent, and that agent made these following tool calls. So you have that sort of visibility. When you have that sort of visibility, then it becomes easier to go and trust those decisions or make the appropriate judgment calls in terms of what you want to, um, sequence it on.
Yeah, that's right. I mean, so some of the listeners in the audience, I, you know, like I still hear that and I'm like, that's all good. I still have a little bit of fear. Um, cuz I guess Tim, to your point earlier about like hard conversations and if you don't get this right, you'll have an agent that consumes a bunch of resources. Like it feels like 64% is impressive, but would you say it's also like maybe fragile? Like it's possible for a few big incidents to really change this number. Like maybe we're in 2027 and the IBV survey is way lower because of something that happened. Do you buy that?
Well, so in certain communities, there's actually some conversation that the expectation is there will be a significant breach in 2026 that will reset the board for how we think about AI.
Um.
So there's some concern about that. I think that number to me feels high.
Because it will be positive until it's not.
Sure.
And what I mean by that is that that's based on implicit trust today.
Mhm.
If that trust gets violated, you're going to see that number plummet.
Yeah. And so this goes back to what you were saying where these systems, we need to understand how do we, how do we dig into what AI is presenting to us and ensure that the information we're getting is something that we truly can better business on because executives are out there making these decisions on strategic decisions around their company. You know, do I go right or left? Do I go into a new market? Do I go after new customers? Do I change the direction in how I engage with customers? These are tectonic shifts for companies that are both costly, hard to do, and incredibly risky. And if they make the wrong decision, it could send them into oblivion. And so there is a bit of trust that you're seeing in this number today. I think we will see something happen that will kind of reset the board a little bit, and then we'll get it right, and we'll start putting in some in some guardrails to ensure that the information we're getting, we truly can trust, and we can move on from that.
Yeah. Hillary, I'd love to get your view from your vantage point working with a lot of boards and executive teams like it kind of feels like, um, you know, there's a stat here in the survey that 76% of surveyed organizations have a CI, uh, AIO. Um, and I think a little bit about how like, you know, it's not like god-given that we have a thing called a CMO, nor is it god-given that someone has like a thing called a CIO. These are roles that were invented at a particular point in time that an organization said, we need to have this role. Yeah. And, you know, these roles, they come and go, right? Um, and so kind of question for you is, do you feel like this, this new role is something that's going to become like a persistent thing in corporate executive teams over time or where where are we in that?
Yeah, I mean, I see a couple of different forms of C AIOs. Um, I wish we could figure out how to pronounce that. Can, can we, can we define a way before we started, and I still got it wrong?
Yeah. No, no, it's, uh, it's, it's a thing. Um, but but I see a couple of different instantiations. Some are basically the chief AI evangelist, some are the leader of the team that has the greatest AI competency, etc. There's, there's a lot of different definitions. And I think in general, whether or not that role stays on a permanent basis depends on whether or not they are trying to help the organization through the hype cycle and the learning curve, or are they trying to consistently deliver technologies. But I would say that in the overall executive construct at the C-suite, the most effective model that I am seeing for organizations that are actually getting return on their AI projects is when that person functions as part of a team. Because again, going back to the cloud era, if you just had a chief cloud transformation officer, the VP of cloud, that person would get a no from the CISO, that person would get a no from the chief risk officer, that person would get a no from the application owners because of schedule. Exact same thing is is being replicated on AI, where those same personas bring their concerns to the table, and it then becomes a blocker and an extender of the amount of time. If they function as a team with joint responsibility for getting to a yes, getting to a successful AI implementation, then it comes to the table, security comes to the table, the AI expertise comes to the table, etc., and they are jointly commissioned for success. And I see several organizations doing that, and those organizations are moving forward with guardrails to your point, because it's all been co-designed, but they are tending to move faster than those that have a CIO that then has to go and shop and ask permission of many other parties in the organization.
Right. And if you want to give us maybe a sketch of like, so how does this role even come to be? Right? Is like the CEO says, like, we have to do this because obviously, you know, having worked with some CISOs in my time, they're they're kind of jealous of their authority. Um, how is this role being carved out of, you know, these sometimes very big enterprises?
Yeah, again, I see multiple models. In some cases, it's also a joint role with the chief data office. So in some cases, it's an extension of the data organization. Um, in other cases, it's really just to ensure that, um, the organization has strategic intention to get as far and as fast as their competitors. So again, I don't necessarily personally, um, meet a single mold or a single place that the role came from. Um, but I do think it's a great signal when an enterprise says, "Hey, we have a CIO," because at least then you know there's a voice and a center of competency, and the conversation matters to the CEO and to the board, or they wouldn't have created yet another C-suite title.
The other pattern that we have seen is that, uh, we're also increasingly seeing some of the chief AI officers straddle into the operations realm more on the business side of the realms, right? Sometimes by design.
Because if you think about, uh, attacking these AI problems, right? Like it's not just a tech problem, like we've always been saying, right? You're going and looking at, okay, how do I take top-down functions and top-down workflows and then reimagining them? How do I ensure that I have the right guardrails in them and so on and so forth? Um, which to your point will need a lot of coordination, not just purely from the tech side of the realm. Like you have to work with the, the business owners, you have to work with legal, you have to work with the compliance officers, etc. Right? So there is, it is a cross-cutting element.
And so those roles in some companies, at least, it's evolving into the operations realm. So I mean, crystal ball gazing is always going to be.
Yeah, I was about to say, like, so next year, more or like, is this trend increasing? What do you think?
Yeah. Right. So I think it's, it's always going to be like, we can all make those prognostication, but, see, there's, there's two ways it can play out, right? If it is viewed as, at least in companies where it's viewed as, uh, you know, this is an instrument for me to go run AI projects across the board or AI initiatives across the board, then it probably becomes completely pervasive. Then you probably don't need it at a steady state where, you know, AI adoption has matured to the level of everyone uses it and so on and so forth. In an ideal state, then you probably don't need someone going and wrangling and doing all of that, right? The, the other side to it is that, you know, we might be completely underestimating the, the, the way this role could be changing, right? So you're not thinking about, you know, we may be thinking about these workflow transformations, but
The capabilities keep advancing day in, day out, right? So we could be looking at some of these roles working very tightly with the CHRO in terms of how do you go and drive the, the people side of the, the conversation, things of that nature. So I think it's hard to guess, but yeah, it could go in a couple of different.
You want to jump in.
Yeah.
Yeah. So, um, couple of thoughts here. Number one is you have to look at the genetic makeup of the executive team. So having spent a lot of time in the CIO realm, having been one myself. Um, I will say a lot of this is is, uh, dictated based on the CIO that you have. So one thing to keep in mind is a CIO is not a CIO is not a CIO. There are different personas that might carry that title. Some are strategic, some are not strategic. They're more tactical. And so depending on the expectations of the CIO and how you view that part of the organization also influences whether you have
Someone that is responsible for AI. The second piece to that is, do you need someone that's going to help lead your AI function or that purpose within the organization that's going to bring together that cross-functional conversation? Absolutely, you need that. You need that AI council. If you are not talking about an AI council within your organization, I highly recommend you figure that out. Number two, number three is that I really question whether that is truly a CXO role.
I'm if you have the right CIO that is more strategic and very business-oriented, I think the CIO is not needed. You still need someone to to help guide that process, similar to a CDO, but is it a CXO level role where it's seen as a peer at the ELT and part of those strategic conversations? I don't think so, because we're talking about a very specific technology and a point in time. Um, if I go back in time, many of the CDOs that were hired several years ago were largely hired because they didn't have the right CIO in the role. And so the CDO was an augmentation of or extension of the CIO, but they gave them a CXO title. And so it's important to understand that genetic makeup and where you're going strategically within the ELT.
Well, as we get into the kind of home stretch here, Hillary, maybe I'll turn it to you. Uh, the final story I want to cover is was a new story in Crunchbase that launched, um, that I thought was just like a good reminder of the moment that we're in. And I think lurking behind the last few comments has been like, are we early in this process? Are we late in this process? Where are we? Um, so the stat that Crunchbase is reporting is that artificial intelligence funding in April reached $37 billion, which accounts for 66% of global venture investment last month. That's a lot. Um, but it also suggests, and I think the thing that I'd love to have you comment on, is that a lot of people still feel like there's a lot of room to compete. Um, there's a way of looking at this which is, well, the big companies already rule the space, there's no way to play the game. Do you want to just talk a little bit about whether or not you think the space is getting more or less competitive with time?
Yeah, I think to to share what we were chatting about even before we came into this conversation, AI has become such a broad technology definition now that sometimes it's like, where does the boundary of AI start and end in this conversation. Um, and I think a lot of of of what I see in startups in this space are actually trying to solve business problems. They are not all only trying to do fundamental large language model innovation or something like that. That's sort of what it used to be. Everyone was going to come up with their model. But now AI conversations and AI startups, a lot of them are trying to fix something for a business and actually deliver an industry-specific functionality or solve a long-standing problem in supply chains or things like that. They're kind of built as AI companies because they are using AI. But I think the whole field and probably the way these types of numbers are getting calculated represents a much larger contribution to
100% of money is going into computers, you know.
Exactly. It's kind of what you're saying. Yeah. Interesting. Yeah.
Um, so do you feel like this number is informative really, in some ways, I guess, is a testament to the fact that just AI is everywhere now, right?
Well, I mean, look, I mean, there's, there's an unprecedented level of, uh, interest and explosion in technology investment period, right? I mean, I think that's what that's what a lot of this is, is saying, and it's everything from infrastructure through to software, right? And and so I think, you know, it's, it's sort of all. Is there anything that any of us is doing that isn't AI at the moment? I, I sort of hope not, right? Because it is a very effective current technology and therefore, um, the way that startups kind of position themselves in many cases, yes, they're using AI, but they're solving a, a different larger problem.
That's great. Well, uh, Ami, Hillary, Tim, this is our first live episode. I think it went incredible. This is a great panel. So, thanks for joining us. Uh, and, uh, thanks to all your listeners. If you enjoyed what you heard, you can get us on Apple Podcast, Spotify, and, uh, podcast platforms everywhere. And we'll see you all next week on Mixture of Experts.