Cutting Through the AI Clutter with Ron Kim, Former Global CTO at Merck

Host Ariel Jalali, CEO and CAIO of Paragon, sits down with Ron Kim, former Global CTO at Merck and founder of Ron Kim Advisory, to tackle the question every C-suite leader is asking: is your organization actually ready for AI, and is AI actually ready for you?

Ron brings 35 years of hard-won perspective from global enterprise technology. He recently completed a five-year tenure as Global CTO at Merck, one of the world's largest pharmaceutical companies, and has previously served as CIO at Exelon Utilities and McAfee. He spent five years advising TPG's $120B private equity portfolio on technology strategy and due diligence, began his career at Accenture rising to Partner and Managing Director, and now offers independent executive AI and fractional CIO/CTO advisory services.

Ariel and Ron cut through the noise on the questions every enterprise leader is wrestling with right now. What does AI and agentic AI actually mean in practice? How do you tell real transformation from shelfware? How should a CIO get control of shadow AI and tool sprawl? And why isn't the first question always whether a process should even exist before you automate it?

The conversation goes deeper into the trust gap between AI startups and the enterprise, the process debt elephant that derails most AI initiatives, and real-world use cases and cautionary tales from healthcare, life sciences and pharma. Ron also shares his unfiltered take on whether the major AI model vendors are truly ready for the enterprise.

The episode closes with a direct challenge to leadership teams: AI transformation is not a CIO problem. It is a culture and change management problem, and the bottleneck, as always, is people.

For C-suite executives, PE operators evaluating portco AI readiness, and technology leaders driving AI transformation and automation, this episode delivers the practical, unfiltered intelligence you need.

Learn more about Paragon: paragoncto.com Learn more about Ron Kim Advisory: ronkimadvisory.com

Paragon - Managed Intelligence Provider (MIP™)

Transcript

SPEAKER_00 0:03

This is a series of podcasts that build on each other to help business leaders and C-suite executives think about implementing AI transformation that delivers measurable returns. Today we're exploring AI readiness from the perspective of a global CTO who is seasoned and experienced, and we'd like to provide our audience with two tangible takeaways. One, how do you assess whether your enterprise is truly ready for AI and whether AI solutions are ready for you? And two, how should technology leaders cut through all the AI hype to find real and measurable value? Our guest today has spent 35 years in the trenches of enterprise technology, leading transformations at some of the world's most complex organizations. With a rare combination of big four consulting routes, private equity rigor, and Fortune 100 operating experience, he brings a uniquely practical perspective on what it takes to make AI work at scale. Ron Kim recently completed a five-year tenure as global CTO at MARC, one of our planet's largest pharmaceutical companies. Previously, he served as CIO at Excelon Utilities and McAfee, spent five years advising TPG's 120 billion private equity portfolio on technology strategy and due diligence. He began his career at Accenture, rising to partner and managing director. He now advises independently, offering executive, AI tech, and consulting and fractional CIO CTO services. Ron is an active speaker on AI and healthcare, including presentations to SAPA, the Sino-American Pharmaceutical Professionals Association. Welcome, Ron Kim. Great to have you here.

SPEAKER_01 1:42

Thank you, Ariel. Great to be here. Always great to speak with you.

SPEAKER_00 1:45

Awesome. So the first warm-up question I give to our guests is an icebreaker usually pertains to their bio. So here's yours. Across these fantastic chapters of your career, which role do you feel gave you the best vantage point to understand AI's real potential and its real limitations?

SPEAKER_01 2:02

I have two answers to that. I mean, the most obvious one is going to be my my most recent role when I was the CTO at a Fortune 100 biopharma company for five years. And I just I just retired this past December 31st. And that's the obvious choice because I was there during the 2022, 2023 plus Gen AI explosion. So lived through the whole life cycle of kind of experimenting with Gen AI, scaling, getting our employees to do take up, then transitioning from personal productivity, like real transformational use cases, you know, trying to get out of uh eternal pilot mode and things like that. So really just learn that whole cycle in that one role. So that's the obvious, uh, obvious choice. I will bring up uh another part of my career that I think helped, and I'll go all the way back to the 80s uh when I was a computer science student. Back then, we coded, you know, by hand, no AI agents back then, AI algorithms. Uh so hopefully your audience realizes, maybe some of the younger members wouldn't, that you know, AI is a term that has been around for for decades. Uh, it's not just new with Chat GPT or things like that. So um we had classes uh in software engineering where we coded AI algorithms you know from from scratch, and they were primitive back then by by today's standards, but the principles were still are still applicable today. So just understanding, at least at a basic level, how those algorithms work, how machine learning works, having to code that that that uh that sort of logic uh by hand. Um I still apply that today. I mean, I think it still is a positive influence on how I view these wonderful AI capabilities that have now emerged, you know, 35 years later.

SPEAKER_00 3:47

It resonates with me as well. A lot of the patterns through the cycles of technology seem to repeat, things getting centralized and more distributed. Um and you know, all the best practices of um you know across the career end up being like a snowball as opposed to stuff that that gets dusty on the shelf. So definitely appreciate that. So, in terms of the um a good starting place, uh we we live in a in a period of rapid technological change. Some say that we're on an exponential curve, and this is the first time we're actually feeling it. There certainly is a lot of hype out there and a lot of clutter. And uh, you know, basic question what is what is AI? You know, when a leader says AI or agentic, what that should what should that actually mean to us?

SPEAKER_01 4:34

I'm seeing this question emerge more and more with the uh companies I've started consulting with um this year. So, first of all, I guess to to restate your question, there there are so many different definitions now of of AI, and agentic is even worse. You know, in 23 when Gen AI first came out, it was a little more clear then because the market wasn't so saturated with products claiming to be AI. So in 2023, when you said Gen AI or AI, you kind of people kind of knew what you were talking about. These days, uh all software products say AI on them. I would challenge any listener to this podcast to find software that doesn't say it has AI capabilities. So right off the bat, you're talking about AI has now devolved or regressed into meaning anything a computer does. Um I'm I'm half joking, but but that it's really created a lot of clutter because um the term AI, it's it's hard to differentiate what people are, and same with agentic. I was at um a large tech conference and on the expo floor and talking to vendors and saying, you know, we have this Agentic XYZ, and I dug into it. I said, well, you know, talk me through what this is doing. And they said, well, you know, it's this it's this agent and it gathers data from a database and produces a dashboard. You know, to me, that's just a piece of code doing something. You know, it's not an agent, right? So, in in my opinion, um, and there's very uh definitions of this, you know, AI has many, many different forms and meanings, you know, machine learning, classical AI, generative AI, et cetera, et cetera, et cetera. Generative AI, which has been most of the conversation the past few years, you know, is is really using algorithms and and intelligence to deduce and create new content, or you know, either summarize content, create new content, um, combine source of content, et cetera, whether that's written or visual or video, et cetera, et cetera, et cetera. And uh, you know, usually from some sort of LLM. So if if if none of those conditions exist and people are calling something AI, I mean, oftentimes it's really not AI, right? Um and then the agentic piece, you know, in my definition, in simple terms, is look, there's a piece of code that's got you know wired into LLM capabilities behind the scenes, and just like LLMs can seemly produce semi-autonomous or autonomous material, text or video, et cetera, um, this piece of code can use that same LLM, uh, those same LLM principles to make semi-autonomous decisions. Uh so they can act just like when ChatGPT first came out years ago when you said, wow, it almost feels like it's it's autonomous. You know, the the agentic term uh I think has to have that same element. Uh but again, I've seen a lot of vendors just simply call a piece of code doing anything an agent. Uh I I'll I'll share another story. I mean, I talked to a vendor recently and they said, well, you know, here's an agent that does your airline reservation. I mean, we've had code that does airline reservations for decades. You know, why is that called an agent? Um, so again, the but just I think people need to be aware, and they're probably feeling it, that the terms AI and agent are used very loosely and uh to mean different things. And um, you know, it's just it's just caused a lot of clutter and it and it's worse now in 2026 than it was in 2023.

SPEAKER_00 7:32

Aaron Powell Another distinction is you know, how do we tell the difference between real or deep AI transformation in the enterprise versus just giving everyone Chat GPT copilot or Claude?

SPEAKER_01 7:45

In my own experience and comp now I'm I'm an independent consultant, I work with a lot of uh startups and VC companies, you know, et cetera. Um so look, if you're using uh AI, some form of AI, to either greatly improve a process that that you have or uh you know, and or automate a process with more intelligence, or even not only make your current processes better, but make you uh enable uh your organization to do things it just couldn't do before. I mean, that falls more in the realm of I'm using AI to really transform my business. You know, I've also seen companies that said, we're an AI organization, and I asked, well, what does that mean? They said, Well, we've given everybody licenses to Microsoft Copilot. I have nothing against Microsoft Copilot. I think it's a great productivity tool. I use it myself, but though, you know, those are two really different definitions of being an AI organization. One is saying the first example I said is look, I'm using AI to just be transformational. I'm doing things that I simply couldn't do before this technology existed. Um, and it's transforming my business process, and there's real business value. Just saying I'm an AI organization and providing personal productivity tools, again, there's productivity there, there's value and things like that, but like they're just different parts of the spectrum, in my opinion. And I'll, you know, I'll give a couple real examples. Um obviously I came from the life sciences industry. I mean, companies in that industry, you know, are using all forms of AI to bring therapies and medicines and vaccines to market quicker. So, you know, as you may know, um, you know, to get a drug, it's a 10, 15 year cycle. And um, you know, if the companies are in that in that industry are looking to use LLM, build foundational models to find what are the causes of disease, what are better candidates, what are better targets, et cetera. And you know, that can really take uh the hope is that that takes a significant amount of time off of that 10 to 15 years. So that that to me is really transformational AI. That's different from, hey, Ariel, I just gave you Microsoft Copilot. Um, there's another example I just I just talked to a company recently. So um they're using a uh Generative AI platform to take products and generate uh video advertisements with no creative agency, zero, no middleman. You know, that just does this automatically. And this is this is a you know well-known use of of Generative AI, people generating videos for it. But again, to me, that's a transformational thing is I can create advertisements, video advertisements that can I can distribute electronically for my product without without using a creative agency at all. It can just happen automatically, and I can have tools and kind of do self-service, whereas before I had to pay X hundred dollars an hour for a creative agency. Again, I'm not dissing creative agencies. There's certain things that you know that they must be used for, but there's a piece of transforming that whole space where, hey, some of this can be done self-service on our own. Those those things are pretty transformational, you know, discovering new drugs, you know, having self-service to functions that I previously could never do myself. Um, I know there's a lot of different uh you know religious arguments about what means what's real AI is and what isn't, but that's those are some examples to make it real. And um again, I you know, I applaud the companies that are really just focused on the personal productivity piece, but you know, that's that's not the the top of the mountain, uh, in my opinion.

SPEAKER_00 10:52

And then in terms of uh company governance and leadership, bringing this up to the board level, um what are some hard questions that the that the board of directors should ask before writing the check for some of these AI initiatives?

SPEAKER_01 11:07

You know, I think over the past few years, as I mentioned, there's so much clutter uh has happened. I mean, the number one thing has got to be value, right? So what if I make these investments, if we take these, um spend some time and effort on these AI or Gen AI or agentic efforts, like what's actually the value? Like I'm all for some form of experimentation and innovation, and that that's great. So if some of your work is just you know the wild factor and doing cool stuff and kind of seeing what's possible, that's great too. But at some point, it has to be I can articulate how this is gonna change our bottom line for the better. Whether it's again, it's processes going faster, better, cheaper, it's processes that I just simply couldn't do them before. At some point, and again, I'm seeing this myself even now, even as a consultant, I'm seeing this now. The board member is gonna say, well, wait a minute, this Gen AI stuff hit in late 22 and 23. And where's the where's the massive value? Like where is it, right? And I think that's what uh companies are struggling with now. And so if I was a board member, I'd say, like, all these great AI initiatives you have, you're talking to me about, they sound wonderful. What's the value gonna be and and when? And some of them are you're not gonna be able to quantify, but you have to have some that you say, look, we're driving towards being able to do XYZ again, better, faster, cheaper with more innovation, et cetera.

SPEAKER_00 12:22

So the next set of uh questions are around uh an enterprise being ready for AI itself. And uh, you know, how do you how do you make sure to answer the question, is my enterprise ready for AI? Um, the first topic is really around shadow AI. It's everywhere. On one hand, you want to encourage curiosity, but how does the CIO get their arms around it?

SPEAKER_01 12:46

This is a tough one, and it changed, it's changed quite a bit the past couple of years. There's no world where just IT and just the CIO does AI stuff. I mean, just each passing day, more AI capabilities are embedded in tools we use. Sometimes even we don't even know about them, right? I mean, you can do Photoshop and it's using LMs to generate, you know, generate images. So I think it's a matter of, you know, where's the practical place? What are the principles to say, what does IT do? What do we need to worry about or be concerned about and manage? And what are the things we say, you know, as long as these things are in certain guardrails, go for it. It's a little bit of the same discussion we had with low-code, no-code tools, but it's just that on steroids. But I I even still today hear about companies saying, you know, we want IT to be the single place where all AI is done. I mean, I don't even know how to answer how to answer that question. It's impossible. So no matter what, there's got to be principles and governance. And I know nobody like nobody likes the G-word governance, but there's got to be nimble governance that says, look, at the very at the very least, I'm, if there's activities going on outside of IT, it's safe and secure. Um, and then there's other, there's got to be other principles that say, look, if you're going to build something and it goes, it's going to be deployed to production, you know, whatever that means, there could be different forms of that. Um, there's certain architectural and and technology process things that you have to follow before it goes into production. And um, you know, these things are evolving uh all the time. I mean, I I will I will tell you, you know, now most large enterprise software platforms have some sort of LLM and Gen AI capability built within them. And then you get into, well, my governance has to look at those and say, where did those models come from? Are those okay to use? Is my data safe or getting exfiltrated, et cetera, et cetera, et cetera? Um the the final thing to your question is around agents. So within the past 18 months, um, every large and probably small software vendor says, we have agents and they can run through your whole company. So is that okay? You know, is it okay if uh Enterprise Platform X says, well, we can go into Enterprise Platform Y and make changes in that system company? You know, we can we can run the whole thing. Um, there's a lot of vendors vying for who owns air traffic control for these agents across the enterprise. I mean you've got you know agent core, semantic kernel, Google Enterprise, et cetera. So from an IT perspective, it'd be good to say, you know, where do you draw the line? What are the things IT controls and what are the things we care about, and what are the things we just know are going to continue to happen outside of IT and if we wrap the right governance and security principles around those, and not a lot of red tape, but making it so you know people can still be productive. And I'll tell you, it is it's not easy. I mean, these are these are issues that occurred before AI, but just A the pace of change of AI just makes this more more difficult and and uh subject to change.

SPEAKER_00 15:31

I'm glad you talked about the uh the land grab in the enterprise uh software space. Everyone's anxious and excited, and they're trying to sort of capture and be that centralized one AI to rule them all. You know, every every vendor wants to plug into their into your stack and be the, like you said, traffic controller. So how do we, you know, how do we stop that sprawl and keep everyone in in their lane? Is it by asking, you know, how does this eliminate three other tools?

SPEAKER_01 15:58

I think the pace of change uh that's currently happening with AI technology is kind of starting to fundamentally change how we decide to make technology decisions and investments. Uh and I was just in a conversation about this the other day. I have seen myself this concept of what I call disposable technology decisions uh starting to emerge. Meaning in the old days, you know, when I was a CIO and I've been a CIO and CTO, you would make technology decisions and say, well, look, you know, I'm making a big bet, I did a big study, I'm gonna buy this technology, it's gonna stick around for you know two or three years, and then we'll have to figure out what to do next, you know, whether to continue or change. Now the landscape is changing over weeks and months. There is no dis there's there is no AI decision where that's gonna stick around for two or three years. I mean, I can I can talk about all the changes happening in the industry, and they're they're so rapid. So then the the reaction to that was I saw companies get stuck in this analysis paralysis, like what do we what do we pick? Cognition for our AI coding assistant, or what about Windsurf? Wait a minute, cloud code's getting better. What about cursor? You know, oh well, wait a minute, cognition bought Windsurf. You know, it just it's crazy. So um, if comp if organizations want to move forward, I think they have to make, in some cases, what I call disposable decisions, meaning, you know, we know this is going to be short term. It helps us move forward, uh, but I'm not as hung up on this has to be the decision that sticks around for two or three years. And uh I, you know, that that has a whole bunch of implications to come with how do you measure the ROI of that? I mean, if you're making a disposable tech decision, you know, it's the so there's a lot of changes that have to happen fundamentally, I think, in uh and I and that are happening about how CIOs and CTOs think about making those uh technology decisions. Because as you mentioned, all the vendors are trying to do a land grab and the land is changing under our feet.

SPEAKER_00 17:38

So uh Jeff Bezos calls them two-door decisions, but I like your term even better. Disposable decisions has a better ring to it. And um, you know, I I guess that to wrap that up, though the decisions are disposable, the North Star isn't. The North Star is always value creation, capital efficiency, um, doing the right thing by the customer, those types of goals. And that's really, you know, the enterprise focus. So that sort of brings in a whole flip side of the coin. So we've talked about is your enterprise ready for AI? Now let's flip it and say, is AI ready for your enterprise? And one of the challenges that we see is that tremendously innovative platforms are getting spun up, it seems like every day, uh, with lots of money behind them, lots of VCs, et cetera. Um, you know, the Silicon Valley kids are are have been very busy, but there's kind of this inherent trust gap with with enterprises, especially as you get, you know, mid-market, upper mid-market. And with respect to that trust gap, like, you know, I heard recently that one of one of the VC funds um actually acquired a hospital so that they would have you know a test bed and for all kinds of economies of scale. And what does something like that say about the AI to enterprise trust gap?

SPEAKER_01 18:54

I don't know if I'd want to be a patient in that hospital, actually, but yeah, I think I think what you said is is really true. Uh the so the again, I've worked on both sides of it, kind of the PE VC side as well as the enterprise buyer side. In some cases, there's a disconnect in what they're selling, what the enterprise is looking for. And you know, one of the one of the things that comes up often, and I think this will resonate with some of your listeners if they're kind of corporate CIOs or CTO types, is all these new tools and vendors come in front of us and they say, you know, you gotta buy this, it's great. And there's two questions that I would ask if I was sitting in a corporate CIO or CTO seat. I'd say one is, are you gonna be relevant in six to twelve months? Because this, you know, this this cycle of acquisitions and new new capabilities coming out. So you gotta, you know, say, how relevant are you in six to twelve months? The second thing is I don't want to keep adding tools into my environments. If a tool or platform vendor comes to me, large or small, they have to have the message of if you partner with us, we'll give you line of sight to get rid of two or three other things. That's a line of uh conversation I I rare, I see some vendors have that, but I rarely see it. And so for any any vendors out there saying, Boy, I have, you know, I have this thing, it's a sure thing. Why do people get turned off uh when I when I try and talk to enterprises about it? You know, it's because the enterprise are getting bombarded with vendors every day with these new AI capabilities. And you know, the first thing we want to say is we don't want to keep adding on to the the ball of string. You know, tell tell me how you can take stuff out. Um so that's that's a that's a piece of I would consider AI, if I was still in an enterprise, I would consider AI ready for my enterprise, more ready if they had that kind of messaging, you know, right up front, rather than just keep, you know, keep uh keep adding on. But that's the big thing is look, are you gonna be relevant in six to twelve months? How do you how can you help me simplify my environment as an enterprise buyer and not not add on to it? And how are you differentiated? And uh as I mentioned before, that the pace of change is so quick, it's tough to make a big bet anywhere.

SPEAKER_00 20:45

There's a lot of hype coming out of the the hype machine. You know, there's this term SAS Pocalypse. They're like, okay, well, we can now vibe code everything. Uh I don't think that's a hundred percent realistic. I think that you can clear some technical debt, certainly. Um and then, you know, there that's one option. The other option is to go with some of these new uh new tools, new startups, and the other option is to do an internal pilot. So the the question is, you know, is there a tangible difference between a startup and an internal, an early internal pilot?

SPEAKER_01 21:19

The benefit a startup is gonna have, again, that I've seen uh uh these past couple of months is you know, they're gonna they're gonna have uh feedback from multiple enterprises that they're you know, you would think that they're baking into their system. Whereas an internal pilot, I'm really gonna be most of my explorer is gonna be how does my own organization work? So I'd say the breadth of the data and experience that goes into what they're building as a as a startup uh is probably could could be broader than just being within a single enterprise. But both of them, I would still apply the same value questions of are you gonna be differentiated in in six to twelve months? What's your capability? And and also can you help simplify the environment? I would I would ask that even of an internal if they was building something internally at an enterprise and we're investing time and money, it's like what's what's the end. What's the end goal here? Like are are you gonna be differentiated or is this something that you know anthropic is gonna come out with in three months?

SPEAKER_00 22:05

I mean to your point, even ask them, let's be honest with ourselves, what's the probability that this becomes a disposable decision?

SPEAKER_01 22:12

Yeah, yeah, 100%. This disposable concept, um and the term is is relatively new, but it's something I'm seeing and something I had even felt. Uh, but I'm seeing it when I talk to, you know, as I've as I've been consulting the past couple of months, and uh that's just the term I coined. I don't know if there's a better, better language for it, but but it's real. It's real. You just you have to you have to have that mindset in order to move forward. It's just part of the deal now.

SPEAKER_00 22:36

And and then the final question, you know, in terms of of change and enterprise uh readiness for for AI, whether it's an internal pilot or a startup, do startups and or internal pilots maybe get beaten down sometimes by those of us that are more status quo, you know, the CIO that's a little bit more status quo. And uh is there a risk of making making these making these efforts gun shy?

SPEAKER_01 23:00

Yeah, I I think it's certainly possible. Um, here's what I would suggest to prevent that from happening again, set having sat on both sides of the aisle. Now in 2026, we're I think we're clearly in, you know, or we're we're emerging from from pilot pilot fatigue. So if a startup or um internal pilot is going to um internal effort's gonna come talk to a CIO or a board or something, um I think the most effective thing would be they have they have line of sight to scaling. Like they're either they either have done it or they've looked, this is not just a pilot, but this is a real thing with real value, and it's not just a you know a cool experiment we're gonna do. You know, I again my earlier time frame, I mean, I think I think 2024-ish was 23, 24, maybe was the period of you know thousands of thousands flowers blooming. I mean, a ton of pilots, but nothing really going to scale. So I think what would if I was a CIO, what would make my ears perk up is, hey, there's a real line of sight for this thing to actually scale and go to production and and demonstrate value as opposed to, hey, this is a cool thing we built and it's got a lot of cool agentic bells and whistles. And uh, if you can have that conversation, I think people won't feel as beat down. And I, you know, I'm I'll go back to a concept I bring up a lot. Technology success isn't really about the software code or bits and bytes, it's about you know how you how you manage the conversations, the human aspect of it. Uh uh address concerns, address doubts, build confidence. That that hasn't changed my whole career. I mean, I from I remember doing projects on the mainframe all the way to now, you know, AI project. I mean, really what makes or breaks them is can can the vendor or the the owner of the project, when they're talking to stakeholders, um manage the change management issues, the concerns. This is gonna take my job, is my budget gonna get cut? Do I really want to do this? In my opinion, those people issues um are as much or more of a factor than the code or the containers or the you know the data, et cetera.

SPEAKER_00 24:51

So and it kind of is a natural segue to talking about what I call the process debt element elephant in the room, which is you know, we've talked about technology, people, and process. A lot of times our our technical debt and our data debt are sort of shadowed by this process debt. The first question there is like, how much of potential AI failure is really a process problem and not a technology problem?

SPEAKER_01 25:17

Yeah, I think I think a fair amount is I've seen this firsthand. I've probably done it firsthand and I've seen it talking to companies I'm working with. Um there's a big rush to throw AI or automation or agents at a certain part of a process. Hey, let's really step back and do the whole process mining. And what they end up doing is, you know, agentifying a piece of the process and just moving the problem around, moving bottlenecks around or things like that. This this goes back to anyone, I mean back to RPA, you know, years ago. I mean, people say, look, I created a I created an agent or a robot to um automate this PDF file creation. Well, nobody stepped back and said, Does that PDF file should that even be created? Like, what's the point of that, right? So I think I'm restating your question, Ariel, but the um, but yeah, I I think the the make or break piece on a lot of this stuff is does the organizations have a set of processes, do they know what their processes are? Is there some sort of process library? I don't want to get too cumbersome, but is there some sort of process place where they say, oh, here's all the things our company does, and it's documented, and this is how we do them. And then and looking at that um with a process entering mindset and then applying the AI or whatever technology you want to apply to it, that to me is gonna maximize the chance of success. I mean, I've just seen that over and over, and I've seen the converse where people just take a piece and say, look, I can make an agent do XYZ, and it it's not, it's not really gonna scale. It's not really gonna there's no real enduring value. I think organizations know that on paper, so I think many people would say the same thing. I think just putting that into action is the challenge.

SPEAKER_00 26:47

I'm starting to to see a concerning type of graphic floating around, and it's basically a file and folder structure that basically looks at different jobs and then has specific tasks in there that are automated for for each agent. So essentially the the existing processes are codified into jobs, and then assuming that a job is a collection of tasks, how many of those tasks can we automate? And and I start getting stressed about that because I feel like maybe we're going down the path of you know, we keep on building technology, whether it's ERP, whether it's SaaS, on top of processes that serve legacy systems and and change and maybe even change resistant people. And um will we do the same thing with AI?

SPEAKER_01 27:35

It's possible. I mean, I I always think when these big, big, huge headlines come out, like AI will replace all consultants or you know, things like that, uh, or AI will replace all coders. I think there's some truth to those headlines, but usually, you know, the it's usually the truth is not on the extreme, right? It's it's you know, so people aren't close to it will say, hey, well, software engineers write code, Claude can write code, so all software engineers should be gone, right? Or consultants just do research, you know, you can get that through, you know, Anthropic. But if people really know those jobs, it's not just the writing of the code. In fact, you know, I'm an old software engineer. I'd say, you know, the actual writing of the code is a fraction of getting functionality from birth to production. There's a lot of other steps that go into that. Design, testing, you know, et cetera, et cetera, deployment, CICD. And again, I know these software tools are getting better at that, but my point is that um when the extreme headlines come out, there's usually the truth is somewhere not quite there. And it's the same, same thing with consultants. Um, how many times have you read? I don't know, Wall Street Journal, New York Times, all consultants will be gone with AI. I was a consultant for 17 years. Uh, I was a partner at a big consulting firm. I have been a client of those consultants since then. Uh, I know people who still work for those firms. I didn't just sit behind a desk and do research. I mean, a lot of that consulting job is like all the client interaction. It's almost like your client's personal therapist. You have to be their personal therapist or marriage counselor, whatever you want to call it. There's a lot of like human work that goes on that I just laugh when I see, like, oh, AI will replace all consultants. Now, will it replace some of their work? Yeah, 100% absolutely. The right evolution, I think, is people, let's say I'm a software engineer. Instead of me just saying, well, Claude took my job, I'm gonna say, well, look, I'm gonna learn how to use these tools, and that's gonna be a force multiplier, right? Or I'm a consultant. I'm gonna learn how to use these tools and become a better consultant. I'm just a force multiplier for me. So I think when you really know the nuances of any particular job, I think you can make a better judgment. Oh, well, will I replace that or not? And again, in some cases the the the job may change, uh, but but I still think there's in in my opinion, in a in a lot, there's some cases where jobs get eliminated. But in my opinion, there's a lot of a lot more cases than people think where the human has to adjust what they do and say, I'm using this now as a tool to become better. I was uh there's a really basic analogy, you know, uh next to my house there's a house being built, and uh I was watching it, and um, you know, I was watching these these construction guys use some tools to like you know build a ditch and extract a tree and tree trunk and things like that. Well, you know, there was a time in history when they had to do that manually. Well, now there's still construction guys, they just use those tools to do what the tools can do so they can focus on a higher-level stuff. So um I know I know it's not exactly the same, but you know, the some of the principles are are the same. So I so I think to your question, you know, yeah, there's gonna be jobs that change, and in some cases jobs that get eliminated. But um I also think there's a lot of cases where people will or can take advantage of this AI and say, this is this is a tool for me to my job will will alter or change instead of maybe I'm not gonna be a hands-on person as much, maybe I'm orchestrating AI capabilities instead of doing all the work myself. But there's still gonna be opportunities for people, especially people that are thinking about it, you know, kind of with a big picture, big picture in mind.

SPEAKER_00 30:50

And you know, getting getting back to the point about you know, uh empowering leaders to more of them to ask the question, should this process even exist before automating it? Are there some some practical guidelines or advice that we could give them to sort of empower them to ask that question more?

SPEAKER_01 31:10

Yeah, I think this is largely this goes back to this is largely one of those people issues. So I was I was working with a company recently in 2026 that that had this, was having the same conversation. And so, you know, they said, look, we looked at some of our processes, like I got called the department heads together. We looked at some of these processes. There's a few of these that just we don't need them at all. Well, that means a certain department head is gonna lose budget or lose people or something. So you got to believe that guy, that person was like, oh no, wait a minute. No, we need these. Let me let's let's start a bit. I'm gonna we're gonna start a couple week study and presentation to tell you why we need these pro. I mean, that's that's what you're getting into. My point is a a lot of the the process questions, you have to just be willing to work through um some of those conflicts and human human pushback. Um, it's gonna take that kind of discipline. And again, you could argue that's not an AI specific thing, but AI is causing us to look at those processes more. Even in the short time I've been an independent consultant a couple of months, I've run into that a couple times. Like, hey, let's let's apply this AI to this process, and maybe we don't need these people, uh, or they, you know, we don't even need this process. And um many people who own those processes aren't immediately going to say, yeah, that's a great idea, let's eliminate what I own.

SPEAKER_00 32:19

Absolutely. Especially if it's just, you know, moving things around or doing data entry or just a rudimentary task, there's there's a lot of opportunity there. So digging a little bit deeper into AI and healthcare and life sciences now, that's you know, been a good portion of your background. Um, wanted to ask you, what are some legitimate use cases versus hype for AI and life sciences, pharma, healthcare, writ large?

SPEAKER_01 32:48

There's a number of very legitimate uh use cases emerging that I I know a number of in in the pharma space, I know a number of uh companies are are working on, and I know a number of AI providers are building capabilities for. Um I'd say the two that are most uh prevalent now and have a lot of excitement are uh clinical trials. So using AI to enroll people in clinical trials, keeping the people engaged, making sure the people that uh sign up for the clinical trials, you know, that they're they're relevant participants, the trial is effective. Um I know the the clinical trials process isn't probably known by many people, and I'm not an expert at it, but uh oftentimes um what you want to avoid is investing all the time and money in a clinical trial, and you're not getting the bang for the buck out of it. So the you know, this is a great use case where AI can determine look, this is the most promising population or the most promising set of characteristics to have this clinical trial and you get the most bang for the buck. And clinical trials are really what gets drugs to market. The second thing in pharma is just around drug discovery. I mean, I mentioned this earlier. Um, one in 10 uh drug candidates uh enter clinical that enter clinical trials ultimately receive approval. Um, and again, as I mentioned, this is like a 10 or 15 year process. So anything that A helps you find candidates that have a higher chance of approval and B can potentially accelerate that time frame. I mean, that's huge. That has a huge impact on society. It doesn't get as much press because you know, some of the durations are pretty long. So I mean, it's people are it's flashier to talk about something that, hey, I AI just made this happen instantly. You know, if you talk about, hey, AI changed this clinical, you know, the duration of a drug going to market from 11 years to nine years, I know it doesn't sound that flashy, but uh, but that's really has huge implications uh on um you know, on the betterment of society. So so just you know, again, around discovery um and as well around uh optimizing clinical trials, those are those are legit use cases. And I uh I would say I'm out of the farm industry now. I'd say every every large farm is is working on that or thinking about that. There's some other earlier use cases that were not quite as ambitious but still move the ball forward. I mean, the one one I'll mention is I mean, I think Gen AI is is uh uh great for authoring materials. You know, just like you can go into ChatGPT, write a story about me, and it just comes up with this story, it's it's really good at doing that, and it's good at summarizing information and consolidating information. Um, you know, in the pharmaceutical industry, there's there's a pretty arduous um regulatory submission process. I mean, these are, you know, these scientists have to spend a lot of time on uh correlating documents, uh references, tables, footnotes, et cetera. I mean, these this is just this is you know weeks and weeks and months and months of time. Well, if you can feed the LLM the right content, the right language, the right tone and tenor, the right terminology to use, the LLM can draft that stuff in in you know, minutes or hours rather than days or weeks or months. And uh also it um frees up the scientists to work on the science. I mean, I I think most scientists who studied hard to get their PhD in chemistry or neuroscience, they probably don't love to spend a ton of time being a copy editor. That's a a well-known use case in the in the industry. So I mean that those those are three examples, you know, authoring, um, clinical trial optimization, and just drug discovery in general, that um it's not it's not uh fantasy. I mean, the the industry is working on those use cases actively, I can say that confidently.

SPEAKER_00 36:09

And it's it's interesting. It there's kind of a common vocabulary that's that's coming together, which is like there's three there's three phases of AI. The first is like as research as research tool. Um second is sort of as a coworker uh or or co-pilot. And then the third is as an actual agenc replacement for for human tasks. You know, earlier in my career I I helped build a life science research center with like nine different content types for bench scientists, and that was like running on search engines and things like that. That was squarely in phase one. I feel like we're kind of now in phase two, maybe 2.5, and then you know, heading head we're heading into three, but I don't think we're there yet.

SPEAKER_01 36:55

I agree with you, and the things I will caveat um this whole conversation with is there's really smart people that have been working in the farm, really smart scientists that have been working in the farm industry for a really long time, and AI, forms of AI have been being applied in that industry for years. I think the current forms of AI and Gen AI and Agentica are wonderful tools going forward to again accelerate some of the processes I just mentioned or make progress. But you know, some people heard the you know, the Larry Ellison comments uh, you know, a couple last year on, you know, we can create an effect. AI will help us create a cancer vaccine in 48 hours. Look, he's a lot smarter than I am. I'm all for setting the bar high. Uh, but people have to realize like this is not a new space uh of AI. I mean, this is AI's been uh applied here uh for years, so I don't think there's gonna be a miracle that happens uh tomorrow. Uh but but it it can legitimately make a huge dent in some of this use cases I just mentioned uh in the in the pharma space.

SPEAKER_00 37:51

Are there some some cases or examples from your career that you can talk about at any level that highlight sort of what worked and what didn't?

SPEAKER_01 37:59

I mean I'm gonna go back to some of the um some of the principles I mentioned. I can I can talk about real examples, but I mean a lot a lot of the a lot of the success of these things, it's it's the people factors. You know, it really it really is. So can you get people uh comfortable with making bold decisions and taking risks? Uh we just talked about disposable technology. Can you get people comfortable who can convince their leadership that we have to make investments and you may not see a payback in the traditional sense? Uh may I can't tell you that there's gonna be positive ROI or not. Can you find people who instead of focusing on, well, I think AI is going to take my job, shifting that mindset to I can use AI to do different things and do them better. You know, as corny as that sounds, I mean, that that's when I look at the patterns, like the pattern recognition across the companies I've been exposed to that have been involved in AI, whether it's my own company or peer companies I talk to or now other companies I'm working with as a consultant, to me, that's the foundational pieces for success. I mean, that the technology is, there's always gonna be cool technology out there you can buy, and the technology does what you what you tell it. But I I ironically, I still think the the human leadership and vision to navigate through all these things I just mentioned, the fast, the pace of change and all that kind of stuff, is gonna be what really carries the day. And that's what I've seen.

SPEAKER_00 39:13

So the uh the big AI model companies, and and they've they've been incredible. They've been shipping something every every few days now. They're telling us that they now have enterprise ready AI. So um enterprise ready AI in healthcare is apparently here. Um, should we believe them? And and and should we sign on with an with a large AI model vendor?

SPEAKER_01 39:38

I think those big model players are always gonna play a part. I mean, they're growing more and more capable. I mean, I'm starting to think that the enterprise, I'm starting to see the enterprise, the big huge frontier and enterprise models are starting to also become AI coding agents. They're starting to replace that whole industry of AI coding agents. I mean, that's the trend I'm seeing. Um, I think those they're always gonna play a part. Uh, I think those large models should be mixed in with industry-specific models with that expertise. In other words, I've worked with and still talk with the the big players. And let's just take, again, I'll take the healthcare life sciences for example. They'll say, well, we hired a bunch of scientists to make it a life sciences relevant model. I just think those big models are are so uh they're so broad. It's kind of like, you know, an athlete, right? You can have a general all-around athlete. Sometimes you need the specialized athlete. I think in a place like um pharma life sciences, the specialized model still is valuable. I wouldn't just say I have can have an all-around model that's got some some healthcare and life sciences elements to it. Uh, and I know I'm simplifying that because I'm I'm sure the models are very good, but but um but at least in in that life sciences pharma space, what I've seen is look, the companies that are really focused, look, their whole staff is deep scientists who've done this before, and that's who's building those models as opposed to it's just happens to be one flavor of a big, huge model. You know, I I think that's that's important to recognize. And sometimes, like I said, sometimes you need the special the specialized athlete, you can't just rely on the all-around athlete.

SPEAKER_00 40:58

Yeah, and uh certainly that that mirrors what we're seeing, and there has to be an intelligent way for their for these tools to hand off to each other and know what domain they're good at. You know, you can't always use a sledgehammer. Sometimes you've got to use a small Phillips screwdriver on something. So it's a it's an integrated toolbox. And I think that's for us, that's gonna be the exciting part over the next few years to see how the fullness of the enterprise tool set comes together. There isn't really a great at this point of recording a great orchestration layer for the agents. Everyone's running around with their with their claw bots uh, you know, uh and hacking those together, but those are certainly not enterprise grade. I think we're gonna start seeing those coming through.

SPEAKER_01 41:40

There's already agent chaos happening at companies. Um because what here's what happens is people say, Well, I have agents in my company. And for my earlier comments, agents are different. They're defined differently, right? And then somebody says, Well, I need to have a layer that kind of watches all the agents. Well, what's what's it gonna watch? Because do you do you do have you defined what that agent, you know, what those agents are and what type of agents you're gonna orchestrate? And then are the agents being built or bought in your company, do they fit that definition? If not, you're gonna have a bunch of things flying around and you can't expect to manage. You know, there's just and I don't, I'm not saying I'm smart enough to have the answer, but I I do I have seen this is a challenge, you know. So I agree with you, this or agent orchestration concept. You know, look, I think some of the big play I mentioned before, you know, agent core and semantic kernel and you know, etc., I think they're gonna have industrial strength things that handle much of the problem, but there's still just some governance to put in place before you can say, Do I have do I have control over that?

SPEAKER_00 42:34

Which brings us back to our favorite topic, which is uh people. And uh, you know, wrap wrapping this up, you know, Ron, 35 years into enterprise tech and and you mentioned it.

SPEAKER_01 42:44

The people relationship aspects, I think, are much more make and break. They've always been make or break issues for a CIO, but I think they're even more now than ever. CIOs or CTOs are much less siloed. It's not, you know, 20 25 years ago, you know, IT is almost like a back office function. Those IT guys handle that, you know, we don't do that stuff. Well, now it's kind of you know intermingled with the business. You got people in the business building tech stuff, you know, how does that how does that intermingle with what IT is doing? So um the places I've seen that have the healthy, free-flowing conversation relationships between IT and business are the ones that kind of avoid a lot of those big pitfalls. You know, the other thing I'll say, and I I don't this has kind of probably been happening for a while, but um, you know, there used to be a time when a CIO or pretty senior person could be kind of you know in the ivory tower and you know, kind of managing things from a from uh from on high. And um I've seen a trend where today's CIOs and CTOs, at least in my experience, even at big company, big companies and small companies, there's a um there's a no jobs too small mentality, right? So even you're a big corporate CI or CTO, you know, getting your hands dirty, showing that you can get your hands dirty, just learn learning some of the details, you know, rolling up your sleeves. I've seen that more now with the Pace of change than than ever before. And CIOs and CTOs that I've have more credibility, whether it's with their own teams or at boards or things like that. And again, I've observed this as a third-party observer that say, you know, they'll say, look, wait a minute. I built some agents and I saw firsthand uh how difficult it is or how difficult it can be to have proper governance over them because I because I've done I've got my hands around, I've done that myself. That's really powerful. And uh I think when things were more stable and technology wasn't changing, you know, maybe maybe a CIO or a CTO could afford to sit a little bit removed from those details, but stuff is changing so quickly that um I think it's a necessity to be successful, to change that mindset and uh be able to roll your hands, roll your sleeves up. And again, no jobs too small, right? Once in a while, you gotta dive deep and and learn. And um look, maybe generationally the the CIOs and CTOs of today already think like that anyway, but I know not all of them do. Let's put it that way.

SPEAKER_00 44:56

So well, it's certainly the era of the player coach, and uh and you're one of the one of the great one of the greats, Ron. So thank you for uh for joining us. I appreciate your friendship and uh partnership. Uh and uh and I wish you wish you luck on everything that you're working on.

SPEAKER_01 45:13

Great, great speaking with you, Ariel. I really appreciate it, and uh always good to connect.