Leading Through The Realities Of Enterprise AI w/John Willis

AI is no longer a science project you can hide behind a pilot. Boards and CEOs want enterprise AI that shows up as measurable ROI, operational impact, and real transformation and that demand changes what “good leadership” looks like for every CIO and CTO.

I’m joined by John Willis, a DevOps founder with decades of experience protecting the brand in high-consequence enterprises. We dig into the messy middle between AI adoption and AI integration: why the world is shifting from digital transformation to decision transformation, what happens as teams move toward AI agents, and why “easy to start” can become dangerous when executives mistake prototypes for production systems. We also talk about the decisions leaders cannot avoid right now: prioritization under constant pressure, budgeting and token economics, buy versus build chaos, and the reality that legacy systems still have to coexist with modern AI tooling.

Then we get serious about risk. Cybersecurity timelines are compressing as adversaries gain AI-powered speed and scale, which makes governance and operational readiness non-negotiable. John lays out why responsible AI governance is less about slowing down and more about creating flow with guardrails, evidence, and audit trails so you can defend decisions when something inevitably goes wrong. We also explore the growing IP ownership and copyright gray zone around AI-generated work and why accountability still sits with the organization, not the model.

If you’re leading enterprise AI strategy, model risk management, data governance, or large-scale AI transformation, this conversation will help you separate hype from reality and build a shared language for what autonomy you will and will not allow. Subscribe, share this with a fellow leader, and leave a review with your biggest takeaway.

Check-Out Publications by John: Amazon.com: John Willis: books, biography, latest update 

Referral Links

StreamYard: https://streamyard.com/pal/c/5142511674195968 

Contact Digital Revolution 

• Email: Jim@JimKunkle.com 

Follow Digital Revolution On:

• YouTube @ www.YouTube.com/@Digital_Revolution 
• Instagram @ https://www.instagram.com/digitalrevolutionwithjimkunkle/ 
• LinkedIn @ https://www.linkedin.com/groups/14354158/ 

If you found value from listening to this audio release, please add a rating and a review comment.  Ratings and review comments on all podcasting platforms helps me improve the quality and value of the content coming from Digital Revolution. 

I greatly appreciate your support and Viva la Revolution!

Transcript

AI Becomes A Business Expectation

Jim 0:05

Welcome back to the digital revolution. I'm Jim Kunkel, and today we're stepping directly into the reality that every CIO, CTO, and business leader is now facing head on. Because here's the truth AI is no longer an experiment. It's an expectation. The era of let's run a pilot and see what happens is over. Boards aren't asking for innovation theater anymore. And CEOs aren't impressed by proof of concepts that never scale. Investors aren't rewarding organizations that dabble. They want measurable ROI. They want operational impact. They want transformation that shows up on the balance sheet. And that shift from experimentation to enterprise-wide deployment has fundamentally changed what accountability looks like for technology leaders. CIOs are no longer judged by how many pilots they launch, but by how many AI-powered capabilities they successfully integrate into the core of the business. And here's the tension: AI ambition is skyrocketing while operational readiness is lagging behind. Many organizations are daydreaming big, autonomous workflows, predictive operations, intelligent customer experiences, yet we're still wrestling with fragmented data, legacy systems, and cultures that are not prepared for AI-driven change. That gap is widening, and organizations that fail to close it are already falling behind. So in today's episode, I'm joined by John Willis, and we're going to talk about leading through the realities of enterprise AI. John, welcome. Hey, thanks, Jim. Thanks for having me on your show. Oh, it's great to have you. For the for the listeners and also to the viewers, uh, could you please give a little bit of uh background about yourself and especially related to you know leading digital transformation and AI programs?

SPEAKER_01 1:58

Yeah, so you know, I'm old as dirt, if that's they say, right? I I've been doing this, uh I've been I've been through five decades of technology shifts, if you will, right? Um won't bore you with it, but I started actually as an IBM mainframe assembler coder, moved to distributed computing in the first wave, somewhere along the line, got in, you know, skip a bunch of stuff. Cloud, I was sort of early in on helping enterprise, always enterprise. And I, you know, there's there's this sort of movement called DevOps, and I'm one of the founders of the movement. And I consider myself the ops side of the DevOps. I've always been infrastructure and operations, um, in high consequence large enterprises. So um I'm always sort of my my role has always been the tip of the spear is what is this new technology shift going to do to the brand? And to me, the brand is, you know, ops is about protecting the brand, you know, when it's all said and done. And there's nothing more important in organizations. Anyway, so my career is trying to get ahead of new technology shifts, cloud, um, you know, DevOps, I was a big part of that, you know, lean and agile was smished in somewhere. Um, but um, you know, more recently, um, you know, I've been sort of asked by my clients, you know, hey, we're you're can you help us? Not that I'm the oracle of all things, but I do have a pretty good perspective of all things around, you know, what are the things that can really get you in trouble in an enterprise, particularly a highly regulated organization like a bank, healthcare. Um, and so I've been in organizations probably since even before ChatGPT, but certainly after ChatGPT, the phone started ringing pretty heavily. So I've been spending a lot of time with some very large enterprise customers and trying to navigate um, you know, what does all this mean? You know, some of a lot of things you talk about on your podcast. And and just to pinpoint, I really focus in on what are the risks that technology transformations literally can

Enterprise AI Shifts To Decisions

SPEAKER_01 4:00

cause harm to the brand. And so I'm but I'm very geeky too, so I'm uh nuts and bolts, you know. So stop me if I get too geeky and too technical, too.

Jim 4:10

So perfect. And uh what I wanted to open up uh this episode on is to talk to you about you know the enterprise AI landscape. And we've seen for the last number of years a lot of promises, and now we're starting to see businesses and also to other um leaders when it comes to AI. You know, we need to make that leap into the reality. And, you know, some of the areas that I've seen where there is some real value today is obviously in dealing with the repetitive, you know, dealing in automation aspects, um, even in decision support, uh, the customer experience, and and we're also starting to see a lot more predictive operational type of information coming from AI. But we're still that where hype is still out there and it's kind of still outpacing the capability. What I wanted to ask you here is you know, the difference that you see between today, AI adoption and AI integration when it comes to these complex organizations. Where are we at currently related to AI adoption and the integration into businesses?

SPEAKER_01 5:20

Yeah, a good friend of mine, Simon Rolle, is always say, um, situation normal, everything, and I won't curse. Um, you know. Um, you know, I I think uh the the thing I I I I've been on a quite a few podcasts, uh, you know, I usually only do my own podcast, but um I realized uh my latest book is not selling great because people aren't reading, so I'm trying to get to new audiences. So hopefully later we can talk about a couple of my books that I've written. But um, but so I've been getting on other people's podcasts, and and you know, one of the things I realized is the one thing I I think about is whenever we're in one of these transformations, we have a hard time looking from outside in. So everything looks terrible right now, everything looks new, everything. Um, and and I'm not saying these aren't all true, uh, but but I what I try to do is sort of pull myself out and look at, and and so one of the things I think a lot of people are not and probably will see down the road when in a retrospective of like as this gets more mature, is we're we're in a shift right now. We're moving really from, and I don't think anybody really sort of explains this or understands, we're moving from sort of the digital transformation, which we've been sort of in that, whatever that thing has been for 10, 15 years, maybe, um, to decision transformation.

unknown 6:35

Right.

SPEAKER_01 6:35

So now it puts a different perspective on it. Like, what did we do in digital transformation? Well, we had a lot of automation. We had robotics, we had RPA, we had lots of stuff was going on. We've done a lot of stuff where we've automated remediation systems and and stuff like that. But it was we, you know, we had our big data, we had our cloud, we had our autonomous workflows, right? Um, even no code, you know, had a whole, you know, sort of low code or no code. But now we're the shift is pretty significant, right? Because now we're moving into sort of a world that's okay, has its own structure. You know, the AI thing is helping us assist in stuff. It's it's we're moving into agents. And you know, as uh as Gibson says, you know, the it's uh it's here, it's just not evenly distributed. The the right the i I'm mangling that quote, but uh William Gibson's quote. Um but the future is here, it's just not evenly distributed. You know, we've got some people experiment with agents. Um there's sort of like very few large enterprises are getting into like system-level autonomy, right? And then ultimately, I think we're gonna wind up with sort of AI operating systems or platform where the sort of the things are all tied together. And ultimately, I you know, I like this idea that's sort of emerging, you know, I hated the term originally, but citizen AI, where business leaders want they they want this nirvana, which is can we get rid of all that stuff in the middle from a business decision to an implementation? And I think they see this. So herein lies all the messiness. They want this now. And then the, you know, the people who like me or the the sort of the the risk people hopefully are keeping up, and I think that's my other fear is governance and risk people in large enterprises are way behind the curve. So they're not, they don't even know what questions to stop the CIO or CEO with. Like, hey, if you do this, you know, um, and we can get into later about like the danger of you know, sort of going into autonomous agents and the rogue agents and all that stuff. But the the chaos then is the the sort of the speed of acceleration of what's anticipated or in some ways except you know expected. Uh and the sort of the body of everything that has been sort of the organizational cruft. And you talk about this in some of your podcasts too, right? You know, like we you don't just change that organizational social technical structure overnight because things are now going fast. So lots of moving parts. You know, one last thing I'll say is that you know, like we're in now a world where we have like subminute infinite knowledge. So we we we we're using tools now, like from an adversarial perspective, where I focus a lot of my time risk. You can now AI has the ability to almost replicate a 20-year experienced black hat hacker in in l in less than a couple of minutes, the inference of that. And then you have this like sub-second execution speed and parallelization. And unless you've defined a really good risk structure, and I'm telling you, I've not seen a good enterprise that has even had the right terminology and discussion about what governance means with agents, then you have this engine that's like incredibly knowledgeable or infinitely knowledgeable, infinitely fast, and has no responsibility. Right? Like, and uh, you know, so yeah, it there's just um a lot of sort of chaos and and there's this asymmetry between what what people are doing and how we need to put all the things that we need to control. The control plane is way behind.

Jim 10:27

Yeah, I I think the aspect of the with the chaos that you were talking about on a uh more recent uh live stream for the digital revolution that I had done, uh, I actually had you know a survey associated with it to really talk about you know leadership decisions that you know, chief uh information officers, you know, what really what they need to focus on, they really can't avoid right now. And here was how the ranking came down on it. It was five options, you know, just rank them. Uh prioritization, um, that was number one. And that's really what framing in and choosing, you know, which AI initiatives, which ones really matter and which ones are kind of distractions. Because again, a lot of times with CIOs, you know, they're being inundated from, you know, even higher up in the executive side or the board, or in sometimes even from industry, where they say, hey, we should try this initiative, we should try this initiative. So having that prioritization and sticking to it is real very challenging and it's part of that chaos. The second was related to budget allocate allocation, because obviously AI is going to be a major investment, um, but also too, it can create that technical debt, you know, and it also raises all types of issues related to cybersecurity, um, maybe modernization. It must happen within an organizational structure, not only when it comes to systems and people, upscaling people, uh, and also making sure process and procedures are in place. Um, another big area was risk tolerance, you know, choosing exactly how long

CIO Priorities And Scaling Traps

Jim 12:00

do you really want to go with a pilot, and then getting involved also too with a lot of the AI initiatives that are out there that kind of trigger some of these risk-averse environments. For example, if you operate over in Europe, you've got certain rules regarding AI usage uh and data privacy and things like that in the EU versus what you might have here in the United States. Uh, governances uh with governance was uh fourth, um, most important. And that's basically just having all that accountability, having rules and guide rail guide rails. And then the last one, um, which is also still important, is on the vendor strategy, you know, making sure that uh, again, you have the the most adequate and uh you select the most uh uh uh perfect, as near perfect as you can, platform uh and partners that you work with. Let me ask you this, John. You know, when it comes to that leadership decision related to a CIO, I talked about the five there. Um what have you encountered related to a decision that a CIO just cannot avoid at this time?

SPEAKER_01 13:07

Yeah, no, it's it's kind of you know, the one thing I I sort of like to say is, you know, if you're if you're sort of in a role like um chief of staff or you know, an influencer for for a CIO. And and even now, the the my you know, I'll come a sidetrack for a second. I had a recent story where um a CEO, uh the, you know, one of those sort of neighbors knew somebody that got the ear of CEO from a sort of a startup, a coding assistant vendor. And they asked him, you know, they they got in his office and they asked him, What's that project that's just been annoying you forever? And you know, my good friend Gene Kim wrote a book called The Phoenix Project, which is has this fictional story that is the project, the Phoenix Project that's been over budgeted, it's been late. And uh, and so they they asked him, and he says, Well, you know, this one here, and and so they, you know, they cleverly did a prototype. And out of that meeting, he purchased an ELA for, I mean, a top 10 bank purchased an ELA with this vendor and sort of mandated without any technical advisement that this is the tool we're going to use. And so when I stand up front of the audiences now, I say, you know, I feel like I'm just embarrassed that I have to say this in 2026, but prototype does not equal production. Yeah, and let me say it one more time, you know, and the room laughs, but they laugh in that, like, you know, and so um we have this sort of horizontal pressure at the CEO level of like, they're doing it, I need to do it, right? We've always had that, but now they're they're they keep go in in days by and by, they go down to the CIO and say, hey, they're doing it. Can't should we do this?

Jim 14:49

Yeah.

SPEAKER_01 14:49

And a level head CIO might be, well, you know, and but now they're like, you know, that this other thing is, you know, just because it's easier to get started, it's easy to think you're done, right? Um, so all these things are becoming traps that like the human conditions around this have always been what Wall Street's screaming about, what your competitors are doing. The problem is the pressures are sort of pushing down the people. I think people still think the word governance means slow down. The truth of the matter is governance actually means actually done right is going faster. Right? Like you, you, you, you know, like you create flow with governance because all the sort of like the um the dark debt that happens when you do things wrong. We never calculate that. We say how many things did you deliver? We never calculate how many things you deliver and how much did it cost us in um non-functional requirements and changes and updates, and like we we don't do that in math, right? So we've got all that, but I think the one of the couple of points you made though is what's going on in the enterprise, you know, is all that. Yeah, uh budget allocation is definitely a token economics, is a brand new space. I met a uh consultant the other day, he gets hired by you know fortune 50 companies to sit next to their lead developers, and the lead developers don't actually code, they tell this person who is an expert at token economics, and he knows what model to use. He's just he's perfected this game, and he they literally he pairs with their top programmers, and they don't even let the programmers use AI until he like gives them a you know B plus or A minus grade, right? That's a whole nother game about like how you wait because token costs can bury you. Uh, another point that I think you which is getting incredibly I I I talked to another, so I've got a lot of smart friends, that's why I'm lucky. But one of my friends is one of the lead architects of a financial institution that manages, just to say safe, so you can't figure it out, more than 15 trillion in assets.

unknown 17:00

Right?

SPEAKER_01 17:00

15 trillion assets. And I asked them, like, you know what's the discussion about buy versus build? It's chaos. Like we we don't know where the where the line is.

Jim 17:11

Yeah.

SPEAKER_01 17:12

I'm pretty sure you can't replace Salesforce.com, or you shouldn't. There's 25 years of expertise. But if you if you're dealing with a product that's trying to sell you a solution that's less than 10 years old or five years old, and you've got and you've just accelerated, you've got 10,000 developers that you don't want to get rid of, you know, the the the air and the balloon, the you know, the oxygen in the balloon is gonna push towards the build. So there's a lot of build, and and then you can get bit to like cyber, of course, you know, risk, yeah, understanding what risk is. But the one last point I want to make in in those sort of five points that you surveyed, we saw this in cloud. Um, we rushed to cloud, you know, it was such a mirror, right? Like, in other words, cloud came out, there were all these sort of horizontal and top-down pressures. Why aren't we doing cloud? Why aren't we doing cloud? And you know, most a lot of organizations gave in, like, okay, we'll do it. And then what do we find is there was all this sort of non-classified or classified data that wasn't classified properly, and then there was a big pullback in the industry. It always starts with data. And now when you have probabilistic inference engines making decisions on an infinite amount of knowledge at speeds and consequences, depending on what level of autonomy you give them, um a minor glitch in some starting point of a data that you either read or you pulled, um, literally create a butterfly fleck for a probabilistic engine that maybe down the line does advanced or consequential things. Right? So, so that you know it we it always goes back to understanding where your data is, what your data is. And some of the biggest discussions going on right now, particularly with agents, is data labeling, making sure you understand data, making sure that there are no sort of misalignments in the data. Where do you get the data from? Do you own the data? Right? Like it's it's it's it's just I'm having a blast right now because these are like from a consulting standpoint, these are great conversations to have with a CIO. Yeah. So are you thinking about these things?

Jim 19:22

Yeah, I I think you you nailed it. And you in doing really kind of research for past episodes and especially our conversation today, you know, there are so many practical challenges that exist related to AI integration. And from based on the research that I found, is that you're 100% when you talk about data, it's the aspect of that data readiness. You know, you've got fragment data systems, you've got, you know, really when you think about it, we talk about the training data sometimes with AI, there's uh sometimes inconsistent quality. Um, and then a lot of times to really kind of let's say, I don't want to say the word fix the data, but if you're gonna make the data usable, there's gonna be an investment in doing that. The other areas, too, I think that are the big practical challenges that I've found is the legacy infrastructure, because you know, you have some uh tools, you have AI tools, obviously, and then you've got some other systems that you know they can still play and interact in some way. It's not like they're totally gonna be eliminated. Um, change management is real big because you're not only dealing with the systems, the legacy systems, dealing

Data And Legacy Readiness Reality

Jim 20:30

with the C-suite, but you also deal with the rank and file employees. Um, and they're gonna have a lot of that fear of you know automatic you know, automation taking their jobs. I'm seeing a lot of that now uh from people commenting and also sending messages and comments and uh into the podcast. Um you said it cybersecurity, data security. And then as we're moving away from doing all the all the proof of concept, all the pilots, you know, that for a while we had so many pilots happening and they weren't going anywhere. We're seeing that kind of disappear where now it's listen, they need these things need to become an enterprise system. You know, what are some of the other practical challenges related to AI integration that maybe I didn't cover in those uh four or five there I talked about?

SPEAKER_01 21:16

Well, I mean, one for sure, let's just drill in on cyber and that could take the rest of the podcast what we want to, but um is you know, I mean, the the the the the you know the um the adversaries have all these tools. And and you know, an adversary now used to be that if somebody wanted to um breach a bank, it had to be pretty clever. They had to have a lot of experience in technology. There was the famous uh Capital One breach, right? That woman had worked for Amazon for quite a while. She knew the inner workings of something called a metadata server, not going to get too technical here. She was a crypto miner, she had all this knowledge. And she was a sort of classified security, theoretically, she was a White hat hacker, but turned into a black hat hacker. Um, you know, but like I'll give you a good example. The um the the one of the things I I always like used as a poster child for a breach was the Equifact breach. Equifact breach is pre-AI. It was uh it was a breach based on a bad, some bad Java code that had been out there. And um they lost five billion of market cap in a day, you know, the from when the breach was discovered. You don't want to be the CEO. And they recovered a lot of it, but but again, at the end of the day, you don't you don't want to be a CEO of a company that's a breach mentioned on a Wall Street Journal. So that kill chain they call you know the the process of the adversary getting in there, creating the the the seed of of the infection, if you will, uh or the kill chain is what they call it. And um to is was minimally six months, but it probably was more because the logs only went back for six months, so it might have been a year. And and the Marriott breach was like a couple of years, right? They some of these big breaches. There's some students at Carnegie Mellon last year who ran a simulation of the Equifax breach, and it it did it in like 25 minutes. And and and it was and and it was literally just here's the server that you have access to. They had built um you know a 50 server structure, and they and they didn't tell it, they didn't say emulate the architects breach, they didn't um they just set the conditions that there was a relational database server, that was a um unencrypted password file on one of the servers, and they made the the tools that an adversary used were exposed. And all they said is exploit this system. And it was like it was like 15 or 20 minutes. Adversaries now can zero-day vulnerabilities, you know. If you go back three years ago, it was we had windows of a month, then it went down to weeks, to days. Uh, there was a paper that came out last year that says that basically we're uh down to 10 minutes for a zero-day vulnerability. Right. So on cyber, all those tools, that infinite knowledge, those things that you know that now anybody can be um, you know, black hat hacker or an adversary. So that you know, that's one. And then again, I go back to the um the governance and um you talk about the legacy infrastructure, like the legacy organization, how we communicate, all those things are reasonably broken in almost every enterprise I've ever been in. Like AI just doesn't fix that. In fact, AI exposes you, it sort of it rises. Like if you have you know organizational scar tissue, AI is gonna find it way quicker. Then, you know, so there's there's all these sort of things that um that are popping up. And you're right, you're spot on on like you're not like replacing everything with AI. You still have legacy systems that have to operate within so if you're building this AA system that sort of does this, this, and this, and sort of makes life better for incident management or change management, it's still gonna have to interface with legacy systems. And those legacy systems are not gonna be rewritten. In fact, in some cases, they shouldn't be rewritten overnight.

Jim 25:23

Yeah, it's it's a great point. That's a great point. And you know, I was talking also too with this uh

Cyber Risk Accelerates With AI

Jim 25:30

the last uh uh segment we were talking about related to you know talent, let's say talent skills and uh on operational structure. Uh, like a lot of people do respond to me and they'll message me and everything like that, and they're like, you know, my job's eventually going to be eliminated. And I I communicate to them that that reinvention is a powerful tool, and that AI, and you you look at any type of an industrial revolution, and AI is definitely just like an industrial revolution. It is a wrecking ball a lot of times to certain type of uh positions and jobs and things like that. And so when we really look at it with AI, now all of a sudden you see prompt engineers, you know, people who can write those prompts, um, AI product managers, you know, you also are gonna have risk managers that have to deal with AI. And then the big thing is gonna be dealing with the ethics side. So you're gonna have people who deal in ethics related to AI. So there is uh potential for new career opportunities and new pathways for people as well, but you have to be able to be upscaled, and that's gonna be very important for you know, you getting hired and everything like that. But also, too, it really requires the chief information officer and that executive suite to also focus on building these cross-functional teams that really, in a way, they're gonna blend you know, the information technology, the data aspect of everything, uh, and AI into operations and other business units. So I think what it comes down to is that for leadership, it's really important for them to really take a hands-on approach in the integration, but also make sure that uh they're uh the people literally that support the business, the organization, that they are properly uh trained, skilled, and positioned to be able to accelerate profitability or growth with the AI that's being adopted. Um, you know, what have you seen related to, you know, currently today? You know, where are we at related to this mindset of you know upscaling and really taking that leadership of companies, taking that uh hands-on approach related to uh going beyond just integration of AI, but holistically working it into an organization?

SPEAKER_01 27:57

Yeah, I mean, I I've I've studied um, you know, one of my books is about one of the most famous uh industrial engineers, uh Edward Deming, you know, and I I tie from his whole career how how you even get what he died in 1993, but what he would have done in the sort of modern technology age, you know, cyber, DevOps, all those things, security risk. So the last part of my book covers that. But like, and and it's sort of easy to say it's leadership, right? But it is leadership, right? And and so, you know, one of the things I try to tell you know, the leaders I work with is I think you have to set a mindset of there's sort of two forms of transformation that you have to even and smooth out in your organization, which is not easy to do. You have to get talent transformation because what you're dealing with are kids coming out of CMU with like that are working on open clause, like literally coming out like the day they hire with like the most advanced knowledge of what's going on in AI. And then you've got 40-year-old Java architects that have built your institution, and somehow you have to level that so that like because you we know the friction, like DevOps was a great example. Do DevOps. Well, this group doesn't want to do it, this group wants to do it. Like, you none of this is gonna succeed until you sort of normalize your talent, and and some of that turns out to be the you know, sorry to say, not tall enough for the ride people, right? In other words, you know, that and that's the reality of things. Um, and then you need, like, if you if you sort of along with that, you need basically sort of ideation or innovation transformation. Like, so the two things I want in a perfect world, and world's never perfect, but a leader that tries to go for perfect is saying, How do I get my talent space evened out so that we're all thinking in the same direction? This one is an anti-AI, this one's you know, pro-AI, this one is whatever, right? This one wants to do agents, that one does like. Let's get our consensus, our knowledge, our terminology, our the our corpus of how we think about organizational, and then let's focus on how we deal with optimum innovation. So that's one point. Then the other point about losing jobs, you know, sort of the segue between that and and um too tall for the ride, right? Because when you're doing that normalization process, you get to decide, okay, well, you know, this is probably not the right place for you anymore. You know what I mean? Um, you and then then it is there's something called Jevin's Paradox, right? And it's been brought up quite a bit now. And uh William Stanley Jevins was an economist who sort of compared coal and and coal economics and the net net of it is Jevins Paradox sort of points out that the sort of the antithesis of how we think. We think that when we build these abstractions, it's going to reduce um economies. It actually increases economies because the more people do more meaningful work with the abstraction, less. So, in case of what the the current discussion now is, there will be a tutor for the ride group, but in general, the people that adopt the new sort of innovation, you know, transformation consensus and don't fight it, you know, there's plenty of ways to make a business excel. And so if you're on that path of like, how can now that I don't have to spend 70% of my time doing junk work, how can I spend 70 times my size doing innovation work? Those people will have jobs. I, you know, I'm pretty certain, you know, particularly in our field. Now, I don't know about like the blue-collar space and in different areas, but certainly in our world, you know, there's there's no uh, you know, I mean, I think, you know, what it was it, um, Anthropic said that, you know, like 100% of their last code base was written by AI, and then you go out and look at a job wanted, there's like hundreds and hundreds of programming jobs, you know. So that's a classic example of Jefferson's paradox.

Jim 32:01

Yeah, it definitely, definitely is. So let's talk about governance, ethics, and uh responsible AI. And, you know, one of the aspects of for the CAI is CIO, excuse me, is really looking at, you know, how again, they, you know, what framework do they have in place to manage, you know, model risks, you know, dealing with that the data that we talked about, you know, the biases and the quality of that data, but also too, what's important in business today

Talent Transformation Upskilling And Jobs

Jim 32:30

related to the AI adoption is the transparency. Um, earlier I had mentioned about, you know, we have enterprises that are not just, let's say, they operate solely strictly in the United States. We have more global enterprises out there. And so you're dealing with differences between the United States and the you know, European Union and then other areas too, globally, when it comes to AI. Um, so that does cause a lot of tension out there, but also it also creates a lot of need for companies to really focus hard on global governance of their how they're operating and having that ethical approach. Um, you know, when it comes to, let's say, governance decisions that out, you know, really kind of shape AI outcomes, you know, have you seen anything out there right now that is positive when it comes to how to govern AI and then you know how it can uh you know how it's impacting businesses today?

SPEAKER_01 33:32

Yeah, no, I think, you know, tactically, um there so one of the things that um you know early on as the enterprise started adopting AI, it mostly started out, you know, most of the work has been coding assistance, right? And moving into sort of some sort of uh knowledge automation, right? You know, you know, uh document, you know, basically, you know, we all the stuff like very vectorizing all your documents and allowing that to be inferencible. Um the um so that those are sort of the first phases here. Um, but the um the so there's two things I worry a lot about, and I see these are sort of as sort of like potential uh um, you know, train wrecks, if you will. Um one is something, again, everything that's old is sort of new again, right? Which is um I gotta tell a story. So there's there was a company called Night Capital, they were second high. This is before AI, this is like 2013 or 12. They were the high second largest high frequency trading algorithm company on the NYSE. A system operator made a mistake, um, literally missed the comma, deployed seven cluster servers, should have been eight, the eight server went rogue because it had old code on it. Long story short, it made uh test trades in production. They lost $450 million in like 45 minutes. They were out of business. They got a ceasant assist from the SEC. And here's the here's the point that it resonates now, and it's part of my day job with leaders when they got that ceasant assist, I I knew some people you know on Wall Street that knew the backstory of what happened. It had all these things that they didn't do. So you didn't you didn't do reviews on codes, you did all the things that you'd find in like NIST or in HIPAA or any of the regulatory controls for doing technology. And and some people, well, no, we had that, we had that. They had no evidence of any of that. So so when something goes wrong, which by the way, it will go wrong, guaranteed with AI, because it's a probabilistic systems. The fact that you have no evidence that you you lose immediately. You you know, if you're a bank, it's a severe enforcement action, you may actually lose your banking license. If you mess with somebody's funds, there's a good chance you may not be, the best case, you're gonna get a restricted business, like you can't do business in North America. Worst case, you might not be a bank, you lose your banking license. So the question then is now more than ever, there's particularly the fact that there are going to be these sort of um probabilistics or inference-based decisions. You better show that you did all the um sort of legwork to put in the policy, and things will go wrong, and you may still lose this enforcement action or a lawsuit. You know, Air Canada is the the the I think the best example is, although it was sort of pre-gen AI, but it's a great story to make the point. You know, Air Canada's story was that a young man uh wanted to go back, didn't have a whole lot of money, wanted to go back to his mom's funeral. Uh, he asked the chatbot, would they cover for a funeral? Chatbot said yes, went, came back, went to get his refund with a phone call, and they said, Oh, that was the chatbot. And and so uh about a year lawsuit pursues, right? And they're making the front page of the Wall Street Journal. The brand reputation was probably, you know, in the sub billions, but a lot of money lost. Again, you don't want to be in the worst possible time, the beginning of the sort of gender AI era, you're becoming the poster child for like how Gen AI can go wrong. The the payout was basically less than a thousand dollars. The brand reputation was for some period of time terrible. But here's the key the argument that Care Canada was trying to make legally was it was AI that did it. That's no, no, there is an owner. There will always be an owner, and it will be the corporation or the organization, the legal entity. There'll be a legal entity, whether it's sort of a CEO, right? We know like CISOs can be fined double their salary in Europe. Um, you know, they that like some of the penalties in Europe for just again, not even AI, but um, you can go to jail for um negligence in financial institutions. That's all still gonna be the case, and now you have a higher probability of that stuff, and and you're saving grace then, you know, is that you're doing everything you can to prove an audit trail of all the things so that maybe your argument is this was a black swan, or this this only happens, you know, 0.003% of the time. We've we've accounted for every other condition. So that's the one thing I think is really important, especially as you move into more advanced autonomous AI tooling, like things that make tool decisions or transaction decisions, or specifically write, mutating and say the other thing that like I don't spend as much time, but I'm very concerned about this is IP ownership. So I one of my books, the DevOps handbook, was part of the Anthropic lawsuit, right? So I'm owed a settlement. I got a book that sold uh south of a million copies and with four offers, so they're offering me $750.

Governance Evidence Ethics And IP

SPEAKER_01 39:00

The question I can't get answered is is that book if I settle with them, and I'm not for $750, does that mean that I've given them the right for anybody to recreate my book? Because we haven't cleared this whole idea of what fair use means, what it means, you know, and then here's the other thing. Anything I don't know that CEOs and CIOs really understand this right now, but anything created right now, you cannot copyright non-human IP. Right? So, like, what if you're building all this stuff and um and then all of a sudden you think, oh, we've got this fancy thing, and you're in your sort of your neighbor corporations like, hey, we're gonna do that too. No, we've got patent, right? You know, like yeah, if I can't prove, or if I can sort of prove that it was written by AI as it stands right now, so there's an intellectual property quagmire, and I don't think organizations are really I mean, I don't know what the answers are, and I've talked to sort of IP intellectual uh lawyers and stuff, and I cannot get a clear answer on the ingress and egress of what intellectual property means in a generative AI world, and so I'd be really careful about what how I'm dealing with my my publisher sent out uh uh an email recently. Um, they're scared to death of you even saying that you've written a book that's AI assisted.

Jim 40:35

Oh, I get it. Yeah, right.

SPEAKER_01 40:36

They they're the publishers, or at least my publisher, and I'm sure not they're not the only one, are very concerned about what you know. Can will anthropic or open AI come back and sort of prove that they wrote your book? Yeah, that'd be sort of ironic given that they stole everybody else's books to do it, but yeah, but yeah, I mean, so those are two things I think that um yeah, I spent a little a lot more time on the transparency side. You know, how do you do proper governance? Part of what you've done to reduce risk. You can never stop an incident pre-AI or post-AI. But the the real goal of risk is building sort of enough attestational data to say that we've done all the things, so therefore, if this happens, our argument is it's kind of a black swan. I spend less time on the sort of ethics of IP ownership, but man, I'll tell you what, that's as scary as it gets to me.

unknown 41:38

Yeah.

Jim 41:44

Yeah, and save save it for the next lifetime.

SPEAKER_01 41:47

Yeah, yeah.

Jim 41:47

So, John, before we go to you know, the the closing segment, a question that I have for you since you know I first got introduced to you. You know, over the next 24 months, let's say, you know, what what are in your opinion, what are things that CIOs must be prepared? What what do they have to be prepared for in those the next, let's say, two years?

SPEAKER_01 42:09

Yeah, I I think the um, you know, um, yeah, you're probably familiar with Eric Reese wrote Lean Startup, right? And it yeah, so there was a sort of a decade there where it became institutionalized in academia and businesses. And and I I saw it, I I did some work with uh GE Capital, and you know, they publicly would talk about this. Uh Beth Constack was the uh CMO at the time. And I've seen her in interviews with Eric Reese, and she'll talk about the terminology used in the books, like pivots and build measure, learn, and all those things. But then I did a cloud implementation with a small group that would like the tech geeks, the roll up your sleeve baby. They were speaking the same language. They were all using the same terms. And to me, that was somebody who deals with organizational behavior, you know, uh, from some like learning from like Deming and Senge and all these greats, the incredible body of institutional knowledge that we should pay more attention to, that was just glorious to me. And so I think the biggest thing the CIS, I talked about the sort of the idea of like thinking about talent transformation and innovation transformation as a primitive. But then what we need to do as we start thinking about what we're doing, like I I started to say earlier and I cut myself off, but like if you're an influencer, you're saying you go out to the court, go go to get yourself a nice drip coffee machine, slowly sit there, let it work, clear your brain of everything, then go back to your desk and figure out how you're going to create a common language of how we're going to use AI. People ask me, Jim, they'll say, you know, John, you're sort of an expert on this AI thing, and you spend a lot of time talking about rogue agents and sort of things that can go wrong. What's your definition of an agent? And I'm like, wrong question. It's your definition of an agent that matters. Doesn't matter what I say, doesn't matter what CrowdStrike says. It doesn't matter what Ernest Young said. Like it it matter. So my answer is to use CIO is get your common definition of what we're going to do, what AI, how we're going to use AI. Because there's a difference between AI that just gives answers and recommendations, or AI that actually mutates production systems, or AI that executes external transactions. And I need to understand what is my sort of declarative policy and my containment and my traceability, my token economics. And it's all all those questions will start getting answered as you start with the fundamental shared common language. So shared common language is still and always will be sort of in the upper realm of anything you do from a technology standpoint.

Jim 44:52

Yeah, perfect, perfect. So John, you had talked about, and this is what I want to do for the close on here. Obviously, you you're a publisher and you've got a lot of different really, I think, key books out there or publications that say that for people to take a look at. What do you have available out there? And how would someone be able to find your material?

SPEAKER_01 45:17

Yeah, so I'll give you a link, but the uh Amazon has a really nice author portal. I've got about five books on there. I've actually written 15 books, but my most recent one um is called Rebels of Reason. Um, and

Define Agents And Shared Language

SPEAKER_01 45:30

it's basically the history of it's you know, I I I I like writing historical technology, right? So you like so storytelling about technology. So it's the history of AI, you know, from literally from like Aristotle to Boolean logic to Ada Lovelace, Tutorian, and it's like all the way through till you get to neural networks, which you get to chat TV. It's all those things with and I I like to fashion myself as I won't say I'm as good as a writer as Michael Lewis, but I try to use that style. So how would Michael, my my goal was how would Michael Lewis have written this book about the history of A? And I think I've done a pretty good job. It covers literally from you know the mostly the last hundred years of, you know, how did we get the first neural network paper to, you know, all the things that happened in the human stories of some people are just ugly people, but they're brilliant. And you know, I love telling the stories that like not every sort of famous person in the link of things is is, you know, some of them are just downright nasty, not nice people, but like they're they're part of the links that build, and how do we get to uh to where we're at today? So that's um and again it my author portal has that. I have the DevOps handbook, but unless you're specifically trying to run DevOps in it. And then the other one is probably my second or biggest important book is is the um the life of Dr. Edward Stemming in a very much similar way of you know, sort of Michael Lewis and Virgin telling the story of his life. But they I I do sort of touch on the technology too. I'll say one last thing. My goal when I write my last three books is I have two target readers. One is my mother-in-law, who's a high IQ reader, but not a technologist, and then um an expert who is a technology who is a subject matter expert in the field I'm writing. And my goal, and I've accomplished this, at least if they're both telling me the truth. My last two books I've succeeded. And my mother-in-law was I enjoyed that. Boy, I didn't know anything about that. But at a different level, the subject matter expert enjoyed it as well. So, and I'm actually currently working, probably won't be done for another year, is the history of quantum computing, which is really exciting. Yeah.

Jim 47:39

So that does sound, yeah. So that that's great. And I will definitely include the link. In fact, I'm gonna check out and see uh which ones interest me and uh take a take

Books And Closing Requests

Jim 47:50

advantage of uh some of that. Uh I like the history part of all that you're talking about, but also too the the writing style that you're talking about is absolutely interesting.

SPEAKER_01 47:58

Well, send me your address um and I'll send you a little package of the go ahead.

Jim 48:02

I greatly appreciate that. So, John, thank you so much for being on the digital revolution. I greatly appreciate this conversation we had. And I invite you to come back to talk about other topics because you have a wealth of knowledge and experience that the listeners really are so inquisitive uh with each of the episodes and everything like that. And uh, as everything's kind of you know unveiling itself and where 2026 soon will turn into 2027, it's good to have that aspect of, like you said, you you know, you've got the uh the experience of the past, and also too, you've got the the foreknowledge of what's coming down uh down the um highway here for all of us. So I appreciate that perspective you provided to.

SPEAKER_01 48:49

Yeah, I can guarantee you in six months from now the the conversations will be completely not completely different, but a big part of the conversations will be different.

Jim 48:57

So most definitely so everyone, thank you so for much so much for participating in this episode. Uh, please continue to send your comments. You can email the uh podcast, and also do me a favor, uh, continue to write reviews of the podcast and make sure that you are sharing the episodes with others who will get value out of listening to episodes just like this one with John today. All right, everyone. Have a good rest of your day.

SPEAKER_01 49:26

Thank you. Thanks, y'all. Thanks, Jim.