Deloitte CTO on the AI Investment Trap:
CIO Advisory 2026

Bill Briggs: The single most effective and tactical session we do is not on multi-agent systems or ops or infrastructure, it's on storytelling.

Michael Krigsman: Companies are caught in an AI investment trap, spending too much in the wrong places with too little benefit. It's a mess.

The 93/7 spending imbalance

Bill Briggs: The findings from Tech Trends, we've done our Tech Trends Report for 17 years, so we published in December. The best stat coming out of it is that 93% of all AI spend is going toward the tech and the tooling, and only 7% is on everything else, which is the culture, the change, the learning, the how do we communicate the vision, how do we be thoughtful about trying to redesign or reimagine how we work and how we serve our markets and our customers.

And that is well out of balance. And we wonder why you see headlines of pilots that don't reach value at scale, we see proofs of concepts that never go anywhere, and we think it's a straight correlation of not investing in all the things that are actually required to embed AI into how work is done differently, how markets are served differently and the like.

Michael Krigsman: I'm a little bit confused because AI is a technology. We're talking about technology-driven change, and so why shouldn't we spend 93% on the technology that's the engine?

Bill Briggs: The spend on the tech is still also heavily biased towards the new and not enough investment in the foundation elements of data readiness and core monetization systems that we need to go in and make ready to participate in AI and agentic workloads. So the overall 93, there's an issue with where the disposition of the spend is.

But to your question, any transformation effort is so much about how do we actually understand the problems we're solving, the outcomes we can strive for, how do we get our people to buy in and engage to be able to help us really focus it in the right places, and then the work itself is gonna fundamentally change.

And historically that's been, you know, 93/7 on an ERP implementation would be the wrong balance because so much of making it actually work is about investing in the softer elements. In AI it's even more pronounced, and the reason is, historically tech has been forced on many businesses and business users and customers in unnatural ways. And so the business processes have evolved to be a reflection of the way the tech works.

And we're in a moment in time where if you apply AI into an inefficient process, you apply AI on an overly complex process, business process, you're gonna weaponize inefficiency and actually probably pay a lot. We'll talk about the cost and infrastructure dynamic I'm sure is a big theme of the Trends Report.

So this moment in time, it's harder than it ever was 'cause before we had the blueprint, the roadmap, which is basically here's the steps that we used to do, and can we do them incrementally better with technology? Now we're saying, hey, we've got to fundamentally potentially blow up all the steps that we used to think had to be how we work, and then apply AI in increasingly intelligent ways to focus on the same outcome. But hopefully many of the steps we had to do before aren't required anymore.

But to get to that level of focus, to actually get to the level of the organization embracing those changes, putting them in action, and then investing to continuously improve them, that's the stuff that's all in the investment outside of the core tech bucket, which is the place that's so sparse right now.

Tech for its own sake has never won

Michael Krigsman: You're CTO, you're a technologist, and so why do you have such a strong focus on this 7% that's outside the core 93% of the technology?

Bill Briggs: You're saying you'd expect me to be arguing it should be 99/1 the other way because I'm a tech? No. I'm a technologist who loves technology because of the potential it represents, but tech for its own sake has never been a winning hand. Putting any one tech as the hero of the story and assuming it's gonna win the day is also a fail. It's a flailing and failed strategy.

So yeah, I certainly think we need to continue to invest in the tech, but if we're not doing it the right way and it doesn't actually change the way we serve markets, doesn't change the way we serve our clients or customers or citizens, it doesn't matter.

And I think we're stuck in this... The other piece I see all the time, Michael, is the feeling that the world's moving so quickly and technology's changing so rapidly and the almost paralysis of, I can't make a decision now because next week a headline will come and I'll have buyer's remorse, or we don't have our strategy figured out to perfection and so I don't want to get started, I don't want to architect myself in a corner, I don't want to make a move I'm gonna regret. And that's also a dangerous trap, which we can pull that thread if you want.

Michael Krigsman: How pervasive is this? I mean, you're, you know, obviously 93%, but is... Are you seeing this in many, many, many different organizations, this same pattern?

Bill Briggs: Different industries around the globe. And we released the State of AI in the Enterprise survey a few weeks ago, and it had a similar bit of finding of, you know, we're at the place where 30% or less have agentic pilots that have reached production at scale. But there's an overarching confidence, there's potential there. There's opportunity there. Like 70-plus percent think that there's something here, and we need to invest more.

And so this kind of awkward adolescent stage of I would argue that tech... We're not waiting for the tech to mature anymore, and the fact that it's gonna continue to advance is a feature, not a bug. But the fact that we're still learning how to treat this... And it's a different kind of tech investment than we've had.

It's, you know, historically you might have a CFO say, "I don't wanna release any capital unless I have complete confidence in all the use cases and all of the potential return." And that's been a trap in its own right.

We just released at MWC last week the enterprise AI navigator, which is Deloitte investing the last year to take all of our industry sector functional knowledge and embedding it in a tool that gives not just here's the places we think there's the most value if you applied AI, if you agentified your processes, but actually ties it to financial and operational metrics in a way that we can give the CFO much more confidence without it being completely limited to, "We're only gonna do these 3 things." It's more of a, "Here's a portfolio of what you can go and the value you hopefully can reap." So a lot going on. It's exciting times.

Treating AI like a new coworker

Michael Krigsman: Folks who are watching, you can ask your questions right now if you're on Twitter, X, use the hashtag CXOTalk. If you're watching on LinkedIn, pop your question into the chat. I urge you, take advantage of this opportunity. When else will you have the chance to ask Bill Briggs, who's Deloitte's CTO, pretty much whatever you want? So I urge you, take advantage of this great opportunity.

So Bill, you're really describing a misallocation of investment that is pervasive. Is that too strong a term to use, misallocation?

Bill Briggs: It's where we are in the movie. And there's this increasing appreciation of as we deploy AI, it's gonna help us rethink the work that our people are doing, which means the change management journey that's been a part of a lot of tech implementations of the past, it's less about training them to be able to repeat the tasks to use the tool, and more about training them to understand how the tech is something they can be using to light up their own use cases. And so it's this interesting different dynamic.

And as AI, and we have an emphasis on physical robotics, as we're seeing more intelligence in the technology, there's this interesting dynamic of some organizations are treating the deployment more like here's the introduction of a coworker versus this is a piece of tech that you've been given access to and rights to use. And that's something new and different.

You know, we didn't have SCP screens being named and being listed on an org chart. You're seeing that happening. In robotics one of the health systems that's been deploying wheeled robots with a humanoid form factor that basically move lab supplies and hospital supplies to where they're needed on the floor, they actually have naming competitions. You know, the head RN is the one that introduces them to the shift, and then they have naming competitions, to make it just feel like more of the team.

My favorite one was Sir Bots-a-Lot. I'm a child of the '80s and the '90s, so that makes me happy on many levels. But yeah, it is a different dynamic.

AI itself means nothing without culture change

Michael Krigsman: We have a really interesting question from Arsalan Khan on Twitter, and Arsalan is great. He's a regular listener, and he's brave enough to be first.

Bill Briggs: The first one. Thank you, Arsalan. That's awesome.

Michael Krigsman: Yes, yes. Arsalan, thank you. Okay, so here's his question, his comment. He says, "Everyone wants to change their business through AI, but no one wants to change their culture to achieve that. And AI alone is not the magic pill." So maybe Bill, can you address this?

Bill Briggs: A direct through line. AI is an incredibly powerful technology, and it's a dangling modifier. I'm an MBA, a computer engineer with an MBA, and my wife's an English major. She's at preschool right now teaching, so she's rubbed off on me well I think. AI itself means nothing. If it's treated like AI is the product, that's a great indicator to me that the organization has the wrong strategy. So it's embedding into how work is done.

And like every technology before it, it's only as good as the team that's using it to deliver whatever the outcomes are in... And we're in a phase where because it still has a lot of market excitement within AI is prominently being featured as a part of a product set, you're seeing it as the headline more often than I think we will in a year or 2, where it's just assumed to be a part of the underpinnings of every technology solution.

I got to kick off the CES research thread this year for the second year in a row, and we used our tech trends to kind of frame everything that was gonna be seen in Vegas. As I walked around the 300,000 exhibitors, I didn't see a single one with a sign that said, "Now with electricity." You know? It's gotten to be just kind of assumed that technology will require electrons flowing through them. But everyone had AI and agent as something to act like that was a differentiator.

It's not, you know, by itself, and a lot of times it's an empty feature that doesn't really matter. So how do we shift the mindset to, "No, we're gonna redesign work. We're gonna think about embedding AI into the way our products work in the market," and that's the point. And the fact that there's a model behind it, that there's an agent behind it, yeah, is assumed, but not sufficient for distinction. How about that?

Redesigning work from first principles

Michael Krigsman: This point about redesigning work, this has been the core issue of enterprise technology. I'll date myself and say I remember the mid-late '90s with SAP, although you brought up SAP before I did.

Bill Briggs: It was a great partner of ours and, you know, it wasn't a knock on ERP. It was just the fact that that whole phase had a lot of tech and workflow that, you know, we got much more elegant as we've gone along the way.

Michael Krigsman: The point is that the redesigning work was fundamental back then. If you think, what was the value of enterprise systems even back then? It was the processes and the high quality processes embedded in the software. So you're really talking about a set of human nature issues that's now causing organizations to invest in the wrong place today.

Bill Briggs: It's still a big piece of our business and a key part of our clients' tech roadmaps of package-based technologies, SAS-based technologies, and the power of saying there is regulated industries and processes where we want the power of standardized, proven, global, be able to handle high transaction volumes and scale and complexity, and the value proposition of that piece of the stack is still very high.

And we've embedded over the years a lot of business logic, a lot of the customer experience, a lot of the things that make an organization distinctive that aren't a part of the core back office treasury, finance, HR function. Those have been bundled inside of the legacy systems they built, the ERP implementations that they've done, the SAS that they've customized, and that's where a lot of the reasoning and that's where a lot of what makes every company special is in that layer that's been forced into the technology stack in a way.

And now's the time to say, you know, how do we actually take a hard look at our end to end and decide which ones are gonna be... And for some it's gonna be efficiency gains, so applying agents on top of a finance function's ERP stack, there's ways we can take what had to require human intervention and a lot of manual tasks, we can do it with AI in a way that it elevates what a financial analyst is doing, building on what has come before, but replacing core features of what the work is.

And that's the piece that when we get organizations that are bold, you start with a lens for reimagining, you know, before you're thinking about tech and automation and AI and agentification, you're thinking about how do we simplify based on first principles, which is the outcome, not the imagined constraints that we've gotta do all 10 steps the way we've always done it for invoice management. A few of those steps are regulatory mandated, and we're sure as hell gonna follow those, but it doesn't mean everything the way we've always done it has to be how we realize the AI potential.

Michael Krigsman: I just want to say for everybody that if you see Bill moving around, just before we went live his internet service provider went down.

Bill Briggs: Mm-hmm.

Michael Krigsman: So Bill is on his cell phone-

Bill Briggs: Yeah, trying to be casual, and jazz, not big band, and the show must go on, so we're... And I've got a sleepy puppy next to me who is not very interested in my conversation.

Michael Krigsman: Somehow she's sleeping through all of this.

Bill Briggs: But yeah, Michael. We'll... maybe was the denial of the service from another CXO podcast was trying to shut us down before we could do our sixth time together. I don't know.

Quantifying AI financial risk

Michael Krigsman: This is from David Batz on LinkedIn, and David says, "How do you manage or predict AI financial risk in a quantitative way? Do you have specific KPIs, for example, that you follow?"

Bill Briggs: Twofold on it. I mentioned the Navigator work we've been doing. That's just embedding all the proprietary knowledge we have of business process and value gain in basically 2 big buckets of operational financial metrics, so we can say, "Hey, if you're a retailer, we think these are the places in merchandising, in warehouse management, in promotions that we can have material impact and savings." So that's kind of the upside of how do we try to quantify what good looks like and benchmark it with a very narrow vertical lens.

The other piece is, we've got a great study that was published maybe 2 months ago on just token economics and AI, the finance and cost implications, big picture, but total TCO, which is a huge strategic issue right now. And so you gotta balance the, all right, here's what we know about where we could deploy and what the value benefit could be with scale.

And for our clients of the Fortune 100, the biggest governments in the world, you know, that could be a material number. But then that feeds into, depending on the choices you make on architecture, on infrastructure, as the good news, have you found a use case and it's scaling. So it defies the headlines as creating value and impact, but the usage is growing.

One, we know that all the AI models in framing, they're all going down in overall cost or performance to cost, that curve is bending in a positive way. But as the volume multiplies because you found a use case and the scaling is scaling with the pace of operations, and that scale is tremendous, then it becomes an issue of, okay, we started on cloud because we wanted speed to stand up new capability. We wanted the ability with elasticity and scale.

That's a great story, but also if the cost is just a runaway curve, at some point, when do we invest in dedicated hardware to run that use case in that stack that we actually control? So the... You can remember, Michael, 10 years ago, it seemed like a complete no-brainer of infrastructure's a commodity, we're gonna move to operating expense wherever we can, cloud everywhere, and it's not... Just like then it wasn't mainframe bad, cloud good, it was always more nuanced than that.

The story is emphatically not cloud bad, dedicated high performance compute and AI chipsets that you own is always the right answer. But you've gotta be balancing that, especially as the use cases that you're investing in gain traction and the usage starts to grow. So we've got the ability to help clients get very precise on not just the upfront investment costs, but modeling the performance, and depending on those choices, what you should expect from the cost side of the curve.

Inference costs and shadow IT

Michael Krigsman: How significant is this issue of unit costs collapsing, but the total bill for enterprise AI rising and just the growth of inference, are organizations even tracking inference as its own budget line?

Bill Briggs: They're starting to more and more, and any seasoned CIO likely lived through the sticker shock of your first virtualization bill, and then the sticker shock of your first cloud as you started moving material workloads to the cloud, it didn't have thresholds and monitoring and governance and cloud operations in place. We have the same thing now of AI ops and agent ops.

So the mature organizations are doing individual keys per developer. They've got thresholds and control points in place, so they basically are modeling the usage behavior they're expecting. And so the idea that you could have $100 a month bill one month and then a 500K dollar bill the next month won't happen. But that requires some maturity and rigor that not every organization's doing.

And the flip side of the promise of AI, a lot of vendors are going directly to the head of marketing or the head of supply chain or the head of ops, and sometimes the CIO, the CTO aren't involved in those decisions or the architecture or the deployment, and that's typically where there's a lot of surprise, and quickly coming back to the fold of the tech organization to say, "Hey, we love what we were able to do quickly, but we realize we've gotta handle it in a much more mature way."

The other piece of those one-off, you know, the idea of shadow IT is coming out the shadow 'cause there's just a ton of excitement and a ton of investment in AI across the entire business.

The ceiling is low because AI by itself typically runs into the wall of it's only as good as the data that's ready, and we haven't invested in the foundational elements of our data supply. Or we had this promise, the pitch was about multi-agentic finance, but our individual finance systems and transactions weren't wrapped in APIs and services to be able to be invoked.

And so very quickly we're kind of brought back down to earth of what we can do, and it's really narrow and it's really incremental. And so all of those kinda lead to we've gotta have a more holistic approach to it. We've gotta have some governance and rigor put in place up front.

And unfortunately, and sorry for everyone from the business, CEOs and boards that are listening in, the easy button idea that AI and now agents was gonna make all of that hard work suddenly not be required, it's the opposite. It's actually more important than it's ever been.

Now, we can use AI to apply to core modernization, legacy system renewal, and the data cleansing and, you know, the fact that the models can actually understand the relative cleanliness of the data and then use that to their advantage. That's all true and real, but it doesn't mean it's gone away.

The trust gap from C-suite to frontline

Michael Krigsman: If CIOs or organizations realize that they have been misallocating their AI investments for 2 years, how do they fix it without losing momentum? And Arsalan Khan comes back, and he wants to know, what is that optimal balance, the 93/7 that you were talking about?

Bill Briggs: I don't have a number. I host a podcast with our chief human officer, Simona Spelman, and she likes to say that every dollar in tech should be 8 or 9 that's going into all the change and all the culture and all the learning, which would almost flip the numbers. I think it's probably somewhere in between.

But the things that we're seeing when it's being done well, there's a bit of a senior level directive, and I love it when it's the CEO, but it could be a line of business leader, or it could be a functional leader that's basically declaring the why upfront. And the why of first principle outcome based, we need to get everyone on board to help us understand where there's knuckleheadery and the most opportunity.

Here's another study we did, though. It was basically trust in AI, and it had an org chart anchoring to it. So you can picture an org chart where the C-suite is the top of the pyramid, and the C-suite over all industries around the globe, the trust in AI from the C-suite was 70%, 7-0.

As you went down every layer of the org chart, it was like a logarithmic scale halving the trust every hop away until you got to the front line entry-level worker, and the trust was 6.7%. So from 70 to 6.7.

And if you believe like I do that the frontline worker who's literally in the heart of the business, where business is actually happening, has the best intuition of the things that don't work or are unnecessarily complex, we're completely upside down.

So how do we invest to get that worker to be the one raising their hand with ideas, not using their own personal AI tools to get their job done, but doing it in secret like it's a bad thing, which happens a lot. How do we get them to actually lean in and... And it's obvious why.

I mean, and I didn't talk to the participants of the survey, but, you know, if you had the headlines that AI is gonna lead to job removal and destruction, then it would make sense that the folks in the more junior ends of the career would have a lot of fear and mistrust on what this technology might mean to them, and basically put up walls and barriers, which is gonna make it really hard for any enterprise to get the real value out of the investment.

Governing a workforce that isn't human

Michael Krigsman: You've raised the specter now of jobs. AI agents are multiplying inside enterprises faster than almost anybody can track, and many organizations do not have sufficient governance frameworks for a workforce that isn't human.

Bill Briggs: We're in the early, and this is one where there's actual debate happening in real time, and having it with a lot of my clients and even a lot of my colleagues. You know, the way I try to think about it in... and this is my own simple terms, so this is the computer engineer with an MBA, not a trained anthropologist or human capital professional.

But I like to say we need to look at what we can learn from everything we've put in place for the HR life cycle over the last 100-plus years of corporate industrial work development. And it doesn't mean that every step that we have to do for our human workforce, we have to have an equivalent for increasingly advanced technology.

But it's a helpful, for me, it's a helpful framework of, okay, what do we have to do with onboarding people? Well, we have to decide, one, as we recruit and hire them, their qualifications and their... what we're gonna trust them to do and how they fit into the organization, and then when they get here, we have to give them a user ID and access to systems, and who's their manager, and expectation frameworks, and how will we... A lot of that is similar, the things we have to put in place for especially advanced agents.

How do we think about performance management? This is probably the most pass the hookah piece of the discussion here, Michael. It's a little early for me, but I... yeah, maybe for you.

The idea of how do we... If there's an issue, an agent makes a mistake, one, who's accountable? That's interesting. But probably more interesting is how do we treat that mistake? And is it a trouble ticket like we would deal with our traditional systems and technology? Is it a training issue of the humans that train the agent, train the model, where we need to make them better or fix a problem in our guardrails or our testing suite? Is it a training issue for the agent itself? There's insufficient training, and it made the wrong conclusion, and so how do we tune the fine-tuning, the inference?

Or, and here's the dot, dot, dot that gets really interesting, you know, is it a disciplinary action? Because there's some malintent, you know, that... And we haven't seen a lot of this yet, but you can see a world where, especially agents that are finding ways to learn and grow, that... And if that's the case, what would we do about it? What's a disciplinary action for an AI agent?

You know, is it, "I'm gonna put you in the relative penalty box in Kill Dash 9, until you see it my way," or... And that's the... So the actual mechanics of what we do aren't quite, but the idea of, hey, we need to start thinking about it in these terms, and what will we put in place now policy-wise as we think about our culture and training and learning for the human counterparts, and then make sure we've got enough of a control mechanism in place, because once it scales and once we see the proliferation of these, it's gonna be really hard to rein it back or put it in place after the fact.

So we just hit a lot of beats there. I'll take a breath and a sip of water but

Michael Krigsman: I recently had a conversation with Claude, and Claude developed an attitude, and- I'm not joking, and-

Bill Briggs: Yeah. Yeah, answers became progressively worse.

Michael Krigsman: So I said to Claude, "You have an attitude." Yeah. And we then shifted the conversation into a detailed analysis of how this attitude came about, why it came about. Was it caused by training data? Was it simply emergent behav... Anyway, so that's another dimension of this, is when agent quality deteriorates for what appears to us to be subjective reasons, like with a person.

Bill Briggs: That's why I think it's helpful, any one of us that has managed teams has that experience where you have to navigate through and, you know, there's some precedent, and there's some policy, and but that's gonna be an increasing part of the reality.

You know, we're used to technology that didn't come in tired on Monday morning, and part of the promise of AI is that it's gonna work around the clock, and all that's true and real. But, you know, as we move into personality or at least the mimicking of personality, in ways that could have, you know, I think this idea of performance management and agent ops. You know, agent ops is typically thought of as a much more technical thing.

We saw one client that merged their HR function and their tech function, so we had... There was the chief digital human resource or talent officer, because of this very issue, of saying the lines between our traditional workforce to our technology is blurring, and we've gotta have a way to think about culture and how do we cultivate the right behavior for silicon and carbon-based workforce.

Success theater vs. real metrics

Michael Krigsman: We have a really interesting question from Chris Petersen on Twitter, who says, "Are the organizations putting out the 'we have tens of thousands of AI agents' headlines getting real value from these agents or just trying to game a system of bad metrics?"

Bill Briggs: There's been a rash of studies, I won't name any one by name, but there's been a whole host in the last six months or a year trying to put an emphasis on the small percentage of actual value being created from AI investments. And we've seen the... In our state of enterprise AI, we see a much more positive.

But the prevalence of success theater is still real. And typically when we're measuring volumes of use case or volumes of agents as the thing that's the bar, it's a tell. 'Cause the value should be we've reduced X months of product R&D in life sciences, or we've had this many less restocks dramatically of groceries if it's a retailer. I mean, the actual business metrics and outcomes that you care about would be the headline if they were there.

In the meantime, it's a bit of a flailing to say, "Hey, we're gonna show effort in the hopes of return." But the, "Hey, let's get every use case we can think of and do a pilot for every one of them" is one of the traps we talked about up front. Because, you know, without the right investments in lowering the ceiling data foundations and core systems, putting in place the mechanisms to actually responsibly lift and scale, they're not going to.

And so you're gonna have these one-off isolated, they look interesting, they can't handle complex scenarios and use cases, they certainly won't scale across the organization, and you've got a press release that doesn't actually materially change any of the metrics that your shareholders would care about or your people would care about.

Outcomes over process adherence

Michael Krigsman: I love that, because you're really now talking about shifting the focus from the process, the process being the hand-waving, "We have thousands of agents"- shifting from that into, "Here's what they're doing. Here are the benefits. Here, they're reducing inventory, they're saving time, saving money," what have you.

Bill Briggs: I think anyone that's been, and like you, Michael, about 30 years doing this, so I'm sure we can find a video of me in the late '90s with much more hair and less wrinkles talking about start with outcome and value first as you deploy technology. So anyone that's out there that's rolling their eyes and thinking, "That's the most trite statement," I apologize. 'Cause I think it's a truth. It's the axiom of tech, period.

But AI, what makes it different is, especially 3 years ago, gen AI started becoming much more accessible. Suddenly it wasn't the tech organization that was championing we need to be investing and adopting, it was the CEO, and it was the line of business leaders, it was the functional leaders, it was the board. And I said it before, but just to underline it, the hope was this was the easy button. The hope was suddenly we can skip all the hard work. We can have our business counterparts being able to harness and deploy, and we wouldn't have to do all the things that were hard before. And it couldn't be further from it.

So the statement is still true. It's always been true. It's just trying to remind a lot of stakeholders that weren't involved in a lot of the technology investment business cases and implementations that this is almost like a law of physics of if we don't have that mindset, we won't go... You know, I'm an old Monopoly player. We won't pass Go, we won't collect $200, we won't see the upside that we hope.

Michael Krigsman: On LinkedIn, Greg Walter says this very eloquently. He says, "Finally, it's all about outcomes versus adherence to a process and rewarding adherence to that process instead of deliverables."

Bill Briggs: 1000%. The difference between- We're gonna deploy technology where the goal of technology is to have everyone work like having to follow the exact same steps in the exact same way.

The idea of, "Hey, we're gonna deploy technology with an outcome focus," like Michael said, which is exactly right, "and we're gonna teach you, give you tools individually," not in the development engineering shop, but the folks actually doing customer service, the folks actually doing work on the hospital floor, the nurses themselves, to be able to then think about how we can improve and evolve that tech too for even better outcomes or outcomes that we hadn't thought of when we first deployed the initial use case.

So that unlock is huge. But if I had to say, if you just have one thing, I would go with Michael's and say make sure every conversation about not just AI but broader tech is razor-sharp focused on outcome period.

Responsible deployment and guardrails

Michael Krigsman: So Paul P. says... This is kind of long, and I'm hoping I get this right. Okay. "Bill, as you might know, every technology is a double-edged sword depending on how we use it. We, conscious, conscientious members of society will use it in a responsible way, and bad actors will always take advantage of it to do harmful things. Now, from how can AI governance and design mechanisms not only reduce harmful behavior but also encourage constructive participation in society and reduce that deadweight loss from a socioeconomic standpoint?"

Bill Briggs: And I'm glad it started with a caveat, 'cause I think it's so true. Fire is great if it's cooking my filet, and it's terrible if I'm burning down my neighbor's house. And you can use any technology ever with the same kind of lens.

So then it gets to, you know, I think you're seeing interesting evolution of individual AI providers and how much they're embedding guardrails and policy into their product roadmap and their usage guidelines. That's a piece, and you're seeing, as I mentioned before, the mature organizations are being very thoughtful in having that not just as a written guideline to be followed, but it's actually codified in the data pipelines, the training and tooling, the deployment, inside their engineering and AI process and platform and what they're using to actually invest and implement to deploy.

So we all have to be active in our respective organizations. You know, and if you noticed before, it wasn't me being lyrical when I said trust includes security and privacy and regulatory and compliance and ethics and morality because all of those things are increasingly important, and I think we're gonna see organizations, their stance of what they do and how they use advanced technology is gonna be a piece of their brand and reputation, of how they serve their markets and also the type of talent they attract.

Michael Krigsman: Clearly, the implications and fallout of AI have tentacles through really the entire organization.

Bill Briggs: For sure. And I think there's some cautionary tales. You know, everyone that bought a Mac Mini a few weeks ago to do OpenClaw at home and gave it complete access to their calendar and to their root, pseudo into their repos, it was a lot of fun.

But from the beginning, there was caveats from the creator of, you know, watch out if you give an agent with the ability to spawn more agents access to everything on your digital life. We don't know what's gonna happen, and a lot of interesting things happened in short order.

And, you know, I think that's nice that it played out and enough people paid attention to be able to then take the lessons into the enterprise. 'Cause clearly, there's much more controls and for good reason inside of government and industry.

But, you know, there is the, you know, any technology with unbounded access, and you could run a... you could write a cron job that could do... Michael, we could have gone back in our days and had a Cobalt cron job that could do all kinds of nastiness if it had the wrong permissions and an ill intent.

So and I'm not trying to dismiss the, you know, AI has a different kind of scale and different kind of potential, but I think the same story goes back to responsible deployment technology includes making sure we've got mechanisms in place to control it from the outset.

Michael Krigsman: And as a matter of fact, Perplexity literally in the last day or two announced a product that is similar to OpenClaw, but it operates inside a secure sandbox, so you can have the benefits of OpenClaw but without the wide open security gaps, "Hey, world, you can have access to all of my data however you want."

Bill Briggs: And seeing the productization of the things that everyone could do even with the open source components of, you know, container, access, you know, having set permissions that are separate from your admin and root. The good news is what the best engineers would do intuitively, because they know better, starts to get productized and codified and up and across the stack. And we'll continue to see more of that.

How AI is transforming IT

Michael Krigsman: We need to talk about the CIO role and the impact of AI on CIOs. AI is fundamentally changing what it means to be a CIO from managing infrastructure to orchestrating an organization where human and AI workforces operate side by side. So what is the... how is the CIO role different today than it was 2 years ago?

Bill Briggs: Here's the other headline, Michael, that in the Tech Trends report, 99% of the organizations we talked to said that they are fundamentally transforming their IT organization, and I've yet to meet the 1% that's not. So if anyone's listening in, this is the chance to be seen, because of this profession, and it starts with the tech leader.

We have a couple things from the report, just other factoids I think are telling. One is that 65-plus percent report into the CEO now, which not too long ago it was mostly into the CFO or the COO because tech was basically an efficiency play, back office, keep the lights on. Another one, 68% said that tech is being seen as revenue generating, you know, where it's actually embedded in product, embedded in market.

And so with that you have the convergence of what used to be IT back office, operational technology, you know, in the warehouse, in the manufacturing floor, in the hospital theater, in the fleet. Those are coming together, and that product technology's coming together. And then the role of the team is different.

And, you know, I would pride myself over all of the development languages I knew over the years, starting with Fortran and the machine language into C and C++ and Java and JavaScript and Python, and I might have learned my last language.

But that point about what are the roles that we need now, and they look a lot different than the program and project management emphasis from the '90s, early noughts. It even looks different than the agile teams and the ways we were thinking about full stack engineering just a few years ago.

The good news is the barrier is really about the attitude, the aptitude of individuals to embrace that it's gonna feel different to be a technology professional. But there's a role for most to play in what we need more of.

It's just a matter of can we let go of... I had a lot of ego when I started doing agentic coding, and the idea that I would be trusting the output of an agent was close to 0 not too long ago. And humbly, I can say there's not a lot of lines of code I'm writing anymore, and this more in my hobby at home than my day job.

But our engineers are having the same experience. And for the best of them, it's actually a really exciting time because we get to do a completely different... It's much more system thinking and architecture, but you also have to manage the agents in a way that is not completely relegating all the skills we've built over 30 years, at all. This, in fact, it's making us even more powerful.

The CIO as storyteller and catalyst

Michael Krigsman: So if CIOs are becoming de facto enterprise transformation leaders, what does that say about how CIOs need to adapt today?

Bill Briggs: 2 big things. One is the influence and the vision piece of the role has never been more important. And so when we bring next gen CIOs together, we do an academy a few times a year and just invest time for our clients in helping the next generation of tech leaders be better.

The single most effective, impactful session we do is not on multi-agent systems or ops or infrastructure. It's on storytelling, of just how to structure and shape a vision and be able to help share that to influence and inspire. So that's a piece of it.

And the other piece is the growing reality, and certainly at Deloitte we have this as a piece of our DNA, that together is better than by ourselves. So how do we start thinking about partnering in the ecosystem for co-invention, co-innovation, which, you know, not too long ago, procurement was a race to the bottom of we're gonna squeeze all cost out of every investment and every provider, and it's completely shifting to joint ventures and co-ventures and co-innovation with many, many, many different types of established and new tech providers.

So all of a sudden, the CIO is almost like a VC portfolio manager and having to make some decisions about who they're gonna partner with, with the reality of those decisions that used to be you'd pick a database, you'd pick an ERP, and you can have a half-life of a career before you'd have to revisit them, is changing much quicker.

So then how do you cultivate those relationships, but also have a change expectant mindset? And that's technically how you architect it, so we can swap things out. But it's also contractually-

Michael Krigsman: The CIO has long been the only corporate function to touch every single department across an organization, and so theoretically, CIOs should already possess these cross-functional skills. If that's the case, what's actually new here? What's different?

Bill Briggs: They've had the seat at the table and visibility into those- But we have a 4 phases of a tech exec framework, and the bottom 2 are ops and technology. And I think they've always shown up really well on those two. And the ones that didn't weren't in the seat for very long.

But the strategist and the catalyst are the 2 that are much more forward-looking. And the strategist is saying deeply understand what the different businesses and functions, what makes them tick, where their priorities are, helping them realize their strategy.

The catalyst is actually the one... My dad was a chemical engineer, so I love the definition of catalyst is how do we amplify and accelerate a reaction without being consumed therein? So how do we actually shape the vision, inspire action?

That's the piece that has been not very prevalent in a lot of tech leaders over the years that's so desperately needed now. And so it's moving from order taker to strategic peer and maybe even to someone that the finance lead or the regional line of business leader is taking inspiration and direction from of these are things we have to do if we wanna take on the next strategic goals.

Human times machine

Michael Krigsman: This is from Dr. Carolina Sanchez Hernandez, and she says, "How do we approach human oversight and human in control over ever more complex AI systems?" And I'm thinking there's so much. I'm thinking agents, for example.

Bill Briggs: We just published our Human Capital Trends report, which is the sister org, brother org of Tech Trends that has a very detailed breakdown of how to make that exactly real and right.

But the takeaway here, and I love this spirit to end it on, we started with a technologist saying we should spend less on technology, Michael, which is probably not what you had on your bingo card, and we'll end with the human times machine is the most essential equation we need to figure out.

And it's not just for oversight and accountability. That's a piece of it, how do we keep human in loop on loop? But also as we think about work redesign and culture and how do we harness the full potential, the tools are gonna continue the tools. They're ingredients, they're not recipes, and we need more chefs able to give us 3-star Michelin meal equivalents that we can then not talk about headlines of agents. We can talk about litany of real metrics that matter.

Michael Krigsman: I love that, the litany of real metrics that matter. Bill Briggs is Deloitte's CTO. Bill, thank you so much for taking your time to be with us today, and especially given the fact that your ISP, that your internet connection stopped working at home, and you're making do on your phone.

Bill Briggs: We persevered. Sorry if anyone got seasick or nauseous from moving around, but we did it, Michael. Always a pleasure.

Michael Krigsman: We did it. And folks who were listening, and you guys who asked your questions, even if I didn't get to your question, your questions are awesome. You guys are an incredible audience.

And right this second, I want you to go to cxotalk.com to subscribe to our newsletter because we have great shows coming up. Our next show is in a couple of weeks, and we have the chief financial officer of HPE, and we're gonna be talking about some of these issues with her. So go to cxotalk.com, subscribe to our newsletter, ask your questions, and keep them coming, and we'll see you next time. Have a great day, everybody.