AI Reality Check:
What Works and What’s New

Paul Daugherty's Career and Current Roles

Michael Krigsman: Paul Daugherty has had an incredible career. At Accenture, he was Group CEO for Technology. He was their Chief Technology and Innovation Officer. He's written two books. He's advised many startups, sits on boards, and is now AI Advisory Chair at the large private equity firm, TPG. Today, on CXOTalk number 895, we take an AI reality check with Paul to examine what works, what fails, and what's next. I'm your host, Michael Krigsman. Let's get into it.

Paul Daugherty: I spent a lot of years at Accenture. I was there for 38 years until I retired last year, and it was an amazing vantage point from which to see the transformational power of computerization and digital technology and everything that happened over those four decades. And to put it in perspective, in a way that some of you will appreciate, the first production system I wrote was in punch cards on a mainframe. So, it's quite an arc from there to where we've gotten to today. And I've always been passionate about technology. In my role at Accenture, it was always about leading what's next and leading us into what's next.

So, I'm still an advisor to Accenture. I still do some work there. I took on a role, as you said, with TPG, which has been really exciting. TPG is a great firm. A lot of partners are based in the Bay Area, at the heart of the technology. They've been a very tech-forward PE firm, large PE firm, 260 billion of assets across a lot of different platforms. And it gives them a way to invest in big companies, the large companies that are impacting the direction of technology, as well as through some of their platforms involved in the Leaders of Tomorrow. And it's been great to come on board and work with TPG, and we can talk a little bit more about my role there.

AI Reality Check: Hype vs. Reality

Michael Krigsman: Where are we today? What's the state of AI? Are we in a bubble? Is all of this hype real? What's going on?

Paul Daugherty: This isn't a bubble. It's real, in terms of the real transformational power of the technology, and what's already happening is real. We're just at the early stages of it, which is where you see some of these bubbly things happening. So, let me talk about a little bit about what's real and what's hype, what's real, and what's missing.

I think what's one example of hype, I think, is around AGI, artificial general intelligence, and one of the problems with AGI is there's not really an agreed upon clear definition of it. But generally speaking, people talk about AGI in the sense of general-purpose AI that can exceed the capability of anything a human can do, and I think we're far, far, far from that. And there's a group of people and a group of eminent people that I respect that claim we're either there or pretty close to it. I disagree with that. I think what we have today are really amazing technology that approximate a lot of the things that people could do, but they're far from real AGI, as I would think about in terms of human capability.

Mustafa Suleyman, who many of you may know from DeepMind and then Inflection, now leading AI at Microsoft. Mustafa Suleyman has had a series of posts that you should go look at if you haven't. The recent post about consciousness and the risk of assuming of giving AI agency of things AGI consciousness because it can lead to very bad human outcomes by misinterpreting what the AI really does. And I think Mustafa explains it really well, so I'd encourage you to read his post. That's... But I think that's a bubbly aspect, and lots of people talk about AGI. Companies are claiming they're already at AGI, but I really don't believe that's the case.

We've had AI around for 70 years. The learning branch of AI with the transformer technology and large language models is real, and it's amazing. If we just froze AI right now, no more innovation, no more products, freeze it where it is right now with large language models, I think you'd have 5, if not 10, years of business adoption of what the technology can do today. Much less the massive, amazing arc that it's on.

Michael Krigsman: If we were to freeze it today, that implies that companies have seen benefits, they're getting benefits. Just what are those benefits?

Paul Daugherty: Generally speaking, there's benefits in the obvious areas you see coding and software engineering. A lot of advances being made there, a lot of great tools there. In customer service types of applications and marketing applications, again, could go into detail. One organization I'm working with that had 30% increase in productivity and 60% increase in net promoter score and satisfaction by introducing new forms of generative AI into their customer service organization. And then in different industry functions that we can unpack a little bit. So there's real opportunity there already.

But just circling back maybe, I just want to finish one thought on the what's missing from AI because it's really important to where things are right now. I think everybody's a little too centered just on the large language models and the scaling law around large language models, and I think to get the next level, to get to anything approaching AGI, and to get to the next level of business improvement, we really need to focus on other forms of AI. There's a lot of things out there. Large quantitative models, physical... Or sorry, liquid AI, a number of other things that are getting more at symbolic AI. Yann LeCun has been talking a lot about this, and I think the combination of the amazing things large language models can do and bringing that learning branch of AI, together with the symbolic AI techniques, is what's really gonna propel us to a next level of what we can do in business and next level of what AI itself can do.

Generative AI's Impact on Work and Organizations

Michael Krigsman: What about the impact on organizations? You wrote a book, and a year ago, pulled out sections from it that appeared on the cover of Harvard Business Review. And you said this, "Most business functions and more than 40% of all US work activity can be augmented, automated, or reinvented with gen AI." So, as you look back a year later, how does that prediction stand up?

Paul Daugherty: Yeah, it was Embracing GenAI at Work was the article on Harvard Business Review. And yeah, that came out of research, as you said, from our book and 1,500 organizations, lots of research that we did. And I would... A year later, we were low. The 40% was low. I think we significantly understated it. We've done some additional research and can peg it above 50% now in terms... And that's looking at work hours that people expend that all of us do, that can be impacted by... or that will be impacted by generative AI. Where we are... So the coverage, so to speak, the surface area that it impacts, is increasing, or the potential's increasing.

Challenges and Early Adoption of AI in Organizations

The actual what's happening is still pretty light. So let's just be real about this today. In terms of adoption, we're still in the early stages. From some other research that we've done, we would take a different cut and say that about 15% of organizations have made significant... One-five have made some significant progress on scaling AI somewhere in the organization, 15%. Another roughly a third of organizations, so the balance to get you to 50%, are close on the heels and doing things, working to scale. And about 50% are still more in the experimentation, trying things stage, and maybe individual use cases and parts of the organization, those types of things.

So there's... So, yeah, there's over 50% of the work can be affected, but that's not happening today. That's the potential, and organizations are working to get there. And I think one of the myths about generative AI now is, I think you'll... You hear a lot of the tech industry talking about it, and it's there's an easy button. "Just buy this and implement it, and hit the easy button, and you transform your business for generative AI." That's not the case. This is really hard, probably harder than the technology that come before. This is harder than cloud and the cloud transformation because it's invasive in your organization, it's changing processes, it's changing roles of people, it's reeducating people, and transforming your organization. You had to do some of that always, of course, with past technologies, but that's... This is taking it to another level with AI, which means it's... We got a lot of work ahead of us to get there.

Invasiveness and Organizational Transformation with AI

Michael Krigsman: Paul, you use the term invasive. What is it about AI that requires this invasiveness on organizations as companies adopt it?

Paul Daugherty: There's two dimensions of it, one of which is the invasive... The technology is invasive, and organizations are starting out with legacy systems, with technology debt that's been talked about a lot for years, so I won't go into that in detail. But the average organization still spends 50% to two-thirds of their budget keeping lights on on legacy rather than new capability. It's much higher than that in a lot of organizations, and that creates... It's not just... And that creates an investment challenge, but it also creates a real anchor and roadblock to getting new capability introduced.

And AI is significantly different. You have a lot of work to do at the data level with your data fabric and getting that in shape. You have a lot of work to do at your architecture and application layer, at your middleware layers, and some of that technology's immature. We can dive more into this if you or your audience would, but we lack categories of products right now. There is not really good mature agentic middleware out there for organizations to tap into, and that's still an evolving product category. So on that technology dimension, this is hard work.

The second dimension that makes it hard is the process dimension, because unlike cloud, which a lot of the benefits of cloud were technical, I can move to elastic capability and external data centers and things, and reduce a lot of cost. And yes, there was business impact, but got a lot of benefits. To get the benefit from AI, it's going invasively into processes and changing the way people work, and that's hard to do. There's change management issues, there's organizational issues often where things cut across silos.

For example, we're working with an organization which that's looking at AI and customer service, but they realize the big benefit for them isn't just in customer service, it's integrating across product management and product innovation so that they can really dramatically shorten the cycle time between getting product... customer feedback and product innovation out there. And that's really hard to compress those two functions into one and make it all work the way they want to with AI, and that's an example of the invasiveness on that process side. And then you can get into cultural change and everything else. That's part of it, but those are at least a couple of the dimensions that make it difficult.

Michael Krigsman: So according to what you're saying, and I don't want to put words in your mouth, AI is a forcing function for relatively broad organizational change, with tentacles and implications that extend in a lot of direction. Is that an accurate way of saying... re-saying what you just said?

Paul Daugherty: Your AI strategy is your business strategy. And yeah, there's a big debate in organizations, do I... They want to do AI. Do they need an AI strategy? I'd argue generally speaking, what you're really... What you're really doing is thinking about your business strategy with AI as a significant enabler, along with whatever other things are in the mix that you're dealing with, macroeconomic issues and everything else that's going on.

So yeah. So you have to get into the what's happening in the business. Now, there's things you can do more easily. You can... You get some additional incremental productivity by rolling out copilot tools and doing a variety of things. So yeah, you can do that. But to really get some of the bigger benefits and look at the impact and how you can get there faster than your competitors in some of these broad organizational and process transformation, that takes more work.

Michael Krigsman: This type of change, it sounds you're describing ERP and business process transformation from 25 years ago. How are we different today?

Paul Daugherty: Client-server technology is what propelled Business Process Re-Engineering, the famous Michael Hammer book and everything that some of the audience will remember, from the '90s. And that, the message there was similar, you can't... You can't create the superhighways of tomorrow by paving the dirt paths of yesterday. And that's similar to the message that I was just saying. I think it's just broader. You look at what cl-... You look at the surface area of an organization, that client-server impact, and it was quite low. Whereas AI is impacting pretty much everything that an organization can do.

So, I think it's... It is a similar message in a similar need to step back and rethink about it. And that's why I call it... Rather than re-engineering, I call it re-imagining. You really have to step back and look at the art of the possible, because you don't just have your human workers and customers in the mix. You have your AI... Your AI agents, and you have AI customers and partners that... that are gonna be part of your interaction that you're thinking about. So, it does really require this kind of re-imagination capability.

Michael Krigsman: Subscribe to the CXOTalk newsletter right now. Go to cxotalk.com, because we have live shows, and they are... We have amazing, really, genuinely amazing guests. So, subscribe to the CXOTalk newsletter. Do it now.

AI Applications and Measuring Business Value

Michael Krigsman: We have a question from LinkedIn, on LinkedIn, from Larry Rubenacker. It's a very specific question. He says, "What do you see as some of the key AI apps for the utilities industry?"

Paul Daugherty: Utilities is an interesting industry because you've got services, customer service and billing and things that. You've got manufacturing, process manufacturing, capability generation, transmission, distribution, et cetera. You've got construction type of capability. So, utility is kind of a microcosm maybe of a little bit of everything that you have across other industries.

So, I think there's clearly opportunities in that customer service area we're seeing across industries. And I know that's... That's an area that I've done some work in, and I know that a lot of utilities have done work in. Yeah, so that's... That's an, I think a clear area of opportunity in the mix.

In the more the operations side, there's been some really interesting work and innovative work that some utilities have done around using AI for things wildfire prevention, in the West Coast of the US in particular, where fires have been a cause of major outages and system issues. They're using AI in a couple ways. One is to detect fires more quickly using AI analysis of satellite imagery and all sorts of things you can get. But they're also doing... One utility in particular is, with my... Working with my former company, is doing some work to look at how do you... How do you equip workers to go into very hazardous wildfire environments to preserve, protect the infrastructure, which is very difficult to do.

So, a lot of implications that in utilities as well. There's some interesting things going on in the... In emissions management for utilities, to help utilities monitor emissions. For example, things tracking methane leaks or tracking methane burn-off in different ways to allow utilities to operate more sustainably. So, I think there's a lot of use cases in utilities. Because again, it kind of pulls in a lot of these use cases that happen across the other industries. Utility just is a challenging environment sometimes because of the rate-based model and the way capital is allocated. It's sometimes hard to invest in those opportunities. But I think there are a lot of opportunities.

Michael Krigsman: We have another question, and this is from Joy Tuhan. And she says, "Paul, it was great working with you at Accenture. What are some KPIs and metrics that organizations should use to show AI is creating real business value?" It's a really important question.

Paul Daugherty: If AI isn't moving your... The performance metrics on the page one scorecard of the company that they track, whatever they... An organization, they're gonna track all sorts of key metrics to their business, from days services outstanding to accounts receivable, to all sorts of productivity measurements and manufacturing other things. And I think if AI isn't making an impact on those things, I don't think you're looking at AI right in terms of the impacts it can have.

So, I think it's a matter of looking at the KPIs that are critical in your business and understanding which of those can be dramatically impacted by AI. In my Accenture role, what we did, to pick up directly on Joy's question, we were looking at, with AI even before generative AI, with just AI, classical AI, so to speak, before. We were looking at our, for example, our software development processes and looking at, how do we improve the efficiency and productivity of those every year? And we were improving them about 10% a year every year. And we would... We'd measure that to make sure we got those outcomes.

So, I think that's... I think the key is not necessarily a whole new set of outcomes, but to be very intentional about the AI projects that you're pursuing and be able to tie those to the page one metrics that you manage as a company. Of course, you're gonna have other metrics for your AI programs. Are the programs working right? Are they successful in delivering value and everything? But that's key. The one other element of that that was gonna mention, but maybe in some other context, but that is critical is, I think organizations also need to avoid AI use-case-itis.

There's too many organizations that I walk into that very proudly show me a hundred use cases of AI across the whole org. They got so many. They're so excited.

Focusing on Impactful Use Cases in AI

They got so many. I'm, "pick seven, pick five, pick a number, but focus and pick a few and really tie them to something that makes a difference in your company and make them happen." I don't think it's victory to have a lot of use cases. I think it's victory to have a small number of use cases that make a difference, so you can show the value, show the impact on your company, and then move to do more.

Michael Krigsman: I have to reinforce that comment. There are so many folks who descri-... Who have chopped up the organization, evaluated the organization, and identified all of these use cases. But as you said, identifying use cases and putting people to work in so many different areas does not necessarily help you reach the finish line.

Paul Daugherty: I think that's right. I think you can solve small problems and miss the bigger problem. There's one large consumer goods organization that I worked with, we worked on AI in the sales arena, and the initial use cases were around how do we get sales leads better to make a sales rep more effective, and all sorts of things. It turns out the bigger use cases were thing, or the bigger productivity improvements, I should say, were cutting across use cases. So, how do you combine a salesperson's productivity with things inventory availability and delivery times so that you could cut across that and get things to customers faster?

And there's a whole different way of thinking, how can you move product to customers faster rather than just making an individual salesperson more effective? And you'll miss opportunities for those bigger re-imagining if you just get caught up in use cases.

AI Deployment Challenges and Talent Transformation

Michael Krigsman: This is from Bavana Bhagat, and Bavana says, "Is the talent transformation with AI for real or will it be another RPA?"

Paul Daugherty: AI risks being another RPA if deployed the wrong way. I think RPA... RPA was, it had this problem of being a little fragile. You implement the RPA technology, robotic process automation for those who aren't familiar with it, which is being able to stitch together and automate processes across different systems. It's very powerful technology, does a lot of great stuff, but I think it hit, it hit limits in terms of what it could do and was, wasn't flexible to adapt to changes on, the underlying systems generally speaking.

I think AI could be subject to the same limitations if not deployed very well. So I think the architecture of how you deploy AI is super important. The way you insulate from changes in underlying foundation models is super important. And I think architects are back. For those of you listening, if you're enterprise architects, we need a lot more enterprise architects to shape these AI solutions going forward, that really understand all these things, because it's super important. And otherwise, you can end up with brittle AI systems as well.

Now, Bavana, in your question you said talent transformation. I don't think... so I'm not sure exactly what you meant by the talent transformation related to RPA and AI. But there is a, there is a broad re-skilling of people, develop software engineers, developers, architects, et cetera, to learn AI, maybe more so and feel certainly more so than people needed to advance their skills with the RPA era. So I think that is part of the journey ahead, too.

Michael Krigsman: Sachin Inamdar says... He's from Accenture. He says, "What are your thoughts on the future of the consulting business going forward?"

Paul Daugherty: Services, generally tech services, let's just include consulting, systems integration, and outsourcing, all that stuff together. Tech services generally, I think is hugely impacted by everything happening with generative AI, and it's probably the biggest impact for... More of an impact than any of the previous technology changes. And what that means is it's at, this, the industry's at the center of some of the changes that are happening, which means it's both a threat and an opportunity.

It's a threat because I think the existing way that consulting happens, the existing ways that systems integration and outsourcing happens aren't gonna be the ways to do it, and are already not the ways to do it in the gen AI world. So, I think the challenge for any company, for Accenture, for any company in the industry is how to be first and ahead in transforming to the new processes.

So, on the consulting side, rather than being, having smarter people and good knowledge capital and walking in with a better presentation to the CFO's office, that was yesterday. Tomorrow what you need, you need to walk in with a foundation model with pretrained and the client's data, and in the CFO's office, be remodeling their business dynamically in the first meeting. And that's a big transformation on a lot of levels, and that's emblematic of what we, what needs to happen in the industry and what organizations Accenture right now are working on.

Michael Krigsman: What does that do to the consulting industry when you have a lot of people whose primary skill was listening, and now they have to, of course, still listen, but there, there are so many more dimensions. And the talent pipeline has not necessarily prepared them for that?

Paul Daugherty: This stuff is hard, as I've been emphasizing a lot as I've gone through. Clients need help in doing this, so I think there's a big role for services organizations to help companies do this faster, better, more effectively. So I think there's a big need for this, for services to do that. But they need different shape to it. So an example, you need more forward deployed engineers in the business to more quickly drive change for the client.

So it's a matter of transitioning the talent, which I think companies have been, Accenture as an example, have been very good at, transitioning and retooling talent. It's about rethinking the workforce. The degrees of leverage in pyramids and things will change, depending on the type of work as you look at this going forward. It's about moving services to software, so what of the services that are provided, where are there opportunities to platformatize, if that's a word, or transition to services, the Palantir model in that way. So, I think those are some of the changes organizationally, talent-wise, technology-wise, Michael.

Prerequisites and Challenges in AI Adoption

Michael Krigsman: And this is from Ebru Bayar, who says, "There are still organizations out there who haven't done cloud transformation. They're lacking data quality and governance. Do you see any prerequisites that organizations should focus on before jumping into defining an AI strategy or executing that strategy?"

Paul Daugherty: Those are impediments, but you can't wait. So it's, it can't be sequential where I'm gonna do my cloud and data and then eventually I'll get to AI. So, there's some companies that have decided that. There's some companies I've worked with who have looked at it and said, "Man, I'm in such bad shape on data. I'm just gonna do that, and then I'll deal with AI once I'm done with that." That's a small minority though. I think most realize, "Okay, I got a lot of work to do on the data side, but I gotta get some elements of AI working. I gotta make progress on my journey." And they're melding the two together.

So, I think it's... you gotta look at your specific circumstances. Those are big impediments. Many companies, I'd say most companies, have some variety of the challenges that you mentioned, Ebru. But it's a matter of figuring out how do you parallel track it and manage that in an effective way.

Michael Krigsman: Let's talk about adoption success and failure, what can go wrong. We have all heard about a recent MIT study that said 95% of gen AI projects fail. Now, there were some issues with that study, maybe you could say it was talking mostly about proofs of concept. Nonetheless, when it comes to gen AI, what does failure mean? What causes failure?

Defining Failure in AI Projects

How do we even recognize it?

Paul Daugherty: It came at a time in the market news and when people wanted to seize on it and use the message that 95% of projects were failing, the headline, in different ways. So, I think it was unfortunate that it got the traction in the way it did because I think that's kind of a misleading conclusion. And I don't think 95% of generative gen AI projects are failing. That's very... I have a surface areas of hundreds of AI projects that I see, and it's nowhere near 95% failure. Generally speaking, people are making progress in different ways.

Michael Krigsman: And what is failure? What how do we define failure? And then-

Paul Daugherty: I mean, the way I would define failure is you don't achieve the business objectives you set out to achieve, which gets into one reason for failure, is sometimes they don't... they don't... there's no clear north star of what they're trying to achieve, or no clear metrics from a business perspective. So, that's how I define failure. There's other... I guess you could define it other ways, but I think that's one way I see it.

I've seen people develop great individual use cases that they couldn't scale. I'd say that's an element of failure, too, because they realize that that... they developed it, great POC on a small scale, but they realized they had... they were missing the data or whatever they might've needed to go scale it more.

So, and I guess there's five things that I find that companies that are successful are focusing on. And I've been talking about these for a while, so these aren't necessarily new if you've heard me talk in the last few months. But these are... there's five consistent things that I've been seeing.

One is the value point that I just mentioned, Michael, back to the definition. You need to focus on the value, have a business case, tie it to metrics and understand what that business case is. And value generally will mean really looking across the use case side is looking at things that, reimagining the process that are doing something significant. So, it's this value piece is important. Sounds obvious, why wouldn't everybody do that? But not everybody does. A lot of people are starting in different ways.

The second point is what I call the digital core in data. And that was the at the heart of the question we just answered. But sometimes companies don't have their digital core or the systems in place. If you're not on SAP's S/4, for example, their newest generation of SAP, you don't get access to their AI. So, you gotta get your digital core in place and upgrade your SAP to the latest version to get access to the AI, as an example. So, I think getting a digital core in place and getting the data in place is key, and companies are generally speaking behind on the data piece. But that's can cause... that can cause failure or limit the effectiveness.

The third piece is talent and under-investing in the people, and this is the talent to do AI. Do you have people who can do the AI? But it's more broadly the talent across the workforce. Are you training the workforce with what they need to do? I think there's some best in class examples of companies that are doing a great job of educating their entire workforce on AI and what they're doing and what their objectives are, and involving their workforce bottom up. And these tend to be enlightened leaders who get it. Bottom up, incorporating their workforce. Great things that companies Merck, Walmart, I think, are examples of companies doing this. That are really building the talent base they need going forward, and the learning systems they need.

The fourth point is responsible AI and really putting the guard rails around it so know you're not getting into trouble as you do AI. I think leaders leading companies, I think, are doing that well. And the fifth point is viewing it as an ongoing multi-year change program, so you're building the capacity for technological and process and organizational change into your approach rather than viewing it as getting one use case or one project right. So, those are the characteristics that I see are companies that are doing well.

AI's Role in Private Equity and Investment Analysis

Michael Krigsman: Let's talk about private equity and venture capital. Just give us some insight into your role at TPG as AI advisory chair.

Paul Daugherty: TPG, large private equity firm, 260 billion in assets, and they span all these different platforms. It's not just buyout capital, it's growth. They have growth capital, they have credit, they have real estate, they have impact funds, climate fund, and others that are focused on impact investing. And it's really an amazing opportunity to combine across those.

And the reason they brought me onto the team is it's a very tech-forward team. They really understand it, they've been investing, they've been very successful investors in a lot of domains, including technology, for many years. But they really wanted to look at where technology's creating opportunities and how to accelerate some of the opportunities that they see.

So, the role as AI advisory chair is to help with the with the overall strategy and investment themes, and work across those different capabilities that I just discussed, looking at where there's opportunities. Helping on deals that they're doing, new deals that they're doing. And also work on pulling together, expanding the network, pulling in other expertise on AI, building other connections in AI that'll help them be more even more effective going forward. And it's, again, been a lot of fun. And given their scale, they can have impact kind of an impact on a big way, which has been very exciting to see. So, again, that combined with the VC angle has been a lot of fun to see it from both perspectives.

Michael Krigsman: As you're looking at startups and investments, of course, the traditional evaluation points have been the product market fit, the team, and so forth. How does AI now fit into investment analysis and decision-making?

Paul Daugherty: It's a strategic thing. I'll go through this first. There's a strategic thing you have to look at that's a little different, and I'll talk about technology-related companies first. But, but it's not just technology, but this, I mean, frame couch this in technology terms.

I think there's five ways that AI can impact a technology company. First, it can be no impact. So the AI doesn't impact it at all. That's number one. Rarely the case, but it could be the case that that's the case.

The second is AI enhances the what, what, the SaaS model or whatever the company does today. And look at Workday as an example. The new agentic... They just had some new announcements at Rising this week, and AI enhancing the capability of what they do. I think that's the second level or second example.

The third level is, the AI becomes more important than the underlying SaaS product or whatever itself. And the AI is the focus. I'd argue that when you look at GitHub, when you look at maybe ServiceNow now, when you look at the future of Salesforce, the AI becomes the defining capability around the product.

The next level would be AI goes even further and commoditizes the underlying technology. Look at Klarna, and the way, the way the CEO of Klarna has talked about the impact of generative AI and how he works with different enterprise software companies as an example. Maybe Zendesk, others, that fall into this kind of fell into this pattern.

And then the fifth level is, is AI just compresses the whole spread in the category because you don't need the software. Chegg might be a good example of this. Chegg, the learning company who, when the generative AI models came out, people flocked to, for free or whatever, for very low cost, the generative AI models and Chegg lost a lot of its population there.

So it goes... That's kind of a little bit of a framework to look at. I think you have to look at that from an investing perspective, really look at that and understand not just today, but if you look at the move of the market space in the next years, what the impact is.

Michael Krigsman: We recently had the CEO of Zendesk on CXOTalk discussing their AI transformation, so if anybody's interested in that, we had an in-depth discussion, so just go cxotalk.com to check it out.

Future Trends Beyond AI and AGI

Michael Krigsman: All right, let's take another question very quickly now because we're just running out of time, and this is from Soufiene BEN SABEUR, and he says, "What is next after AI beyond AGI?" And really quickly, please, Paul.

Paul Daugherty: The things to look at beyond what we currently consider as AI and AGI, I think look at physical AI. I know you said beyond AI, but not enough people are focused on physical AI, which is where really Elon's going with X and XAI and everything he's doing, and where World Labs and organizations are going. So think physical AI and AI meets the real world as one thing.

Look at quantum. It's still general purpose quantum, still number of years off, but some specific impacts of quantum happening fast. And related to this world, real world blending thing is robotics and the massive advances in robotics and these new physical world models that are coming helping us make... We'll, I'll predict on the show right here, we're gonna have a moment probably bigger than the ChatGPT moment around physical AI and robotics and what they can do. That's coming relatively soon, and maybe even bigger, bigger to us in realization than, again, the ChatGPT moment.

Michael Krigsman: How soon?

AI's Impact on Data Management and EMR Systems

Paul Daugherty: I think that's within a couple years. In terms of really seeing these physical models and what they can do. Say two, maybe two years, certainly about five years.

Michael Krigsman: Stephen Redmond on LinkedIn says, "How can organizations get ahead on organizing and cleaning their data while not losing momentum on AI adoption?"

Paul Daugherty: Really think about your business case, because what I see time and time again at organizations is people coming up with great data plans and data strategies, and it gets shot down by management, whatever, because there's not a business case tied to it. So I think you need to think about how to link your AI benefits or whatever benefits you have, an ERP transformation or whatever, with getting your data state advanced, and link things together so that your data transition... your data modernization doesn't get cut off and shut down, which is what continually has been happening to companies for the last decade or more.

Michael Krigsman: Rajat Kumar says, "What will be the future of EMR, electronic medical record applications using AI? Are some of the major players integrating AI into their apps, and what kind of integrations will that..." Basically, the future of electronic medical records.

Paul Daugherty: If you look at the recent announcements from Epic at their event about a month ago, those of you that are in this space, they made a broad set of announcements, and really, an expansion of what you think of as EMR. They're expanding Epic's role well beyond what you thought of as EMR to get into these care provision activities where AI will really have big impact.

So I think if Epic's vision of the future is right, you'll see the EMR providers critically, playing a critical role in automating a lot of aspects of patient care, revenue cycle management, and the. There's other views of the world which say that'll build up around the EMRs, and question will be which vision plays out.

AI's Influence on Jobs and Responsible AI Practices

Michael Krigsman: Arsalan Khan says, responding to your comments earlier about enterprise architect, architecture, he is a recovering enterprise architect. He says, "All jobs will be affected by AI. Are there any that are still untouchable by AI?"

Paul Daugherty: Physical skills that people do when you go out and do things in the real world. I think those will take a lot longer for AI to impact. I think anything where you're dealing with knowledge and information, it'll have impact. That's, there's... We have a lot of data, I can put, give you a link to send around, based on different labor categories of where gen AI will automate versus augment, and where there's no impact. I can send that out maybe as a follow-up, Michael. I'll give you the link, and that does it by industry as well as function.

Michael Krigsman: We need to talk about responsible and ethical AI. Everybody, every technology company says, "We are doing responsible AI." What's the reality behind that today?

Paul Daugherty: Less than 20%, and that's probably a generous number, of organizations are really doing responsible AI the way I would characterize the need for it. To do responsible AI right, I can't get into each one of these, but you need three things.

You need principles; everybody's got that. You need, I want to be fairness, fairness, privacy, security, explainability. So, so everybody agrees on the principles, most organizations have that.

The second thing you need is processes. How do you do it and make sure that you're doing AI responsibly? Is it baked into your software engineering practices, et cetera? Most aren't doing that.

And the third thing you need are tools and measurement. Do you understand every AI use case across your organization? Do you know what the risk level of each use case is? And do you know what's being done to mitigate any bad outcomes from irresponsible outcomes, so to speak? Most organizations aren't doing that.

So, we need to get beyond the principles and lip service to responsible AI to implementing really rigorous, industrialized, responsible AI. And I said, less than 20%, probably a lot less, are really at that stage.

Michael Krigsman: That was a harsh commentary.

Paul Daugherty: It's true. I've been preaching on this, so to speak, for about eight years now. Involved a lot of the organizations that have been advocating for responsible AI, and it was preaching totally into the wilderness with 0% listening for a long time. So I think the 20% I view as an accomplishment, we're getting more seriously. But I think too many organizations don't really understand that it's a business issue, because the biggest risk of AI is eroding trust. If AI is used, deepfakes, whatever, think about all the different ways AI can tarnish your brand.

Those companies that really develop trust in the way that they're using AI are gonna have a competitive advantage over other companies. This is about business and strategic value of your brand. This isn't about a feel, a feel-good thing. I think too many people still think of responsible AI as a feel-good thing, rather than something that's strategic and essential to your competitive survival.

Michael Krigsman: What's your view of government regulation of AI?

Paul Daugherty: I think it just needs to be balanced. I think we need government regulation that encourages innovation, leading innovation, and prevents and has the right focus on preventing the real harms, deepfake and some of the real obvious harms that are around us. That's... There's a lot more to say on that. But that's the one sentence.

Future Technologies and Personal Reflections

Michael Krigsman: And one sentence, what technologies are you watching now?

Paul Daugherty: One new thing I'm watching that I think is in the underappreciated, undervalued category, or really under-thought about category maybe now is crypto. There's a lot of people thinking about crypto, but it tends to be the crypto-converted that are thinking about crypto. But crypto as a general purpose technology for tokenization, not just of financial assets, but of other things, data, manufacturing, et cetera, I think is poised with recent steps the US has taken to really legitimize it a lot of ways, to really have a much more profound impact on businesses and everything we do than the non-crypto community is aware of right now.

Michael Krigsman: What is your trajectory for the next steps in your amazing career?

Paul Daugherty: I'm gonna keep doing all the fun things I'm doing now because this is what I love to do. This arc of technology has been my career and my life and wanted to stay close to what's happening. Beyond that, it's more time with my family, which, love to do and I'm having a great time with that, and hopefully getting a little bit better at guitar, which has been my retirement project since I left Accenture. I'm trying to gear up and spend more time on guitar. But all this exciting technology stuff is taking time away.

Michael Krigsman: I did notice the guitar in the background.

Paul Daugherty: Yeah. I'm getting pitched on a lot of startups that can help me learn music faster. So, that's where I'm hopeful.

Michael Krigsman: And with that, we're out of time. A huge thank you to Paul Daugherty. He is currently AI advisory chair at private equity firm TPG. He's an advisor to Accenture. He's doing a lot of other things. Paul, thank you so much for taking your time to be with us.

Paul Daugherty: It's always fun, Michael. Thank you. And thanks to your audience. This is an amazing audience. That's what makes this so fun, in addition to all the work you put into it. And I love the questions, so thanks to all who are tuning in.

Michael Krigsman: Yes. For the folks who are watching, there's so much information that has been packed into today. By Monday, the edited video... We'll do some light editing. Will be up on our site. And then, next week we'll create a summary. Watch this video. If you're interested in any of the topics that we've discussed, the replay will be there. You can take your time and you can unpack it. And right now, subscribe to the CXOTalk newsletter. Go to cxotalk.com. We really have amazing shows coming up. We'll see you next time. Thanks so much everybody. Thanks to Paul Daugherty, and have a great day.