← Digital Disruption with Geoff Nielson
Loading summary
Transcript65 lines
- [00:01]
Interviewer
Hey, everyone. I'm so excited to be sitting down with Scott likens, who leads AI engineering and emerging tech R&D for PwC. What's cool about Scott is his experience across the emerging tech landscape and actually implementing them in an R and D capacity. What I want to do today is see if we can pull back the curtain on the research and development actually happening with these technologies right now and learn what secret capabilities actually exist and maybe even find out what he sees on the horizon and. And the next horizon. Let's find out. Scott, thanks for being here. Maybe just to start, for people who don't know, can you tell me a little bit about, you know, kind of your story, you know, what you've been up to in your career and what you're doing right now?
- [00:43]
Scott Likens
Yeah, so it's been quite an interesting career. I mean, currently I'm the global chief AI engineer at PwC, so based here in the US but looking after everything. AI engineering around the world. And as a big firm, that means a lot. Part of my job is also looking at emerging tech, so research and development, you know, beyond AI. And obviously AI is the word of the moment, but there's just a ton of innovation happening, whether it's quantum computing. Blockchain continues to grow like crazy. Virtual reality, even with synthetic reality kind of being supercharged by AI. My career started in software, actually. I was studying to be a pilot and to pay for flight school. I was working at a software company and it just happened to be one, building one of the first Internet browsers. And for me, that really was this glimmer of imagination of what could happen in technology. So I started early days of the Internet building the browser, transitioning into really just the next wave of everything. So living through the browser and the Internet boom, the E commerce boom, the social media boom, the cloud boom, big data boom. And now what seems to be more of an emerging tech and AI wave, which has just been amazing. So really interesting journey.
- [02:00]
Interviewer
No, that's awesome. And one of the things that I thought was so cool about your story is that you've had this kind of broad exposure to a whole pile of emerging technology that you've worked with organizations to work on the R and D side of as well. So I'm curious, from your vantage point right now, Scott, what is most exciting to you in tech right now? Like, what's on your radar as you know, the technologies we should all be looking at or thinking about.
- [02:27]
Scott Likens
So, no doubt, no doubt, AI is, is the bulk of the conversation. And I think that finally has turned the corner in the sense of, at the executive level, at the board level, you know, really pervasively throughout an organization. Everyone understands AI is, is a lot more real. And you think about it's 70 or 80 years of research and development. So it's shouldn't have been that surprising. But I think we've been through so many cycles of promise, but not, you know, value. We're seeing that value and with that investment, it's also supercharged innovation across the board. If you think about some of the challenges of AI, it needs tremendous amounts of energy. So there's tons of innovation happening in energy and energy transmission and green energy, you know, really good things that are going to benefit beyond AI. But I mentioned a few. I, I think one of the big areas we're focused on is what's happening in quantum computing. Whether it's around cryptography or just quantum computing itself in the sense of how it will support AI and some of the needs ahead of us. So a lot of excitement for me and I think there's been a lot of progress just in the last two, three, four years in quantum. We've seen a lot of breakthroughs. I continue to monitor blockchain in the sense of how it's becoming pervasive as an infrastructure. Not in the crypto world and investment space. Stay away from that. Maybe. Personally I don't. But in an enterprise world, I think blockchain is really maturing in the sense of payments and tokenization of assets both digital and physical, and giving us a really strong foundation that we just didn't have before in this way. I love how AI is infusing into the creative areas, so in multimodal, so generating video and human image and likeness, even into holographic endpoints. And I think the next wave that we're really focused on right now is kind of the embodied AI. So the physical connection of AI to robotics and the Internet of things, which again continues to grow behind the scenes. So now infusing AI at the edge, embodying it into something physical, I think that's probably a wave that's going to hit us here maybe this calendar year in a big way, if not already.
- [04:37]
Interviewer
Yeah, no, that's super cool. And it's something, to be honest, I haven't had a lot of exposure to. So I'm curious, what are you starting to see emerge there? Are there specific use cases that people are talking about or investing in and kind of the physical side of AI? And I'm curious as well, like holographic AI That's a phrase I have to be honest, I've never heard before. What does that look like?
- [04:59]
Scott Likens
Well, I'll start with that. We've, we've created a synthetic version of me, synthetic Scott, almost two years ago. And using AI, I can capture my name, image and likeness. So I can capture myself now and really use AI to have someone interact with the synthetic version of me. And that can be done through a screen. So we can do that through video. But one of the more interesting advancements is holographic boxes. So I can have a three dimensional version of myself, sounds like me, looks like me, moves like me, it uses my hand movements, my facial movements, but also giving me the power to speak any language and interact in any language and have knowledge that the carbon based Scott doesn't have. Right. We can program through large language models, information about any topic. And not only that, but we have the foundational training of the large language models. So putting this holographic representation of me allows me to maybe go to a meeting without having to go. And I've done that and I've had a meeting in Korean, which I don't speak, but synthetic Scott does. So it's just maybe a different endpoint, tying together kind of converging the AI generation, the multimodal version of me, the voice match, the image and body match, the movement match. And really in the last two years that went from me as an avatar, so cartoony looking avatar, didn't really look like me, to 4k representation of me in live video. And it's really kind of changing that pattern of how people interact. So not just typing a prompt, just speaking to me in any language and having me respond, and it's not actually me. So that one's pretty interesting. I think it's kind of changing that human machine interface quite quickly. And if you.
- [06:45]
Interviewer
Stay on that for a minute, because I'm curious from a capabilities perspective, like, have we passed that uncanny valley? Like, do I need to be worrying about, like, am I talking to Scott right now or am I talking to an AI avatar of Scott? Like, is it that level of quality we're seeing?
- [07:02]
Scott Likens
Absolutely. I think we've passed that in the sense of a video screen for sure. There's technical ways to understand if it's generated or not. But to be honest, right now the human, it's beyond human comprehension. And there's some pretty public cases of teams meeting, you know, video meetings, where it's not actually the person because the voice can be mimicked, I think with three or five seconds of My voice, it. It gets very accurate. And with anything beyond that, it. It gets to a point where humans won't tell the difference. Now, machines can still tell the difference in the sense of using AI to detect AI, but I think it's going to be really hard to tell for humans. Now, when we get into holographic, you know, there's still a little bit of interaction delay. It's not specifically real time back and forth. But that, again, I think that's a solvable problem, and it'll get to the point where it's. It's really hard to tell whether it's me or not. So we're looking at, you know, can other technologies help us understand and create that trust factor? And maybe that's where blockchain comes in. We can authenticate and, and understand is this an approved version? And there's, there's another layer to it, so I can capture my image and what I look like and speak like. So once I've done that, my team can have me say whatever they want. So it's not only the image, it's actually the content. So we have to actually authenticate is the content approved and the way I deliver the content. So a pretty interesting and difficult problem to solve.
- [08:26]
Interviewer
Well, that authentication piece is really interesting as well, and it seems like something we need to fast follow on. Right. Because now we're in the realm of deepfakes, it sounds like. Right. Like, is this an approved Scott, or is this, you know, a nefarious Scott who's trying to get me to sign over, you know, a million bucks for, you know, God knows what?
- [08:45]
Scott Likens
That's right. And this has come up, obviously, in Hollywood and IP protections around image and likeness. So there's a lot of people thinking about how to not only secure it, but then is there a commercial model around it? Because now I can get out to more places than I could physically. Right. I don't have to fly. I can save travel and carbon. There's. There's opportunities with it if it's done responsibly. So that's, that's really where our focus has been, is that responsible use of the technology.
- [09:13]
Interviewer
Right. And responsible is a word that comes up in a lot of the, you know, that comes up in a lot of the, you know, the presentations that you've been giving. And I'm curious, like, what's your, what's your worry level with some of these technologies versus excitement levels? Because, you know, we talk about nefarious uses, but even you mentioned Hollywood, you mentioned IP as well. How concerned do we need to be that are we empowering people here or are we potentially robbing people of their agency in some capacity?
- [09:45]
Scott Likens
I'm probably the ultimate optimist, so I think I'm excited. In general, I think there's a lot that we can do now to put the guardrails in place. And I've said this before with, with the boom of gen. It's the first time I've seen the word responsible or trust built in early in the conversation versus let's figure that out later. So I do think I'm hopeful a lot of people have taken responsible AI as a fundamental pillar to everything they're building. There's definitely a worry that nefarious people will use this technology in a bad way. But I think we've seen that throughout the decades of every technology, and it's always been kind of a constant battle. So as long as we're all aligned in the beginning that we have to do this responsibly, we're implementing those things because they're the right things to do. I think we're in a better place than we typically are. But there is a lot of work to do. There's unsolved problems in the sense of how fast the technology is moving and how fast can we make sure it's done responsibly. There's a lot of intellectual property rights questions out there that haven't been answered either by regulation or by legal court cases. We've seen some, but there's still a lot of gray area. So not solved. But I'm hopeful that people want to do it the right way and that we've all started thinking about that in the beginning, which is different, a different pattern than other technologies. I said in the early days of the Internet, we weren't thinking about securing it, we were thinking about growing it, connecting it. And if you think about the E commerce boom, really the pivot point to that was kind of this, this escrow version. If I'm going to buy something, someone's going to escrow my money until I get the product because I didn't trust it. Right. And that was an add on. And then we started to build in more controls and security. So now E commerce, everyone trusts. But in the early days, no one did. Right. You didn't know if you're going to get your product. So I do think AI's generative AI wave has been different, but still a lot of hard problems to solve.
- [11:44]
Interviewer
Yeah, yeah. And you know, it's one of these things where you Know, my sense is, and I'm curious if you're seeing the same thing, Scott. Like people are struggling more and more to keep up with like today's gen technology, let alone like what's happening literally every week that comes out. And so it's exciting and I'm, I'm, I'm really glad to hear you say that. Like responsibility is starting to be baked in and I love, I love that optimism. Right. Because it's fair. You're right. That hasn't always been the case. So that is a nice kind of glasses, half full approach.
- [12:18]
Scott Likens
Yeah. I think the pace is the problem right now. And there's a great author, Rachel Botsman, who said speed can be the enemy of trust. As things move faster, it's hard to trust them because the ground is moving underneath us. And AI for the last three years has been exactly that. It is moving faster. There's more innovation happening than I've seen really in my career, which is a long time. So it's exhausting to keep up. But there's also this excitement around this innovation and the invention that's happening. And in general we're going the right way. You see people thinking about the mission and ethics of things and there's obviously some examples where they're not. But in general, the whole industry, the whole set of innovation is, is really trying to be accretive to what we're doing in business, in society, for ourselves personally. And I think if we stay focused, we stay in alignment, we need, we need some regulation and guidance and guardrails. But in general, the technology, even though it's moving fast, I think it's trajectory is good and it's positive and we're seeing benefits across the board.
- [13:21]
Interviewer
Right. So you talk about speed and how fast this is all moving and generally moving in the right direction. You know, I wanted to throw at you a quote that I got I think from a colleague of yours, Dan Priest, something like that. But, but he had said, you know, speed is more important, scale is less important, innovation is the most important. Can you kind of, you know, unpack that and what it means for people who are, you know, participating in this race?
- [13:48]
Scott Likens
Yeah, I think it's trying to call up this, this pattern is different than we've seen. Again, if we proxy back to the Internet or E commerce or social media, it, the cloud, big data, they followed these similar patterns where there was some innovation, but then it was all about scaling it to get to be the biggest and to own everything. And we saw that, of course, with the social Media, all kind of conglomerating into central apps. With AI, the innovation keeps you ahead. And that's hard sometimes. For big enterprises, innovation is something they say they invest in, but it's really not the core of the business strategy. And I think because this new generation of AI really does affect every aspect of an organization, you have to stay maybe not at the edge of innovation, but you have to keep up with it. So that's where speed means how fast we can change our organization, our workforce. How can we skill our people at a very different rate? If we give them a tool today, it could be different in one week or two weeks. And that's just not a motion that most organizations understand. So we think that the innovation that gives you that roadmap, that view of what's coming next. And again, you don't have to do every single thing, but you have to understand the trajectory it's moving in. Especially with generative AI, there's been really amazing inventions that can be infused right back into the organization. That's where speed matters. And we're a big organization, 150 or so countries, 375,000 people, that's a hard workforce to change overnight. But for us, it was, we saw the benefits. So moving fast with our people, giving them access to that new innovative technology, and then we got the scale. So it's kind of that pivot around scale will happen, right versus you back.
- [15:36]
Interviewer
Into the scale versus front loaded.
- [15:38]
Scott Likens
That's right. That's right. As you start to get the business to understand where we can fundamentally reinvent processes that maybe have been around for decades or hundreds of years, even depending on the industry reinventing those are hard. But if you're innovating at speed, the scale will happen. So it's a little bit of a different pattern than we've seen in big technology revolutions.
- [15:59]
Interviewer
Right. Can we unpack that word innovation a little bit? Yeah, it's a word that gets thrown around a lot, and I'm certainly guilty of that myself. When we're talking about innovation here, what does that mean in the context of an organization or leaders? What are we doing there? What's kind of the outcome or the process?
- [16:20]
Scott Likens
So really it depends on where you sit in an organization, because there's innovation happening. As the engineer, for me, innovation is technical invention. So there's invention in AI and the innovation on the back end of that is how I apply that into the business to actually change something fundamentally. So a lot of the times technology is an incremental improvement. I'm going to get 5 or 10% better, you know, efficiency based on this technology. To me, innovation is not doing any of the patterns we did before, but getting a better output. And that's where we really work with our clients. To say the innovation in a process is reinventing it. If you did 10 steps yesterday, doing those faster is not innovation. Do two steps today with AI filling in the gaps and getting a better output, that's innovation. So there's business model reinvention, which could be innovation. There's technical invention which could drive innovation. So it's really a combination of how we apply that into a business. That's where we see the benefit and the results. To me, that's the big eye, Innovation.
- [17:24]
Interviewer
Right.
- [17:25]
Scott Likens
Small I innovation is more invention. There's pieces we're inventing that didn't exist or we're using inventions out in the market, in open source or with our partners. The innovation happens when we're changing something fundamental to the business. We're getting to market differently, we're creating products differently fundamentally. And you see the business impact happens then.
- [17:46]
Interviewer
Got it, got it. And I understand, Scott, in your role, you mentioned engineering, this is sort of what your team lives and breathes. Right. You guys are, I don't know if I'm allowed to call it a lab, but you've got kind of your, you know, innovation, emerging technologies kind of engineering group. Can you give me a little bit of a peek behind the curtain? Like what does that look like in your world and what are some of the things you're working on either internally or with other organizations?
- [18:14]
Scott Likens
Sure. If I think about the process we take, we think in Horizons. So Horizon 1 is taking some emerging technology and applying it into the world with ourselves or our clients that we could make an impact today. It's still probably the first time it's been done, but it's a, it's a tangible idea and a tangible technology. So implementing AI agents is Horizon one. We know we can, we can do agents today. We can have multi agent build outs. Those agents can work together. They can supply some opportunity to a business process or efficiencies within the business, etc. Horizon 2 is trying to understand what's happening in universities or in research labs or our own lab building something that probably isn't finalized. And that's where I'd say some of the quantum work we're doing is in Horizon too. We know that we can start to do AI optimization better with quantum. Now, it's still early because there's not a scaled quantum platform. There's some options, but at the enterprise level it's still not ready. So that's horizon two. Horizon three is looking further out and that's maybe an area we're looking at what's happening in space technology and communications and things that may not apply to the enterprise in the next year or two. But we know that that's going to spawn invention and innovation that we can use. So trying to understand how to apply maybe what seems irrelevant, making it more relevant to enterprise business. So we think in those patterns and then within my team we have deep research and development done in different areas. A lot of the times it is skill based. So if we're in quantum computing, you have to have quantum understanding. There's specific skills and math that you have to understand. If we're talking about satellites, that's a different skill. That's how do we actually build a physical satellite and sensors. So we start to kind of pattern the skills with the horizons and the art and magic of it is how do you then bring that into an enterprise relevancy. So where would this help an enterprise? A big insurance company or a retailer or oil and gas company. So we have to have that sector lens. So we have a many to many funnel in the sense of what we do in our work. We work very agile. So we have very small pods of people focused on delivering in a week or two weeks, not long drawn out research projects. Because we have to understand if there's going to be value there and we want to then go scale that. The other aspect is we kind of have these incubators. So once we get to a point where we see there's some value, we've tested it with a client or two, we want to incubate that into a business that we can bring out and scale to many more clients. That then takes traditional skills, the change management and the program management. The skills that would mean we could take this to a lot of clients and help them find value with this new idea. And that's the hardest part. You're selling something that didn't exist before. And you know, I think about going back to blockchain. You know, it was probably two years of experimentation and working with clients on ideas that really didn't exist in the market. And now it's exploded and there's timing of that with the administration change in the US and what's happening with the price of crypto. Like there's a lot of reasons, now is the time, but now it's exploding. You were seeing a lot of interest and that's where that incubator just starts to scale.
- [21:49]
Interviewer
Right. Well and I expect in your world to Scott, like the hit rate at this stage of the game just cannot be 100% right. Like it's always going to be fractional. There's going to be wins and losses and that's kind of a built in model of what you're doing.
- [22:01]
Scott Likens
Well, it's interesting what we find is when things don't hit it's, it's likely because they're just ahead of their time. And I, and I look back and we pull things off the shelf from four years ago that was just pure research and now is ready because there was maybe some dependencies in the data we needed or the compute we needed or the physical hardware we needed. But now is the time. So it's, it's almost like we, we look at our backlog from previous years and say oh wow, this one let's, let's pull it back out. It's ready. So we don't throw anything away because we've learned something from it. But sometimes it's timing. Is the market ready?
- [22:37]
Interviewer
Right. Are there any of those that you've pulled off the shelf in the past 12 months or so that any technologies whose time has finally come or is coming?
- [22:47]
Scott Likens
One that comes to mind is digital twins, which is a concept that have been around for a long time. They were always really bespoke and one off and now with generative AI helping us interface into digital twins with the amount of data we have coming from IoT devices. So the, the data now is there from physical, the physical world like digital twins now is one of those that we've, we've researched systems dynamics models for, for a decade and now you're seeing real value, you know, in enterprise usage of those. So that's one. A lot of the AI work is, is, had been done in the research phase and maybe that was a Horizon 3 and now it's really accelerated.
- [23:31]
Interviewer
Cool. The digital twin piece, we're kind of now backing into it from the other side. But we were talking earlier about these kind of physical manifestations of AI. Can we come back to that conversation for a little bit? What are we starting to see there emerge in terms of capabilities and maybe also as use cases.
- [23:49]
Scott Likens
So the one I love is bipedal robots. So robots that stand and walk like humans. And you've seen an explosion of bipedal robots in the last two or three years and in 10 years ago there was some very well known labs that were building one and you could See some videos. But now there's an explosion and part of that is the generative AI interface, the natural interface in the sense of how humans interact with them. So we can speak in natural language, but also voice to action. They're able to now use generative AI to have those, the robots learn. And you're seeing an explosion of intelligent moving robots that applies into the drone space, which in the US at least had really softened over the last five years. We didn't hear a lot about drones, there was regulatory and licensing issues, but now indoor drones, land based drones, those kind of to me are all the same thing. We're embodying AI. We're using the power of both generative and good old AI together in a physical instance. And, and that's where it becomes a lot more real. You know, we can fundamentally change manufacturing and warehousing and logistics. Self driving cars are embodied AI. They have AI happening all throughout them. You're starting to see a lot in the defense tech world. So not an area that you know, we do a ton of work in, but just the amount of invention happening there. So from a drones perspective, both in water, on land and in air. So that is all on the backs of this AI explosion and generative AI and just the reasoning and thinking that these models can do has really advanced what's happening in embodied AI. And then in a more micro level, the IoT world having edge AI embedded. So not just a sensor, that's kind of dumb, just sends a reading. Now that sensor can be much more intelligent and actually make decisions or take action. So now we can instrument the world in a different way. Smart cities and smart districts, of course, smart cars, they're not only generating data, but they're actually taking action. And that's because of AI. And we call it convergence, the convergence of AI and robotics and IoT and maybe blockchain to authenticate or make payments as things are happening. So that's where the real value comes in, is when we converge these technologies.
- [26:09]
Interviewer
Right, and are you starting to see some of those convergences in your, like at that horizon one level in your lab? Is that still Horizon two?
- [26:16]
Scott Likens
No, absolutely. I'd say the first wave for us was in the IOT world where we're seeing the ability to have much more intelligent sensors which then create better data or take action which then we could build into an optimization of energy, of water, of people. You know, they absolutely been rolling that out. The bipedal robots still probably Horizon two, you know, there's experiments, those still seem to be A little bit out, maybe a year or two for scale, but there's some individual use cases you're seeing. And then I think Horizon 3 is this space and what's happening beyond just the world that we walk around in. And it's a broader connection from a communications perspective, from a sensors perspective, et cetera.
- [27:03]
Interviewer
Right. The space stuff is interesting and it's got me thinking about the 20th century in general. And you know, I'm not an expert in this area, but my sense is there's been a lot of use cases where, you know, DARPA or the military or, you know, these public R and D institutions came up with these advanced technologies and then they were kind of brought into more, you know, if I can call it civilian or commercial purposes. Is, is that still a trajectory with seeing or is it almost the reverse now where it's more kind of commercialized R and D labs and then like where is the space tech coming from? I guess.
- [27:40]
Scott Likens
Well, I think there's a lot more private industry that is, you know, inventing things related to space than we had before. It was, it was really driven by, you know, governments and the, you know, the investment that they had to have. But now with the lowered cost of, because of AI, the lowered cost of product development and you know, engineering, there's a lot more private industry. And I think it's amazing, you know, and there's some great examples, without naming names, you know, many private companies launching rockets at a pace that governments can't even come close to. And you know, that, that competition I think will increase. It will, it will be to the benefit of governments because they're going to have to keep up. And it's to the benefit of us because it's technology we all can see and understand and start to figure out ways to proxy that into enterprises. So I think it's amazing and I think it's just an example of how that innovation will lift everyone together, lift, see what I did there? Into space. But the low orbit satellites and the communications and sensors gives us more data and it's just kind of a self fulfilling cycle where more data equals better AI and better AI equals better products and better ways to get more data. So it's really starting this flywheel that we think is amazing for business.
- [29:03]
Interviewer
No, that's awesome. And I mean Horizon 3 is such an exciting space for just trying to figure out what's next and what's next after next. In terms of Back to Horizon 1 for a minute, are there any projects you can talk about either that your team has done in the past year that are ongoing. Now, that might surprise people that technology is kind of here or that you're excited about. That may not be on people's radar.
- [29:32]
Scott Likens
It might be on the radar, but just to give an example of how real it is, the agentic AI era that we're living in, this was Horizon one for us last year. Even though it probably wasn't a term people were using, there were not any, if only a few, agentic frameworks. But we started in the world of software development knowing that agents could help engineers just build better technology. And now we're at the point where we can scale agents with our engineers not only to generate code, but to look at regulatory obligations, to look at, obviously security obligations. Can. Are we meeting all of our safety and responsible AI requirements and empowering our developers to put out output at 2.3,5x of what they were doing? So for us, it's been a huge enabler. And those AI agents just a year ago were probably not on anyone's radar. And now we're infusing them into everything we do and thinking about fundamentally transforming legacy technology, migration, looking at old mainframe software and languages nobody knows, being able to use AI and agents to unravel something that's 40 or 50 years old and it's been running perfectly, modernize it and maybe not have to have humans recode that, have the agents do that for us in a way that we trust it, that we have a comfort on the output. So that's one that was last year, probably Horizon two, and suddenly, boom, it's here, it's at scale. We're doing this across many other functional areas beyond software development, thinking about customer service and finance and operations and human capital and starting to really scale that out. We launched something called Agent os, which is our ability to work across all the agent frameworks in any cloud in a much simpler way, so that agents can. Can work autonomously together. That's rapidly happened. And I just think that's an example of how fast these horizons could move when the innovation is understood and scaled. And that's. That's happening throughout almost every sector and every function.
- [31:47]
Interviewer
And I've got to compliment you for a second on the branding of Agent os, because that's, yeah, beautiful, beautiful branding. And right away you're like, oh, I get it, that's cool. Why doesn't that exist? How ready for primetime is that right now? Is this something that people are already implementing and the legacy modernization piece? I don't know. To me, it's quite Interesting because it's such a gnarly, wicked problem. Maybe some other people think it's boring, but are we starting to see people actually implement this or is it still talk right now?
- [32:17]
Scott Likens
No. So first agent OS is ready at scale. So we take a kind of an approach that if we can drink our own champagne, use it ourselves, we're very complex organization. This is something we had to build out of necessity. To your point, it didn't exist. So thinking about how agents across frameworks can speak in a standard way, share memory, all these things that didn't exist, we built that, but we rolled it out internally. We've been on this journey over the last 18 months for our own AI chassis. Again, 100 countries at scale has to be safe, secure, protected, confidential data, et cetera, et cetera. So once we get through that pattern, then it's ready for our clients. And that, that is the point we're at. We're, we're working with many clients. This is something that we can, we can bring in, accelerate in a week or two, have agents up and running and start to see business value so that, that's ready. Legacy modernization, you know, obviously is, is a huge area of cost for clients, especially industries that have been running on mainframes for a long time, or maybe they even transition off that about a year, maybe a year and a half ago, I said, you know, I could see a world where we don't even have to move the system with a world of agents. Could we keep the system isolated? And if I think about specific, like life insurance, where that Policy runs for 30, 40 years and has to, you know, you have to trail those off. Could we just isolate it, have an agent that's intelligent enough to speak to the old system and the new system? Could we build a new system and not move the old system? Do we actually have to modernize it? And I think we're starting to see people look at different approaches. One is, we're seeing right now on the legacy modernization, can we use AI to actually unravel what's there? You know, you're not going to find the original coders. It's maybe in a language, you know, of course cobol, but maybe in assembly. Like there's languages that are really tough to find experts. Well, AI can help us unravel that and create, you know, the new version of that, maybe not even in code. Maybe it just gives us the agile stories and then from there we go code it. But if we could do that in a day or two or a week, that's a huge difference from the old approach. So we're definitely seeing that. Let's go after those legacy systems. Can we extract what's there using AI and then build it new? That's one approach. I think another approach is can we leave it and can we create an agent that's intelligent enough to interact with the old system and the new system and bridge that? And I think we'll start to see some of those patterns. That's not at scale, but that's one of my hypotheses around where agents can really play a strong role.
- [34:49]
Interviewer
No, it's super, super exciting. And I'm sure, as you know, a lot of these legacy mainframe systems, they're limiting the pace of innovation for organizations. Right. It limits the speed of effectiveness, it impacts what they can do. And so for organizations who want to go down this road and maybe more broadly, organizations that are looking to implement agentic AI across their organization, what's the role you see corporate IT playing here? Do they need to be leading this? Is this happening around them? Is it replacing them? How do they fit into the puzzle?
- [35:28]
Scott Likens
Well, of course we'd love to help those clients, but I think part of the challenge is corporate IT has to do this upskilling that, you know, of course we've lived through. They have to adapt the pattern and the way they approach these problems. We can't just go after it the old maybe waterfall way or hybrid agile way, where we're just creating a two and three year transition. I think they have to change that mindset and build those new muscles around how these tools can help accelerate that. We have a tool called code intelligence that'll go in there and extract all of the way the code works and generate the new stories, it'll generate all the documentation, create a little website that that's something that the corporate IT has to kind of change the mindset. And some of that is the building those muscles around what gen AI can do, what it can't do. You know, it doesn't do everything where it can play a role and be more efficient and helping their team actually have much more output.
- [36:20]
Interviewer
Right.
- [36:20]
Scott Likens
So I think that's the transition. It's more of this human problem. The technology is available, there's vendors like us that'll help you, but you have to change the mindset. I think it also affects budgeting and strategic planning. You know, if I see a five year plan on technology, I just shake my head and say, how do you know what's happening in five years with, you know, the pace of, of change today. Not that you shouldn't have a kind of a target vision, but the reality of technology is, is going to be very different every year versus maybe every five years before. So I, I think corporate its to adjust that, that pace that they're working at and the way they, they bucket these activities and just change, change the skill set as they go along.
- [37:00]
Interviewer
So when you look at the, when you look at the IT organization of the future, you don't see it being, you know, replaced or displaced. You, you see a different skill set than we have today, I think.
- [37:11]
Scott Likens
Yeah, I see it adapting. I, I see, I'm already seeing a little bit of a change in the traditional organizational models in, in the sense of moving to much more. You know, you call it whatever you want, but pod structures of multi disciplinary technology and business together. So I, I think the org model changes a little bit and the skills have to change because you have to take advantage of, of, of AI of course. But no, the, the role is there and, and that's, you know, we've been studying jobs. We'll publish our second version of the AI jobs barometer here I think in May. And what it shows is that it's not taking jobs away, it's changing the tasks in those jobs and it's changing them faster than we thought. But, but it's, the jobs are there, they're just different. And the people that are adapting and using AI, they're actually making more money. So there's a good story there. And it's creating new jobs of course, and you know, it's been accretive so we haven't seen this drop off. There's some jobs that have gone away, but in general it's just what those job titles do versus the job title going away. So I see that throughout IT and other areas of the business.
- [38:17]
Interviewer
So to put a little bit of a spin on that, and I'm curious first of all, if you'll agree with this or not. But what I'm hearing and what I'm seeing firsthand is that I mentioned this earlier, we have these capabilities now that seem to be in many ways far outpacing organizational abilities to adopt them. And I mean just to say it bluntly, it seems like organizations are just not going fast enough to be able to embrace these technologies. And so I mean, I wanted to ask you like what's going wrong out there and what do organizations need to be doing differently or what do they need to be like, what's the mindset piece? What do they need to be doing what do they need to be thinking about, you know, to better get ahead here?
- [38:57]
Scott Likens
I agree 100%. I struggle even with my own teams, I don't go faster. And it's a really a human problem. I say biology doesn't move as fast as technology. You know, we have to a cultural change in an organization and each, each org is different. There has to be strong leadership from the top about the pace. And that means you're breaking the budget cycles, you're breaking the planning cycles, you're breaking strategic, you know, organizational cycles. That has to come from the top. But I'll tell you, over the last two years I, I've been in front of more board of directors and more C C level across the board about this topic. So I think that's been solved. And, and I think at the lower levels and people coming in out of college, they're using AI. It's kind of this middle level that they look at it as a threat. And I think that's just the wrong way to look at it. If you lean in, you say this now gives me a superpower. No matter what I'm doing, if I can enable my teams to do more, it's not about getting rid of our people, it's about producing more products, it's about getting to more markets, it's about being more efficient than my competitor. That's all upside. So I think we have this weird problem where the top and the bottom really align. That this is a good thing in the middle is saying, whoa, whoa, this changes my world. I've been doing this for 10 years, 12 years. Yes. And it's going to change and we'd like you to change with it because it's a better outcome for everybody. So it's a human problem. And I think how you communicate that, how you create the change management around that is really important. And I do feel like it's a little slow. And we've seen some of our clients really just over the last year change and go fast, ironically a lot in the regulated industries because I think they see so much opportunity and they've been really doubling down on the investment. A lot of it's about workforce training and the communications around it that this is, this is a good thing in, in general for everybody and giving them the tools and skills. Now that the challenge is, the training changes all the time. You know, it's, it's like how many tools do you have? And as we, we roll out our own tools, we struggle with how do we communicate this. It changed Again, I just gotta say we gotta tell them that that's the new world. It's gonna change every month. It's not something one and done. These tools are gonna get better as the AI gets better and you gotta be curious and you've gotta lean in and you gotta have that passion to change with it. But it's a tough problem.
- [41:19]
Interviewer
You said something there that caught my attention, which is, you know, you alluded to we give them the tools. You know, to what degree are you recommending that? This is kind of centralized and it's, you know, here's the tools. Like I don't know, the image that came to mind was still like, and obviously this is very old fashioned but you know, Moses coming down from the mountain with like the tablets like AI versus you know, it being kind of grassroots, people led, employees at the bottom saying these are the tools we found. This is what's enabling us. Is it one, is it both?
- [41:53]
Scott Likens
It is a bit of both. But I think in the world of Gen specifically we have a unique opportunity at the end of the day, one foundational model, whichever one you use, can serve the whole organization. So we have a point of entry to that technology versus the previous world of technology. We had a lot of options, there was a lot of competition and anyone could install and use it in our however they wanted within their business unit, machine learning, et cetera. Now we have one way that we have to make sure it's done responsibly, it's secure, it's private. So we can't let people use public tools because that means data could leak and it's hard to understand. So there's some business strategy around. We wanted to provide the best tools and we did. We've provided cutting edge tools at, at pace, but we didn't want to have a hundred tools because then it's, it's impossible to manage the, the risk. So it is somewhat of a centralized enablement. But that tool is so flexible that it, it's not constraining. It can help a tax accountant, it can help a consulting strategy person, it can help someone doing an audit. In our business really that same tool is, is so generally intelligent that there's no constraints. But to us it was minimizing the risk, giving them enough innovation. And the bottoms up is about how I use it, not about which tool is better. If I'm this language model or this language model, don't worry about that. How are you going to change what you do every day? That's where we want the innovation. So we Open that up. If you want to build something custom using the standard architecture, go crazy and let's share that with, with your colleagues. So we want that innovation around how to use it, but not necessarily this tool versus that tool. That's what my team can do and we'll evaluate them in a very deep way against not only our business rules, but the innovation. And we feel like we've kept a good balance there. There's other strategies but I do think you have to have some controls on it.
- [43:54]
Interviewer
So you provide the sandbox. But they can build whatever they want.
- [43:59]
Scott Likens
That's exactly right. That's exactly right. Yeah.
- [44:02]
Interviewer
Yeah, I love that. I imagine there is a tension there though for your organization or for others where if you're not there fast enough, you know, it's the same shadow IT problem now, you know, a shadow AI problem. If you're not there fast enough, you're, you're kind of racing them as they come up with new use cases. And I don't know, maybe that's a good thing or maybe it's a bad thing, but at least it's a challenge.
- [44:21]
Scott Likens
Yeah, no, it has, it has pushed us to be very fast. When new models are released, we have them same day or next day because if we wait four, six weeks, then we do get frustration, we do get people trying to find a way around it. So I think it's made us better and increased our pace of delivery to give them the best thing to then go innovate on. But you got to be in the sandbox. We've got to protect our own internal ip, we got to protect our clients data, we cannot take any shortcuts there. So while I want to go fast, I can't take any risk in certain areas of our business. It's, it's, it's just non negotiable. So it is a bit of a tension but I think we still get great, you know, ground level innovation within the, the boundaries. The guardrails, it's not a controlled environment, it's just guardrails like stay within these, do whatever you want and work with us to scale it.
- [45:13]
Interviewer
Got it. There's, there's another word I wanted to unpack here and this comes up all the time by the way, which is tools. Right, the tools we're talking about, about Scott, you mentioned kind of broad general applicability. Are we talking about enterprise versions of things like Copilot, GPT? Are we talking about you know, more industry and sector specific tools or maybe even, you know, business function specific tools? What Tools do you consider to be, you know, part of your toolkit here?
- [45:42]
Scott Likens
So I'd say last year was, was mostly about those enterprise enablement. We've created Chat PwC which is our, our customer private version. It has access to OpenAI models, to Gemini models, to cloud models if you're the right kind of level, Llama models and other open source models. So for us that was, you have Copilot in Microsoft, of course, and other tools where that's to me a general use. It's, you know, using it on a document or a meeting. Chat PwC was something we built to be general PwC use. We're embedding our data, our processes. We're building agents that do things specific to what we do as PwC. No matter where you're at in the business, that was valuable. And then you had these enterprise versions which were secure, but they were still general. They don't have your data necessarily. So this year it's moving much more into the world of agents, which is where you get more domain specific and there's probably some areas of fine tuning or small models in certain domains. But in general it's, it's more agent specific to a domain and, or a function within an industry. So there's very interesting data in this area. We want to create an agent that knows everything about that data. It's not more generally usable, but it's going to be very precise. So we are starting to see, and we're building dozens of those now to say this would be used in this specific instance and it's going to be very good and we'll continue to make it better. So don't use a general model. Use a specific agent that was using a general model behind the scenes, but in a different way so that we get the precision we need within areas of our business. And then tools, to your point, is a very broad term in the world of agents. Tools mean something. I'm calling a tool which is an API to go do something versus a large language model is doing the planning and reasoning around calling tools which in effect are APIs to a piece of software or code or another model. So there, you know, it's an interesting term right now because of the way people are using it.
- [47:48]
Interviewer
Got it. So you know, to zoom out just a little bit, you know, as you talk to leaders who want to implement, whether it's, you know, agentic AI or any of these kind of Horizon one technologies, what's kind of your best advice for what they should be thinking about or what they should be doing to break the cycle of slow and actually get to success.
- [48:11]
Scott Likens
So I always say waiting is not a good strategy in the world of AI right now. And last year there was a lot of wait to see which model will win. I said that is a bad strategy. So I say waiting is not a great starting point. So some experimentation. My big advice is the workforce investment you need to make. So I believe you're going to get benefits. No matter what large language model you use or what hyperscaler you use, you're going to see opportunity. We're probably barely tapping the actual benefit available. It's all about the investment in that workforce. What are the skills you need? I personally have changed the team to be AI engineers. It's a combination of data science and engineering which typically would have in different colors, groups. Now I want everyone to have some data science and some full stack development because everything we're doing is kind of a combination. Well, how do you do that in your organization? It could look different, but the skills you need to deliver at pace to look at your funding and road mapping models. How are you effectively managing this and then that overall organizational design. Is there a central group of experts? Is it a hub and spoke? Couple different models depending on the organization. But you have to start looking at that workforce transformation to enable the upside. Otherwise you're just going to get individual efficiencies here and there. They're hard to capture. You're not going to see the big return on investment. The bottom line won't be impacted. You have to go in, pick a business that's willing and has opportunity and team them up with a new way to operate the technology transformation. And that's where you're just going to see explosives explosion of return.
- [49:46]
Interviewer
I agree with you. I think that's very much in line with what I'm hearing and what I've seen in practice. The workforce transformation piece, just to kind of put a finer point on it, is this largely about upskilling current staff? Is it about procuring external talent either full time or working with vendors or suppliers or some combination of the above.
- [50:10]
Scott Likens
As the ultimate optimist? I think it's a combination. I think the. There's plenty of great people that are, that want to be upskilled. They want to make this transition and you should, you know, go after them first. There's kind of the. I forget the percentages, but the zealots, they're all in. There's kind of the big middle that's got to be convinced and then there's ones that never going to change. So the ones that are never going to change find different stuff for them to do. Convince the ones in the middle, use the, the zealots or the passionate to, to help them. I think outside talent always helps kind of be a catalyst to change. And with the speed of delivery, I think you need a little bit of that. Whether it's a partner or vendor or hiring in some of this new mindset. We're seeing that quite a bit working with very traditional companies who are in fact one client said can you help me write this? So it sounds like a Google or an Amazon job wreck. And it's a very traditional industry because they want to attract that different type of person. So I do think there's some kind of catalyst to bringing in the new mindset. But I think you have to focus on the core. Your, your team has been delivering something good or else you wouldn't be in business. And taking them through that transition makes so much sense, but it is, it's difficult. You have to plan for it, you have to invest in it and you have to change things that maybe have existed for decades. And that, that's the tougher part that cultural change. Cool.
- [51:28]
Interviewer
That's. Yeah, it's. It, it's. It always comes back to the human piece. Right. As much as we talk about the technology to get it to work, it seems it's the people. The people, the people. Scott, I wanted to ask you a completely different question and you may find this one a little bit challenging given your moniker as ultimate optimist. But I'm curious in your mind what technologies are overhyped right now? And when I say overhyped in your language, that may mean what are probably closer to Horizon 3's or Deep Horizon 2s that are being marketed right now as kind of the next big thing.
- [52:08]
Scott Likens
Ironically, AI agents I think are overhyped most, to be blunt, and I'm not going to call out any specifics, but I think people are falling in the trap of creating agents like we did rpa, where they're creating agents in a brittle way that is just a step by step workflow. And the true power of agents is autonomy and action. And I think that's a ways off, that's challenging to trust an AI to be autonomous and take action. So are you going to let purely AI commit that transaction to close that piece of business? I think we're a ways off from that. And people are selling it as agents are the answer to everything. And I don't think they are Right now. So I think that's one. I don't know, I don't know if any other are necessarily oversold at this point because we, we've, we've seen the quantifiable. A lot of the AI stuff I think is pretty quantifiable. Some would say quantum's over, over, over hyped currently. But I don't know. I, again I see that the trend line on that is very different. So people used to think it's 20 years away and now some would say it's two and you know, it's somewhere in the middle there. But I don't know think that's necessarily overhyped but maybe under or understood misunderstood in the sense of, of how to get there. Different. Very, very different approaches.
- [53:32]
Interviewer
Right. Do you have a, do you have a prediction for, you know, Q day or when we get to, you know, the need for post quantum cryptography?
- [53:41]
Scott Likens
I don't, I think it's closer than people think, depending on what you mean by Q day. I think the one we were really tracking aggressively is the post quantum cryptography. Yeah, post PQC in the sense of when, when quantum could be a real threat to, to encryption. And I think, you know, that was probably a 2035 date for many people and I think that's moved in. I think that's moved in. I think more and more people are starting to say that has moved in and maybe not necessarily just because of the, the, the Q day is there, but maybe quantum approaches to optimization could cause a new attack vector for cryptography and we're starting to see some trend lines there. Could quantum inference help AI inferencing which could maybe create another attack. So maybe there's different attack vectors for cryptography that could move that post quantum cryptography day in. So we're tracking that quite aggressively.
- [54:35]
Interviewer
Yeah. So if I understand it, there's just some convergences happening here that we can't quite parse out what the impact is going to be. But it feels like, it feels like they're starting to come together in a big way and take off when it's.
- [54:49]
Scott Likens
Hard because there's, there's much more nation state investment in quantum and we're not going to ever really know what nation states are doing. They're not going to share that. So it's hard, hard to judge based on just university and enterprise research.
- [55:04]
Interviewer
I really appreciate it. This has been a really, there's been an awesome conversation. I feel like we covered a ton of ground, a lot of really interesting stuff and it's nice to hear kind of what you guys are doing in your own group and with your customers. You've got a really unique viewpoint on it.