Subscribe
Share
Share
Embed
What does it really take to scale AI across the enterprise? How do leaders move from ambition to impact? In Winning with AI, world-renowned AI pioneer Andrew Ng and Bain partner and digital re-invention expert Sarah Elk host candid conversations with global CEOs on the frontlines of AI transformation.
Together, they explore the opportunities and obstacles of deploying AI at scale: reimagining operating models, empowering teams, and leading culture change to prepare their organizations for the future.
Winning with AI helps leaders of global organizations navigate times of radical change with clarity and vision, uncovering the very human challenges – and opportunities – that transformation brings.
About the hosts:
Dr. Andrew Ng is an internationally acclaimed thought leader in artificial intelligence. He founded DeepLearning.AI, cofounded Coursera, and serves as an adjunct professor at Stanford University’s department of computer science. He was named one of the Time’s 100 Most Influential People in AI in 2023.
Sarah Elk is Bain & Company’s AI practice leader in the Americas, and a leading voice on digital reinvention. She partners with CEOs and senior executives to reimagine how their organizations operate – aligning strategy, operating models, and culture to turn AI into sustained competitive advantage. Sarah is the co-author of Doing Agile Right (Harvard Business Review Press), and her perspectives on leadership and transformation have been featured in the Harvard Business Review and Forbes.
Bain & Company:
Winning with AI is brought to you by Bain & Company, a global consultancy trusted by the world’s most influential business leaders. With decades of experience guiding organizations through growth, transformation, and leadership development, Bain’s executive insights offer what it takes to lead at scale.
Robert Smith:
The economic benefit of what we are seeing is orders of magnitude higher than the economic benefit that we saw going from on-prem to cloud, so it's worth taking that risk.
Sarah Elk:
Robert Smith is the founder of Vista Equity Partners, a global investment firm that focuses on enterprise software and tech-enabled companies.
Andrew Ng:
And through his vantage point, he uniquely has a view on what many businesses are doing on AI. And so I'm really looking forward to learning and hearing about this broad view and what lessons apply to all of our businesses.
Sarah Elk:
I'm Sarah Elk with Andrew Ng, and this is Winning with AI. Welcome, Robert. Thank you so much for spending time with us today.
Robert Smith:
My pleasure. Thanks for inviting me. Good to see you both.
Sarah Elk:
I would love to start with just a very open question on, when did your AI journey begin?
Robert Smith:
Interesting. So, our AI journey actually began over a dozen years ago in terms of how we were utilizing it for certain of our businesses to do things, like in our sports business for instance, how do we now take and strip what I'll call dated information, effectively from what happens actually on a field? To putting it into a data set can then be repurposed for customer sets? And that evolved quite naturally as we started to have the introduction of, call it, the transformer papers, which has then led to, "Okay, how does this now lead to a generative AI environment?"
And we were fortunate that we had relationships, of course, with some of the large hyper-scalers, also a relationship with Sam and with Dario that gave us access to platforms to test and train. I think they realized that because of the breadth of our portfolio, we provided an interesting set of use cases for this technology. And that's really where things started to evolve over the last almost, I guess, three years ago, a little over three years.
Sarah Elk:
Amazing.
Andrew Ng:
One of the things I was fascinated by was about the agentic AI factory. I think I'm the guy that coined the term "agentic" and when I came up with the term, I wasn't expecting it to take off like that. But when I hear about luminaries like you talk about the agentic AI factory, sometimes I wonder, did I overdo it? But it's clearly been very beneficial and successful for Vista. We'd love to hear more about your experience rolling that out.
Robert Smith:
Yeah. Thank you again, Andrew, for coining the term, because it gave us something to coalesce around. Which is, as you know, when you're leading through transformation, you need things to coalesce around to bring ideas forth in ways that can progress the intent. For us, one of the things we did, well over a decade ago, was there's an introduction of this hosted environment for software. In the early days, we were on-prem, go client server to on-prem and them prem to this thing called hosted, which ultimately became the cloud.
And over 12 years ago, I had a relationship with a young man that nobody knew much about by the name of Andy Jassy. And he was building out some capacity to support one of the large businesses that enabled him to really build out a compute infrastructure. And I said, "Hey, one of the things we want to do is migrate our businesses from on-prem to cloud." And we had all sorts of people calling us, "Hey, we'll build this for you and build compute." But he had compute. And I said, "Well, listen, I want to build a factory that converts on-prem software companies to cloud, but I need compute."
And so we struck a deal way back when that we could now drive our companies into their cloud infrastructure. Of course, now we have Microsoft with Azure and others to support that, but the whole concept again of a factory to actually take these enterprise software companies and convert them into a different form factor, in that case being hosted or cloud, was something that we pioneered well over a dozen years ago. So, now you have an introduction of a new technology, this whole generative AI technology. And so we said, "Well, listen, let's use the same methodology. Let's build a factory."
And so we started contacting some of our partners and relationships about how to do it. And Satya actually called and said, "Hey, listen, I hear you're building this factory. We'd like to provide the foundry tools for your factory. We'll be the foundry to your factory." And I said, "Sounds like a good idea. Let's see if this works." And so we, of course, sent teams up to Seattle and part of it was with a couple of our companies. I think you and I chatted a little bit about our customer success business, because we thought that was the most factory ready. And then some other people who could actually help us architect what a factory should look like for the rest of the portfolio company, so that's the narrative.
Again, history, as I say, doesn't repeat itself, but it rhymes. And so what we're looking to do is rhyme that same capacity. And now with 92 software companies to date, we have over 30 that have gone through our agentic factory. And that's basically enabling these enterprise workflows to embrace, call it, agentic technologies, i.e. agents who are actually working and navigating through these workflows to access very specific content, in context, and managing dynamic data repositories to deliver ultimately, in many of these cases, deterministic outcomes using some of the probabilistic inputs that come from utilizing this technology.
So that's really the journey that we've been on. So it's the second factory we've built, and now it's continuing to go through a moment of scale in our factory.
Andrew Ng:
I'm glad you mentioned Andy Jassy. I think, as respected as he is, not everyone still remembers how he changed the face of computing, but you were there in the early days. It sounds like you know what he did.
Robert Smith:
He is a brilliant pioneer. He gets nowhere near the credit he should, but he will actually go down as one of the great leaders and luminaries in our industry. And people hopefully will take some of his lessons of leadership and apply them to their businesses.
Andrew Ng:
Yeah. Well said.
Sarah Elk:
That's great.
Andrew Ng:
Very well said. Hey, of the agentic factory, you mentioned just now the customer success agent. That was actually one of my favorite examples of a [inaudible 00:06:19] agentic AI workflow because a lot of people talk about cost savings. That's something that people used to do, "Let's automate it, get cost savings is fine, that's okay." But I think one of the reasons I thought what you built for the success agent was remarkable was because it allowed you to not just take costs out, but serve a lot more customers, and therefore drive business growth. Did I get that right? I'd love to hear more about that.
Robert Smith:
Yeah, sure. So, if you think about knowledge workers generally and customer success agents specifically, they often have a wide number of tasks that they have to execute on behalf of a specific customer. And some have a number of customers and the same set of tasks for those customers. And in some cases, you need access and understanding into various parts of the relationship. It might be what trouble tickets are outstanding on the one hand, what deployment might be in the other, in certain divisions or departments or across different institutions. And so an individual who is managing large customers, often has to have access to multiple databases and multiple workflows to actually provide a comprehensive view, or in some cases, a very narrow perspective on what is actually happening to make a customer relationship successful.
And so the agentic, call it deployment, for us means, okay, how do we actually take... You think about a person in that position, what are the tasks that they are involved in every day and can you take those tasks and create agents that actually perform those tasks either more efficiently, more effectively? And so then you have a series of agents that now are performing groups of tasks and then you can orchestrate them.
And in the enterprise, importantly, often you have to also record what they do, so that you can actually go back and evaluate either what went right, what went wrong. In some cases, you need it for compliance or regulatory requirements. And so now you have to build out a whole different set of, I call them, administrative agents. So, you have task agents, orchestration agents, administrative agents, and then some of the administrative agents actually have to provide telemetry, right? Report back what happened, record the utilization of tokens in that environment, how efficient was it? Is there a better way to do it going forward? So, that is, again, I call it a complex workflow.
We have now, through our 90 companies going through... Again, 30 of them have already progressed through the factory, come up with four archetypes that do this sort of work. The one I just explained is, we call it the smart sidekick, who actually is able to evaluate various workflows and then bring them forward to aid the worker in being more effective in delivering what it is that the customer's expecting and hopefully delighting that customer with that work.
A couple of other architects, we call it the always on, right? So, these are agents that are constantly monitoring broad scale implementations. Some are orchestration agents. We can call those the conductors, managing a process that might be occurring in a financial institution where there's some regularity the way things are done. And then of course, things that are, I call them a workhorse, which is saying, managing broad deployments. You think about in the government. How do you deal with a government entity that has to deal with everything from snow removal to potholes to DMV, right? So, there's different things and they may not be that deep on the one hand, but they're broad and can be called upon by, in some cases, millions of people at the same time, whatever it might be.
So, coming through our factory, we've determined that there are certain agentic archetypes that require different forms of either configuration or delivery of systems and some require a shorter latency than others. So that's part of what I call the outputs that come from having a factory approach across a broad number of companies.
Sarah Elk:
Robert, with the agentic factory approach, I'm curious with the world changing as fast as it is and you mentioned complex workflows, more agent orchestration, how has your factory approach changed over time? With 30 companies coming in and having gone through it, I imagine the capabilities in the factory, how you're approaching that, has changed as well?
Robert Smith:
Yeah, it will. It has to evolve. Our teams, our company, our philosophy, especially if you invest in technology, if you're not building the design point around evolution, you've got a problem, right? Because our technology changes constantly. And so we have to have organizational design that constantly evolves.
And it's important because as later and later tools, in this case, LLMs and models are delivered, they have enhanced capabilities. So, things you had to do in the past, you can potentially do less of in the future, right? So, now you have to evolve the way that you're writing and the way that your agents are doing discovery and evaluation of the environments so that you can become more efficient, so it's part of the process of innovation.
And the good news is, because we have 90 companies across 70 different industries, we can do what I call a series of designed experiments that we can say, "Oh, well, this is effective to this environment, so it should apply to these others." And gain some economies of scope that way.
Sarah Elk:
That's great.
Andrew Ng:
Yeah, but most CEOs run only one company and don't have the luxury of running 80 plus things the way that you do and be able to test and spot patterns across so many different businesses. What are the patterns you're seeing in terms of what works and what doesn't work in terms of AI adoption? And what lessons would you share with business executives that run only one company at a time?
Robert Smith:
The thing that I tell my teams all the time is, "You have to take advantage of the resources that you have." One of the resources we have are 90 companies, so let's take advantage of that, which means try various combinations, permutations that yield outcomes. Like all these things, in innovative environments, you want to fail fast and then adopt quickly, right? And adapt quickly based on those failures.
So, when you asked the question of what recommendation do I have for executives? It's the same when we have to each individual CEO of our portfolio companies, which is, in some environments, we're saying, "You need a scrum team." Okay? Where you carve them completely away from the rest of the organization.
In some, you go to... And I like one of my CEOs had a really great approach and she said, "Listen, let's start with a zero-based planning of the company." Because the size and scope and scale of that company is, we can just go AI-native on this, but we have something unique. We have context, we have data, we have workflows that no one else has, let's just rebuild from the bottom up. You see what I mean? So, that applies in that case.
And so then, as we look at another business, it may be one or the other of those, or it could be a different approach. But what we're encouraging them to do is to take some risk in this because the economic benefit of what we are seeing is orders of magnitude higher than the economic benefit that we saw going from on-prem to cloud, which was honestly two and a half to three times economic rent pickup, okay? So, it's worth taking that risk and you have to encourage them.
And then you also have, to the extent you can, which is to point to successes. We just had a CEO summit, so I had, I think it was 72 of our 90 CEOs, all in one place, two days, okay? Which is a lot, but they get a chance to not only learn and learn more about the capabilities of our agentic factory, but the point is they also demonstrate to each other what worked in their organizations and what didn't. Finding the right leaders who can embrace the change, ensuring that those leaders have the resources.
But one of the most important is mindset, believe it or not. It's not the technology, it's the mindset. And the best way to move a mindset is to show people. And so we bring our people together to show them the results. And by the way, they show each other. And so CEO to CEO says, "Here's why this worked. Here's what I did and here's why this is effective." As opposed to us preaching to them, they get a chance to share with each other and it's a different efficacy in my mind.
Andrew Ng:
Yeah. What we're really curious about, why is this so hard? I remember for many teams I led, over the years, when I was running the Google Brain Team, which is a team that wrote the transformer paper you mentioned some years later.
Robert Smith:
Well written, by the way.
Andrew Ng:
I was doing corporate... Yes, the team did a good job. I wasn't personally involved in that paper but the team did a good job. It was that same corporate change management that I was doing at Google. I was finding successes, highlighting the leaders, letting each other show each other the results, creating stages, the successful projects to be highlighted, supporting others to get promotions and get pay raises, and using that to encourage more people at Google, at the time, to adopt modern AI.
But the thing that always surprises me, when my team at AI Spy, when we work with businesses, is even though this is the standard corporate change management playbook, to do all of these things that are all completely sensible, it is just so hard for businesses to do this. I'm actually... Oh, and I'm asking you this partly because I know you've written this book on leadership inspired by the writings of MLK. But why is this so hard for people to do?
Robert Smith:
Yeah. We all know that when you introduce change, those who've been doing in a certain way are reluctant participants. And those who will benefit from that change aren't necessarily the, our hesitant adopters, right? Change is hard for some people. I like to think about organization design. Call it, and I actually got this from a friend of mine, an 80-18-2 rule. 80% of the people like things the same way every day. About 18% like incremental change. There's 2% who want to change everything every day, right? And the combination of those, as a senior leader, you need to know who your 80s are, who your 18s are, and who your twos are, and how they affect each other.
And part of what our job, as senior executives, and I'm assuming, again, the audience here is senior executives, is to identify the twos in a way that they can get the proof points to convince the 18s and the 80s. And often it is... We all, logical, rational people like to see proof points. Markets like to see proof points. And if they don't see proof points, they have a healthy degree of skepticism.
And the thing about our market, which is interesting, 96% of software companies are private, okay? Enterprise software companies are private. And so the proof points aren't delivered quarterly in a way that the general investing population understands the dynamic of change and how rapidly it's coming. And then of course, the media sometimes says, "Oh, wow, look what just happened." Well, it's been happening for 12 years or six years, that sort of a thing. And so it's those incremental things that are often hidden from the view of the broader public markets.
But again, most people are comfortable with the way things are, even though if you just... Logically, things are always changing, always changing. And it's just a matter of at what point in time do you actually realize that you have actually been changing and evolving and that it's important to do that to progress, frankly, as a society. So those are all the elements that we see. And there's certain teachings like in Martin Luther King that talks about the importance of looking at facts and then enabling people to make change and enabling the human spirit to be elevated and lifted by that change, which actually makes us all more productive in so many different ways.
Sarah Elk:
Yeah. I think that's so consistent with the principle of scaling bright spots. So how do you take the 2%, prove that the new way can work and that it's better than... It's the more compelling future versus where you are right now. I'm curious in how you've designed the collaboration across your companies. It sounds like you just brought your CEOs together as one example. Are there ways that you've been thoughtful about how to get that change to scale faster or how to scale those bright spots from the 2%?
Robert Smith:
Yeah. What we do... We have a well established process called a BPSS, Best Practice Sharing Summit. And the CEO Summits are part of that. We also have a CXO Summit twice a year. We also have a BPSS for all the chief marketing people, one for the salespeople, one for R&D, et cetera. And so as a result of that, each one of those departmental... And again, this is across all the portfolio companies. So we may at any point in time have 450 people sitting in, I like to find hotels that are reasonably priced. So they'll be sitting in Utah in September.
Sarah Elk:
Yeah.
Robert Smith:
And in doing so, they will then present and teach and train to, with each other for three days, different best practices to make marketing organizations more effective or development organizations more effective or sales organization more effective. And it's that same motion. So again, designing an organization to embrace the evolution, again, of change is how Vista's designed. We are designed to evolve and to enable our companies to evolve. So it's not... As a chemical engineer, I don't like instances of success. I like processes that lead to success consistently. And so that's the way that we approach value creation across our platforms.
Sarah Elk:
Yeah. It's interesting. There's a discussion that's been happening amongst CEOs on, is that the winning capability actually? The idea that you can innovate and scale and do that at pace, does that become the competitive advantage that the process itself of being able to drive that reinvention is as important as choices you make in customer offering because you're continually... It's that always-on process you're describing where the technology's changing, the organization's changing, and the ability to harness that is the winning factor.
Robert Smith:
In my business, I believe so, because we are in, I call it a continuous process business. We are continuously investing capital, creating value, and returning capital, and that's happening all the time. What often happens is people will think about life in instants, what is the average life of a CEO of a company? And are they just looking to optimize for the period of time that they are the CEO? And that creates a whole different set of decisions, which creates a different set of outcomes, right? Because they're not necessarily building that business for the long term. They may be building it for just an exit. And astute buyers will figure that out.
And I've always been of the view that build durable companies that have the capacity... You build in these systems, it's not just a function of the work you did with them, but you built in some infrastructure in those businesses so they can sustain their growth, sustain the profitability long after you've either been an investor or been a steward of that business. And so that's the approach that we take because I'm of the view that build companies that buyers when they're evaluating, say, "Oh, this is sustainable with or without Vista's participation, and therefore it's worth paying a premium for them."
Sarah Elk:
Yeah, there's certainly a lot of perspectives on the durability of the software market these days. Yeah.
Robert Smith:
Yeah. And look, I think some of that perspective comes from... Again, we talked about people's unwillingness to change because there's a narrative out there that says, oh, ARR should last forever. And well, okay, you got this, the terminal value of these software companies is assured because people need this software forever. And this whole dynamic of recurring revenue, which is how we price. We went from on-prem to cloud, which led to a different business model that went from license and maintenance to subscription, right? There's a couple of things that happened there. There's a technical change from the way that the software was delivered. And there was also a business model change. People conflate the two as one, but it's actually two different things that happened.
And as a result of that, that people say, "Oh, well, what's the terminal value of that business?" Well, it's going to be a function of the ARR, which is a function of the subscription, which is the function of the seats that are sold. And that's going to go on forever and that's how the pricing's done and how you get the fast growing, high multiples, high valuation for terminal value. Oh, wait a minute, Andrew and his team came up with this new thing, right? This new technology that changes that. Because one seat is more productive than maybe 20 seats utilizing these tools.
And so if you're still pricing on a per seat model, well, that changes your terminal value calculation if you're priced on a per seat model. And so now there's a new business model that says, what is it based on value? Because if I can deliver a set of agents that in working with an individual delivers the capacity of a hundred people, I should be pricing that, I think our fair share is 50%. Some might argue differently. Okay. We should be pricing that at a much higher value, which is actually a multiple of the per seat model when you actually do the terminal value analysis.
And once enough proof points kind of come out, people say, "Wow, these companies are actually worth more because the total addressable market is much bigger than what it was on a per seat basis." But that's going to take some time to diffuse into the marketplace. The technology has to diffuse in a marketplace, so the productivity of it is realized. And then the business models have to diffuse in the marketplace where the realization of the revenues and the growth is more visible so people can point to it and say, "Ah, that's why these companies are worth more."
Sarah Elk:
Yeah. What do you think, Andrew? 50%? Is that the right value share?
Andrew Ng:
No, I've never been able to get away with that, but it's respectable that you can.
Sarah Elk:
I would love to know... You've been on this incredible journey and have taken bold action on your AI factory and other things that have been, I think, differentiated and unique relative to what we've seen from other players. What has been most surprising to you on that journey so far?
Robert Smith:
In this most recent journey... Let's focus on that. There are certain individuals, people in our ecosystem, who are intellectually curious and really want to see how this plays out and really want to get involved in that. And then there's certain who aren't. And I call it trying to believe that the world is what it was five years ago, eight years ago, 10 years ago. That's a surprise to me in all frankness. And you see that play out in some respects in some of the public market investing and not really understanding the difference between enterprise software and consumer software and the impact of things like context and trust and scale and the importance of that in operating for businesses and companies.
I like to say in the enterprise, close enough isn't good enough. You can't have probabilistic outcomes for a wire transfer, right?
Sarah Elk:
Absolutely.
Robert Smith:
Where my wire transfer is mostly right.
Sarah Elk:
Yeah, exactly.
Robert Smith:
So anyway, those are the things that are surprising. But in some respects, it takes a while for not only the technology to diffuse, but the knowledge and the understanding and awareness. So I guess it's not too surprising. It just takes time to get through these curves.
Andrew Ng:
Just one last thing I'm curious about, is what you did at Morehouse made the press. I think we all read about it. You're offering to pay off the student debt to the graduating students. You're clearly a champion proponent of higher education. What's your view on that in light of what everything tech and AI is doing to us?
Robert Smith:
Yeah, we still have the responsibility to liberate the human spirit, and these tools can do that. And removing certain burdens of access to opportunity, education being one of them, I think is all of our responsibilities. I think this modern AI, I think, can play a big role in improving the quality of life. And in one of the ways is the quality of education. One of my kids personally is now using an AI tool, agent every day as part of his everyday learning experience. And I am fascinated, excited to see his progress. And now one of my other children is now going to start here in the next couple of months in that environment as well. And then I've got a couple more.
It's just, to me, when you see the efficacy of it and the ability to access it... And so part of what we have to do as an industry is make sure that we are enabling access to opportunity, and one of those is education so that people can actually liberate their creativity, liberate their problem solving ability, liberate their capacity to make our planet, frankly, a better place to live for all of us. And we just have a responsibility to do that as effectively as we can in our sphere of influence and domain.
Andrew Ng:
Yeah. I think that's a great note to end on. There's a lot we can do through higher education, through other things to lift up everyone.
Sarah Elk:
Absolutely.
Robert Smith:
So let's do it.
Sarah Elk:
Let's do it. Thank you Robert so much for joining us.
Andrew Ng:
Thank you so much, Robert.
Sarah Elk:
We really appreciate it.
Robert Smith:
Thank you. It was my pleasure and I'll look forward to further updates later. So it'll be fun to talk later after we deliver more of this.
Sarah Elk:
Would love that.
Andrew Ng:
Looking forward to that.
Sarah Elk:
Would love to get the update.
Robert Smith:
Great.
Sarah Elk:
Thank you.
Andrew Ng:
Thank you.
Sarah Elk:
This was Winning with AI. Thank you for listening. We'll be back with more episodes coming soon.