Greg Shove on Why Most Companies Are Not Seeing ROI On AI (yet)

0:00 There's one word for all this, chaos. One of AI's superpowers is it allows us to jump our capability boundaries. And we use capability boundaries to keep everything working, keep the systems working, right? Keeps people in place, in org charts. An education can become a software company pretty much overnight, right? A designer can become a marketer and we can all start coding even in building disposable software. There's one word for all this, chaos. We're just entering into an era of extreme chaos.

0:31 Hi, my name is Greg Shove. I'm the CEO of two AI companies and I spend my life Monday to Thursday in corporate America where it's one tale of AI, which is like, "Hey, this is sort of useful, but not great." And then Friday to Sunday I live in San Francisco where everybody can't live or work without AI. So, we're going to talk today about the tale of two AIs. >> Greg, thank you for coming back for the third time.

0:54 You and Al are resident expert in how What is the level of how well corporations are integrating AI into the organization? So, what is the current scorecard and what can they do to become better? >> First of all, thank you for inviting me back for the third time. I'm not sure I'm an expert. I don't think any of us are yet. We will all know who the experts are in about 10 years. Hopefully by that point all the AI influencers that are in my LinkedIn feed are gone and it will will will actually have some experts, right? Those who actually built and transformed with AI.

1:32 Listen, here's what we're seeing. The grade for enterprise AI adoption is probably C- at minus at best. But, for 10 to 15% of the organization it's A+. The reality is there's all kinds of ROI from AI. It's not that there isn't any, it's leaking. >> Say more about that. Leaking. >> Yeah, well, the ROI is being kept by the employee. >> Mhm. I think that's very true. >> 10 to 15% of every organization are the growth mindset, are they ambitious, are those that are curious, are those that listen to this podcast.

2:08 And they figured out AI, whether their employer wanted them to or not, they figured out AI quickly and they're using it on their, you know, either the official AI that the company's paying for or their own AI that they're paying for or maybe they're not even paying for it cuz you get great I get AI for free as you know. And they're keeping the gains. And of course they should. >> Do they keep it because they want to or is it, let me give you an example, we have designers for example now that obviously use AI to kind of come up with ideas and also to render and stuff like that, but they also now can write marketing copy for these dog toys for example for BarkBox. Now the issue is of course that that's not normally their job, it's the marketing team's job and so you have this kind of released energy but it's in the design team, not necessarily in the marketing team and so we don't see like any improvement for example in FTE reduction because yes, you have people doing somebody else's job, but you don't necessarily have them overlaid neatly in the organizational design that we currently have.

3:04 >> Yeah, that's that's exactly part of the problem, right? That I think that's also the second stage problem. The first stage problem is I'm getting some time back, allows me to do other things, should I do more work and make my job potentially less secure or should I go off the dog and take a yoga class? Like every knowledge worker knows how much work they have to do each week in order to be considered a decent or good performer or even a superstar performer. Like every engineer knows how much code to write in order to be considered a good engineer. Like we all know this, whether intuitively or even sort of indirect numbers. Like we all know this and so if AI can help us get that work done sooner, we're going to use AI, at least the ambitious of us are and the ones who kind of can figure this out. We're going to use AI to get that job done sooner. Basically, what listen, remote work gave us Friday afternoons off. AI gives us Friday mornings off.

3:53 >> How do you think how do you organizations Is it a lost cause for organizations to try to >> cause. It's not a No, just be patient. If you know, be patient first of all. Second of all, don't have your CEO run around bragging about layoffs. And don't have your CEO running around blaming AI for layoffs. This is such frankly, that that these large organizations are talking about AI layoffs already. It's not related to the AI. They over hired.

4:22 Their business is changing. They need less people in certain areas. The fact that they're blaming AI or or you know, crediting AI for the fact that they can lay off as Amazon did, you know, thousands of people at head office. No, they have too many people at Amazon head office. There's 325,000 people trying to optimize amazon.com. I think that's a few too many. Like AI is not so well adopted inside of Amazon that they can identify with precision where the layoffs are going to be. That's just fantasy. What's going on is, yeah, they've got too many people and certainly in some teams the team is being made more productive by AI, but I would argue the employees keeping the gain primarily because the organization, first of all, doesn't have a point of view about AI. Like is it good to be using AI and is it good to be stepping out of your boundaries in doing some other work? Are you going to be basically messing up the organization, messing up the workflow, and so on? So, I think part of this, first of all, is sort of cultural and the concern about layoffs and the concern about how relevant will I be and how you know, how secure is my economic livelihood? That's natural. Employees are acting rationally in this in this moment, I think. And the second is what you said, Henry.

5:30 Like it takes a while to reorganize the workflow. It takes a while to figure out, you know, how the team will now get stuff done. And you know, why would you expect that to happen after a year? It's going to take in some cases, you know, two or three years. >> I stumbled into this use case the other day that I'm now obsessed about because I think it's one of the first kind of good examples of AI really taking over a workflow that was meant for humans. And so the example use is that students that are trying to get a job, they send a thousand applicants now instead of 10.

6:01 It used to be effort being something that you could use to actually figure out who was the good applicant. Obviously now you have AI just reading all these applicants and so this human to human workflow is replaced by an agent to agent workflow and therefore doesn't work anymore. Have you seen other examples of that because it that seemed to be like such a clear one and I would imagine that this is just the first of many that we're seeing as we're putting agents into our workflow, then actually the workflow itself break down.

6:26 >> Yeah, I think it's happening all over the place. I mean I think this is kind of AI workflow, right? I think I think we're writing more marketing copy because we can and we're writing more code because we can and some of it's good and some of it's going to be thrown away. We're building more, you know, automations and GPTs or you know, Gemini gems than we probably need and hundreds of thousands of those will just kind of fall by the wayside and not be used.

6:47 >> Can I ask you one one thing on that? Actually, do you think it's because that we think about in the wrong way? We've kind of taken the SaaS software. I'll give you an example. I had a birthday last weekend and had a bunch of people coming over for dinner. I needed something to basically put all the people around the table. Now, I Googled for a second I'm like, you know, table seating software and like a thousand different things came out. I was like, so I just went to Replit and saying, hey, I need this and then wrote this thing, had it, you know, made the table plan, had a few times where I needed something for the agent to change like for example, I went from three tables to six tables and then I asked to reconfigure it, it did it brilliantly, right? Now, obviously I will never use the software again likely. Like it's not I'm not productizing it, I'm trying to make a business out of it.

7:28 >> So it's single use software. >> Yeah, disposable software. >> It's completely disposable software, right? And so we thought that SaaS soft So we were talked about I think last time that we were unbundling basically SaaS in the same way that the music industry unbundled the album, right? So first it became we went from albums to singles, but now we're not even having the singles anymore. Now, we're just basically writing the tunes as we need them and then we're throwing them away.

7:51 >> Yeah, yeah, sure. >> their live performances, yeah. >> Yeah, first of all, you're kind of a nerd, right? Let's just be clear about that. The year writing a disposable software to figure out who should sit at dinner. How many tables was it? Was it that hard? >> You know, there's 115 people. There's six tables. So, it wasn't super trivial. It wasn't like eight people. >> Okay. We're four people. >> it was I thought it was Thanksgiving dinner. I'm like in our house we just wrote little name cards and walked around and dropped them on the table. I don't know.

8:17 >> No, no, no, no, no. I wouldn't have put it put it I wouldn't have put it on me not to do that for four people. >> Okay. yeah, let's be our headset version, all right. >> That's absolutely what's going on. And for that particular set of opportunity, we also need a new mindset because we do think of it as software going through a product management life cycle. You know, we were at a a dinner recently in San Francisco. We hosted a dinner recently in San Francisco and we had at the CIO of a major tech company, I can't say which one, at the table and his point of view was this is chaos for me, this idea of, you know, unleashing employees not just to use my GPT or Copilot, but actually to let them build, you know, custom GPTs or automations. Like from his perspective, just absolute chaos, right? Disposable software. Like what CIO or wants disposable software inside in the organization running around connected to corporate data, right? So, there is so much change that has to happen. I want to go back to the your first question, Eric. There's one last piece, phase three of this, I think, when you think about sort of capturing the the gain.

9:17 What I'm also not seeing is that managers, leaders of these teams, and these teams are somewhat or maybe fully AI-enabled, but even if they're somewhat AI-enabled, you as their manager have an obligation, a responsibility, to come up with what is the new work we're all going to do. All we're talking about is, you know, let's go optimize everything with AI. I call it cut and create. Let's cut workflows and tasks and and potentially vendors and data source like you can cut a lot if you are good at AI and you and you >> going to be baseline, right? Like I think you said that last time that the 30% is just basically like the baseline of what we have to find in even >> the create? And and and really more importantly like come up with more interesting and higher value work for your team to do. Actually get them excited about, "Okay, we did cut."

10:06 Hopefully not many people if any, but in some teams obviously that will happen. But kind of more importantly in this in this space we now have in our day, yeah, you know, as a manager come up with some good stuff. Question to you Greg or or to you Jeremy. >> I totally agree with you. We do the all these workflow, we make them a 10-tick and suddenly we find way of getting more optimized. That doesn't grow the business, right? Now, we also talk about that this is AI's like electricity and not just like the internet. And so instead of just getting like an electric horse, we have to think about factories and working at night and all these kind of very kind of fun foundational different kind of approach to work. When you're somebody who is tasked with being a manager in an organization, how the hell do you go about even if you're the CEO, how do you go about thinking about what this future's going to look like when you're basically saying that you probably have to completely redesign your organizational structure, you have to re-maybe imagine what you do as a business, and all these features in between that can now be enabled through AI can become something completely >> I think the first thing you got to create obviously a great question and maybe the hardest question to answer, particularly for incumbents, right?

11:11 People have existing businesses and managers who have existing day jobs. I think the first thing you got to do is find some time in your own day. And I think that kind of Google 80/20 rule probably works here. I think you've got to figure out how do you run your business, you know, your team, you know, your set of tasks and responsibilities in 80% of your week or something like that, and actually carve out real time to think about this and frankly turn on your strategic chops. And which I think is probably atrophied for a lot of people. This is something we don't probably talk enough about.

11:43 But to affect this transformation you just described Henrik, it's not easy and it takes quite a bit of thinking. And figuring out, you know, what should we do? And yeah, you can go hire McKinsey, I guess, but rather you don't do that. >> Can I say one thing there? I I I don't want to slow your roll and I don't want to detract. >> That's all right. >> want to interject here that I think maybe what we have to do is specify before we experience the gains that that's how we're going to use the found time. Because there's this thing, I don't know if you're familiar with I'm somewhat obsessed with it recently called Parkinson's Law, which is the phenomenon that work will expand or contract to fill the time we get it. And the challenge with productivity gains is if there's not something new to fill the gain time, the work just keeps expanding. Right? And so I think perhaps, I mean, it's one thing I love what you said about a leader's responsibility to say what's new, but part of the challenge as you're mentioning is they can't think of what's new. So therefore they have to carve out the time. And it strikes me that one of the most wonderful and inspiring ways to redeploy the time that we've now gained through efficiency or that we or the promise of time gained is to create the space that we've been saying we don't have. We don't have time to think strategically. We don't It's like so, what does that mean practically? Block the two-day offsite. Block the hackathon. Block pre-block before your team gains the productivity, before they get their 20% back, say preemptively, here's It's not You're not walking your dog.

13:15 We're in La Jolla together as a team. >> Right. Sure. >> You don't block that yoga class because we've got the weekly brainstorming session, right? But you have to I so I did wield the calendar as a weapon. That's my That's my thought. >> Yeah, no, I love it. It's like we all see it for ourselves, right? I mean, I've got focus sprints locked on my calendar each week. They always end up being Slack and email time.

13:37 You know, it's there's so it's so hard to protect your calendar to your point, Jeremy. that I do that part of the business. >> you about this cuz you obviously think a lot about this. You are, you know, ahead on the curve of all this AI stuff. I would imagine that even for you, finding the time, energy, and maybe even like team around you to say, "Okay, let's just wait for a second. Let's just figure out what distance are we in?" And as AI makes it easier and easier to teach people stuff, for example, then not necessarily like what do we do, but who do we serve and what is the problem that we're trying to solve for them.

14:14 Do you actually find time to do that? >> Yeah, yeah, I'd say mostly. I mean, I I I wouldn't give ourselves or myself, you know, an A+ grade either. It is exhausting. It reminds me of a pivot. Like I I've been an entrepreneur long enough to have done a lot of pivots. And if you've done a pivot in a business of any size, but you know, for me, startups. So, organizations under 100 employees. They're hard. They're really hard. And they take a lot of mental and physical energy and resilience and so on. It feels like that to me now. Meaning, we're not That's how you get everything in your business. But that level of effort to sort of rethink and maybe redirect, right? Maybe maybe pivot some part of the business, maybe the business model or the product model or the service model, you know, whatever it might be. Yeah, I I we're pretty good at it. Meaning, every 6 months we take a hard look at are we doing the right thing and and what are the metrics actually showing us and are these metrics or signals valid? Or do we need to be more patient and wait? I do worry sometimes that we're moving too quickly.

15:14 We need a We need more data and a little more sort of market validation. Sort of yes or no kind of thing. But yeah, it it feels like this is what we want leaders to be doing in this moment. Particularly if you're in this is deep in these industries that are so impacted obviously by AI, language-intensive, knowledge-intensive work. Yeah, but who I mean average manager doesn't really want to do this. It's it requires a system. They want to go home. They want to go home and feed the kids. You know.

15:44 >> I think the issue is also that I people aren't asking these questions as their AI questions. But very fast they become non-AI questions. They become geopoint strategy questions. And so there's this kind of weird vacuum where everybody's like, "Okay, let's get more AI into the organization." And then we you know, learn how to prompt, then we start to do agentic workflows. And then I think a lot of people are like, "Yeah, but to what avail?" Like, you know, nothing everybody will be able to do this. So what are we actually doing? And then there isn't really kind of like I think a terminology of like, "We're doing the AI pivot because we kind of have to, right?" Or we're doing the AI pivot because >> I've got to introduce this quote because I just love it. It's been something I've been thinking about a lot. I heard Simon Sinek say, who by the way I largely disagree with on the topic of AI. Let just let the record reflect. I don't agree with him. But one thing I do really I really resonated with and I know you too well, but I think it has implications on the enterprise question.

16:40 He said, "The difference between a startup and an enterprise is a startup's ambitions lie beyond its grasp. And an enterprise's ambitions lie within its reach." And I think we fundamentally we have an ambition problem. The reason managers are going home is because they aren't inspired by a new vision that's so unattainable they can't afford to go home. >> Yeah, first of all, love the quote, wish I'd come up with that. And and they're not rewarded, right? They're not incented to to to think that way. They aren't >> they they they aren't aligned in creating that long-term value.

17:12 >> That just creates risk for them. Like whoever said, "Go experiment and fail and sort of adopt or accept a lot of failures inside of a large organization? No one. In a startup, that's all you do. Otherwise, you don't survive. You don't you don't make it to the next month, right? So, yeah, I think the reality is for large organizations or mid-size organizations, something's going to have to break for them. And it it's usually something like your largest customer calls. This is what we're seeing and hearing for example in the media kind of advertising agency creative services all that part of the of industry. What we're hearing more and more is saying to their vendors, their suppliers, we're going to pre-negotiate now. By the way, like like has happened in manufacturing for 25 years, we're going to pre-negotiate our discounts every year cuz we know you're getting cost savings from AI.

18:01 And so, we want those cost savings built into our contract cuz you want us to sign a 3-to-5-year contract. So, we're going to see prices drop every year and let's say by 10% cuz you're going to get those gains on your side cuz you're going to effectively deploy AI. That's what for example advertising is forcing. Yeah, they're forcing, right? And all of a sudden the so oh right? Like my my my best customers are starting to pre-negotiate. Law firms will start to see this from their largest clients.

18:26 They're going to start to see their largest clients saying, "No, we want built-in price decreases just like General Motors has done the last 20 years with all their you know, component suppliers cuz the assumption was you'll deploy automation on the manufacturing floor and and get these advantages in terms of cost. You're going to pass them on to us cuz we're a long-term customer or we're signing a long-term agreement." So, that basically supply chain behavior is coming to knowledge work. So, we're going to need shocks like that, you know, the loss of these customers. It basically it's got to hit the EPS. It's got to hit the the income statement in some way. Opportunity's not enough typically for incumbents to move.

19:03 Sometimes it is. But it's more of a cost, right? If Why is that? I just be because of the risk, right? Just because the to go chase that opportunity is tough. You've got to build a team that behaves like a an entrepreneurial team, right? Behaves like a startup. You've got to provide capital more than you ever wanted. >> how the future is for organizations, that they need these, you know, SEAL teams rather than these army divisions?

19:27 >> Yeah, I think that's right. And even then, are you willing to stick with the SEAL team long enough and tolerate their failure? Look at you know, Mary Barra, General Motors. You've probably talked about her before. Like, the fact that she bought Cruise and then poured in billions of dollars and the thing was a dry hole, right? It was a bust. And the good news is she didn't lose her job. I think that the worst thing for innovation in corporate America would have been that she got fired, right?

19:51 Because it would have just told every other CEO that your board won't tolerate you taking those kinds of risks. To me, it felt like a risk worth taking, meaning look at Waymo, look at Zoox. That this feels like it's going to be a big business. They just didn't have the right startup, you know, in that race. You know, they backed the wrong horse. That that happens sometimes, right? Obviously, it happens most of the time if you're a venture capitalist. So, >> On innovation and on what you can build.

20:17 We had Dan Shapiro on at one point and obviously, you know, having been part of of the kind of like spinning out that company for Pre-IP, you know, tracked it quite a lot. And what's interesting was Dan is doing is that he's kind of like untethered to what he makes, right? You know, he creates software, he creates podcasts, he's a little bit Do you think that we're seeing these multimodal companies kind of emerge and will some of these bigger companies also like you're the same thing, right? You know, you do education, but you do software and you have, you know, this new thing that help you figure out what to do and it's all quote-unquote all over the place. Is that the new normal?

20:56 >> I think it might be the new normal to get started. Let's at the same time, we're going to need like the basics done well, right? You can't be playing in five different games and probably winning all of them at once. So, I I I think maybe at a certain stage, the new normal is you're not bound by capability boundaries. I do think AI, one of AI's superpowers is it allows us to jump our capability boundaries. And we use capability boundaries to keep everything working, keep the systems working, right? Keeps people in place, in org charts. Oh, you do that job, you can't do that job.

21:32 >> Can I challenge you on that specific one? Don't you think that one of the issues with that statement is that we're so used to think of a world of scarcity. We don't have time, we don't have people, we don't have money. And suddenly, the issue is that we need to learn how to think completely in abundance terms. We have endless people and we have endless time cuz it's literally if you can spin up 35 agents and the agent to manage them agent, you can do 35 people's job, which is now something that you couldn't do just like a year ago. I I don't know if I believe this, but like could the statement be that the issue is that we have not learned how to work like an Nvidia processors yet. We can't have all these systems running at the same time.

22:15 We used to parallel process and it's like a for us to unlock this new technology we have to unlock ourselves. >> Yeah, no, I'm agreeing with you. I'm saying that AI superpowers allow us to jump capability boundaries, but it's going to be hard for the system to accommodate that because the systems are pretty rigid and this idea that you're not bound by your time, you're not bound by your intellectual capacity, you know, you're not bound by your capabilities both individually and organizationally.

22:41 Like I think what's really going to happen is organizations will jump these capability boundaries, some of them faster. And so, to your point, individuals. Yeah, and education could become a software company pretty much overnight, right? A designer can become a marketer and and and we can all start coding even in building disposable software. Like all the There's one word for all this, chaos. >> Mhm. >> We're just entering into an era of extreme chaos. And organizations are designed basically to reduce risk, reduce variation, >> Yeah, cuz cuz the organizations are designed to deliver earnings on a quarterly basis.

23:18 >> Mhm. Right, predictability. >> Yeah. So, I want to go back to Henrik's sort of question or hypothesis. I think it's very provocative, which is you know, how much of this is going to be sort of more s- serendipity, but sort of you know, managing the chaos and sort of and not being held in place or or not having one business or sort of not chasing just one opportunity, but maybe three. I think that feels right to me, Henrik. But at some point, I think you're going to have to like pick one and go back to the basics of just like execute really well, grow the business with margins, keep customers happy, you know, kick ass, take names, deliver on time. I just the basics. I I I hear sometimes it's well, you know, it's it's time to like don't be a leader, be a steward. You're stewarding the change.

24:03 Yeah, sort of, you know, but no, I think you're going to need to lead. I think you're going to need to like make decisions. >> I'm still torn though on that. I mean like I I I think all logic has always been, you know, I've done incubation for a long time, right? As well. And you do portfolio entrepreneurship until something start to kind of pop off and then you just chase that down the rabbit hole, right?

24:23 But I'm just inspired by this thinking that that might not be what you do in the future. In the same way that, you know, I don't think it maybe because we've been talking to a lot of the folks that were in the early days of AI and they basically had to rethink what artificial intelligence was, right? And we all have these In the past, you know, we had these systems AI systems that basically thought like we kind of put a rationale towards problem. And suddenly they were like, no, we have these processors now where we literally just throw it everything.

24:53 And that's what we do. And so, I don't know, like I maybe it's just because that that is how my mind works. >> You know what it reminds me of, Henrik? It reminds me of what Brian said to us just the other day from you. That keep the GPUs full. It's a totally different paradigm. Greg, this is a early researcher at Salesforce who's now at you.com. He's one of the co-founders. But he said his whole mindset is keep the GPUs full. And that is it's a to Henrik's your point, that's like it's a totally it is an abundance mentality.

25:24 >> Yeah, yeah. >> But most people most people aren't thinking like it. Greg, Brian told us this episode isn't released. We're kind of talking shop, but hopefully it will be by the time yours is. But he talks about how every day his goal is to get to the point that he has something meaningful to hand off to the GPU. So that when he arrives in the morning that the experiment, like you think about like working in pharma, the experiment is done, right?

25:45 >> But Jeremy, first let me give you an example. I've started this new thing where I write like basically a to-do list for myself. And then I write to-do list what I think an agent can do. And so if I have when I look at my like to-do list, I kind of like I put one on one list of the other. I it's basically a business and a markdown note. And then in the morning, I go to Clockwork code and say, "Hey, look at this to-do list and just start. Like you just figure out as much as you can on this list that's kind of like for you."

26:10 And the and the crazy thing is of course like 25 minutes later is and it goes like, "I'm done." And then you look at this list with 10 different to-dos and it's answered pretty well. Enough for me to have to do the work, right? But then you do that and then you're like, "Okay, now like an hour and a half have gone through the day and this was basically what I have intended to be like my day's work, right?" And so it's kind of fatiguing cuz you did then you just have to feed the beast, right? And then you look at all these to-do lists you've had forever and you're like, " there's onslaught." And it's also it's it's fatiguing, but it's super inspiring, right? Because it's just this this fire hose of productivity.

26:45 >> Yeah. All I'm saying, Henrik, is at some point that fire hose of productivity has to be directed at I'm not saying in your case it's not happening, but at economically valuable activity that someone will pay for on a regular basis. >> You mean not writing software for how to seat people at my Thanksgiving dinner? >> As a as hypothetical as hypothetical. >> >> Yeah, and then eventually investors want their capital back with the return, right?

27:10 Honestly, I think you see this tension playing out right now at Open AI. You know, that the Sam Altman memo that was leaked about hey, we're we're code red to respond to the threat from Gemini. So maybe we shouldn't put ads into chat GPT and instead we're going to you know, not yet anyway, and we'll focus on making the product better and better better. You know, I think there is that real tension. How much are we sort of keeping the GPUs full in order to deliver an amazing customer experience and how much do we have to start to really think about you know, all this money we've raised and investors at some point are going to ask for you know, a return on that investment.

27:44 So I I do think that tension is is real and it is real for all of us whether it's a five-person services firm or a Fortune 500. >> Okay, we have to go back now because at the start you said, and I quote, 10 to 15% of enterprises are getting an A+. The enterprise grade is a C. And then we we kind of went down the rabbit hole of leaky ROI and individuals reaping gains themselves. Is that the case at A+ organizations or can you say what's different in the 10 to 15% versus the norm?

28:13 >> Yeah, I actually I actually said 10 to 15% of every organization was getting ROI but the employee was keeping it cuz there's 10 to 15% of the organization that are >> not 10 to 15% of organizations >> I would say the everywhere. Yeah, I would say the A grade is probably 5% or or less. Like I think it's a small number of of organizations that are actually getting you know, sustained consistent value and our our research are Yeah, I think it's a small number and our research confirms that. We just finished our bi-annual you know, AI survey benchmark survey and that's what the data said surveying 5,000 organizations. So So listen, I think what's what where people are doing well, if they're doing it well, is and we've talked about this before, that it's not rocket science, but you got to do it every day kind of, you know, consistently, which is why we're doing AI and why AI and AI manifesto. You have to start with that building block.

29:03 Why are we doing this and how will we do it as an organization? Like what's our operating culture around AI? Will we celebrate it? Will we tell people it's not cheating? Will we actually do hackathons? Will we encourage people to build GPTs, even if they're disposable and we throw them away, that's okay. Some of those GPTs or automations or agents will be good and we'll get benefit from them and so on. So, that sort of that I think that building block is is this unleashing employees. Second of all, giving them great AI.

29:31 Thirdly, give it everyone great AI. I still so many meet so many organizations that are giving only a part of the organization the good AI and maybe everyone else doesn't get it yet or they get sort of the free version. I mean, this is nonsense. >> I was talking you know, I was talking to an organization just yesterday, like global pharmaceutical company. Said, "Yeah, everybody has Copilot." I said, "Is this free version or the paid?" He said, "Well, about 10% of people have the premium version, but everybody" I said, "Well, is it dumb AI or smart AI?"

30:04 And and he said, "Well, I kid you not." He said, "The problem is we don't want to tell people the difference because even if they choose smart AI, the next time they open Copilot, it defaults back to the the dumb one and we we think it's too much work." So, the implication is, therefore, we're going to continue to allow people to use the dumb AI. It's like What? I'm like I'm literally on this call and I think I had to turn off my camera since like my face.

30:29 >> This is what's going on in corporate America, right? And and there's there's cost reasons for that, like the the the good AI costs more money and they and they don't trust their employees to use it the right way or get the value from the good AI. So, just give them the free AI. It's the basics. That's why most employees and in our survey, this is what the data said, are using AI to summarize emails. Basic basic use cases.

30:50 Yeah, listen, as a parent, would you if as a parent of two kids, would you give one kid AI and one kid not? And say, "Hey, both of you go to school, but >> We should do that in AP test. Do you think my sons would enjoy that? >> own On your own kids. Yeah, good luck with that. See if you're >> What a great question. I mean, how how By the way, I mean, seriously, everybody can relate to that.

31:08 >> Like you you you you >> You would do that. Like as a CEO, you think this is the right thing to do? Give somebody an organization good AI? Like you're special? And by the way, those people with the good AI are going to work with people who have the dumb AI. Like this is going to be In the meeting, they're going to figure this out. Oh, you give marketing good AI and give sales the dumb AI. Well, I think they're going to Those teams work together all the time. I think they're going to figure out at some point what's going on. And you know, it's just So, what >> What do you think is the best advice to the folks that are getting to the point where they would like to give people access to what's it called? N8N or Replit or Lovable or like something that is a little bit more meaty than your Co-pilot.

31:48 And yeah, I was going to start by your question of the the and technology officer just getting panic about this idea that somebody could sit there and code internally in the organization with the real data. Yeah, is that an expectation you would dare or how do you what do you think about it? >> Well, you've got 75% weekly active usage on Co-pilot. >> Sure. I I If you've built >> all your, you know, Co-pilot Studios or all your automations and you're you're kind of running out of gas with your existing sort of investments and you have employees wanting more, yeah, sure.

32:20 We don't need that You don't need to buy more AI. We routinely talk to prospects, companies that have bought three, four, five, seven, nine different enterprise AI tools, have deployed them, air quotation marks, which means they turn them on and did a, you know, a CEO email and maybe one lunch and learn and no one's using any of it. Like let's lay a really solid foundation and a really solid foundation is getting most of the organization using AI for more than summarizing your email and doing that on a regular basis which in my mind is weekly and should of course become daily. This is the standard we're we're we're aiming for.

32:59 If you want to be a super company which is really the only the only companies that will attract investor capital in the next decade are super companies. And this is how super companies behave. Most super companies are startups cuz it's so much easier to build a super company from day one. It's so much harder to take a company, a legacy firm, an existing organization and turn it into a super company. But it's clear super companies will win. Super companies will attract all the capital and their market caps will reflect the fact they're super companies.

33:30 So that's what we have to become and that's how super companies behave. Everybody gets great AI. There's no, you know, there's no discrimination inside the organization. They're encouraged to use it and they lay that very solid foundation cuz before giving them the tools to build agents, let them understand their workflows and which workflows can actually leverage an agent. And and so in in my mind a little bit more crawling before you run and but listen, I I get it. It's easier to buy software. If you're a CEO >> being a buying software, you had a new feature come out which I was fascinated about and that's how I understood it. You were trying to solve the problem that a lot of people go, "Hey, I have all these workflows in my life, but I don't really know how to make them intelligent. I don't know how to add AI to it."

34:15 >> Yeah. >> Could you walk me through a little bit? What is the diagnostics that you found out to do do on people that come in with that problem? And what solution did you codify into your system so that people might use your system, but they could also to kind of take the learning and and use it for themselves. First of all, the data is clear. The data says that even if you learn how to prompt at work, you're uncertain or unclear about how to deploy AI in your own workflows.

34:42 At home, we don't have any doubt. Like at home, we know right away, right? We can talk to AI, use it for parenting advice, healthcare advice, relationship, and so on. For some reason, we go to the office and we kind of freeze around, "Okay, I know how to use AI. Use I use GPT at home, but how should I use it at work?" So, this idea of, you know, use case discovery and use case coaching became so clear to us about a year ago.

35:02 So, Prof AI now has a new agent, and it's a use case coach. Basically, Prof AI, our our system knows who you are, knows where you work, and knows what job you're in. So, once you know that and you have an AI-powered system or coach, you can do a lot cuz AI is so, you know, AI is so performant and so capable. So, Prof AI will coach you on use cases, and you can do that a couple different ways.

35:22 It'll suggest use cases to you. Prof AI will just serve up, "Here, try this. Try this. We know what, you know, we know you're a content marketer, and you work for a CPG, and and you live in Brazil, and you work in Brazil, so that, you know, here are the use cases we think make sense for you." Or you can start by just chatting with Prof AI and saying, you know, "What do you do every day?"

35:40 And or "What are you doing this week?" And what, you know, "What are the tasks that you need to get done so you can clock off Friday at noon?" and so >> the best way to to discover different workflows? Is that just to do a calendar mapping? >> I think for That's one way. I think for most people, they need some help. Again, for that 10% or 15%, those early adopters, those growth mindset, they can probably natively on the AI figure out this stuff. Just by trial and error, like we all have. but I think for a lot of people in in large organizations, they also want to know what's safe.

36:13 >> Yeah. >> And so, Prof AI is is built in a way that can sort of suggest and coach kind of safer use cases, and then we can begin to share them, which is really cool. Once you figured out >> That's where That's where exponential gains come from, right? >> You got it. Yeah. >> So, Greg, I get the use case coach. That makes perfect sense to me. I will I'm reminded of something you said in your first visit to the show now at 2 years ago, which is crazy.

36:40 you said in regards to section generally and I think we kind of shared some cynicism around how much people want to learn. You said, I think the direct quote is, "People don't want to learn product strategy, they want a product strategy." I would be curious to know with the use case coach, one thing I heard you just say is I'm curious how if there are implications of that insight about product strategy on use cases. And specifically, what I'm curious about is does the use case coach merely suggest use case or does the use case coach just do the thing for you, i.e. give the product strategy? And why or why not?

37:18 >> Yeah, great question. It does not do the thing. So, the the use case coach and Profit AI is a coach. It's a it's a 24/7 AI coach. So, it is there to help you understand, you know, what the use case could be, how you could construct it, if you will. It'll It'll create the prompt for you, which you can then cut and paste. And and and then go work with your AI, whichever AI you prefer or that you're you know, your company's paying for. So, I'd say it takes you 2/3 of the way there.

37:47 Much faster. So, the time to value We need to shorten the time to value for employees from your first moment of kind of exposure or playing around with AI at work. Again, they get the time to value at home really quick. I think at work that that that there's too much of a lag. And so, the whole point of the use case coach is to get people's time to value you know, to within hours or minutes, right? In terms of play around with these with the Profit AI agent and you'll have three or four or five use cases immediately identified and then go go try it, you know, see if it works.

38:21 >> But knowing what you uniquely know about people's desire to learn and they call it friction, why did you make that design decision as far as what the product would do and what it would not do? >> Because we want people to join the AI class and benefit from AI as an accelerant. So, we want them really proficient in these technologies and tools to get the gains versus frankly eliminated.

38:54 And I think the more you're just doing it, you're building agents that replace humans. And that's not our mission. >> Whereas by It's almost one way to think about it is almost the IKEA effect that if the person has to turn the screws themselves, they'll take it the last mile, so to speak. One, you get enormous drop-off in the last mile, so that's which is kind of the argument against, but the argument for is that actually becomes the means of discovery.

39:22 The person who's willing to invest that last turn of the screw is going to be way more likely to share, going to be way more likely to enthusiastically engage, right? Because it required part of their investment. Is that right? >> I think that I think that's right, and I think it keeps the human more relevant for longer. I think that's a good thing. Mhm. Frankly, like >> Can I ask you We have this I have to think of Autos, which basically helps entrepreneurs build a startup using AI, right? One of the things that we are learning is that the models and Autos capable that we now have to design the system. And when we design the system, we have to think about should we just let the agent do this work or should we purposeful kind of say that the user have to do it? And we know, of course, that then to Jeremy's point, there's going to be a drop-off, but you also want the entrepreneur to feel invested in this idea and having to do this.

40:13 And we talked You talked about the 10% people that know how to use it, and then the others. Where are you? And And Jeremy and I had this conversation the other day. Where are you on the teachability of the 90% versus the Iron Man suit that we're now handing to the 10%? And what's going to happen? Is this going to be a place where these people that you talk about are the new staff members that you talked about earlier? Are they just going to be those 10%ers? And then the 90%ers right now won't learn it? Or some of them might make great over Oh, how should we think about this dynamic?

40:50 >> Yeah, I think I think we should think about it like like any change that we will have people that stay on the sidelines as long as possible, right? The laggards and the and the skeptics. I think most people, you know, will want to make the change. I mean, they're they're not stupid, and they will begin to see this future, and they want to be in it. They need to pay their bills, and if they work in the knowledge economy. So, I just think that that we are we need to be again, we need to be more patient here, and be more supportive. Yeah, we can use occasionally the stick if you're a CEO, like if you don't make this change, you know, you're not going to have a job here. Yeah, in some in some places and in some moments that might work, but we have to do a lot more >> this is the Is this the leaders now having the social media moment where there is the point where we owe it to each other community, society morality to have more patience with people and help everybody understand how to use these tools?

41:53 >> Yeah. >> Cuz you were like a you were capitalist, right? You know, I was reminding you of our last conversation. Like, so where where should we How should we compute this at this point as leaders? >> I think we should be impatient, but supportive. I think we should be demanding, but supportive. Like, particularly if we're in a company or industry that is in the crosshairs of AI, then the the clock is ticking. It it it never happens as fast as as sort of, you know, the media might suggest in terms of the disruption. But then when it does happen, it happens it feels like suddenly and but more dramatically. So, if you are in the crosshairs, whether you're again, your legal firm or a management consultant or or you know, a data services firm, you know, that this is coming and it's probably accelerating at this point and then and then when you get caught by surprise, you have a hard time recovering. So, I think as a CEO, we need to be demanding and patient and sort of ambitious, but we need to put in the support required. Yeah, I think it's okay to say this is a shared responsibility. I as a CEO of the organization can't do everything. You have to show up with a level of your own ambition and your own effort you know, to get to kind of to turn the corner on this AI thing, but you know, we have to do it together and we have to do it relatively quickly. But we will support you. We will provide the coaching. We'll provide the best tools. We'll provide the managers with the training they need actually you know, get their teams to be AI enabled and then use that time in a better way and and so a a lot has to happen here. And again, it's just it's just so much easier to doing it with a new company with a new organization versus you know, applying it to an existing one. But I think we need to be impatient but supportive.

43:28 And I think employees own this responsibility. I I believe it's shared. We got to get our head in this game though. Otherwise, I think we won't know how to react to your point. We won't know how to react and work with these AIs where we maintain our value at as humans. >> And maybe it's just that we will have to change. I mean, the Nicholas, my business partner, he made this point yesterday that sometimes there is elements in the body like the spleen that doesn't really do anything. And untold that it breaks and then you have to go and operate it out. But when it goes wrong, you have to operate it real fast, right? You know, like And so, maybe increasingly it is okay to think of us as a architect for these agents, and then our job is to sit there and be ready for something that doesn't work. The code can only be as crazy cuz at this point we don't know how it's written because all these agents have been writing it all. The marketing copy is just spitting out and it's an automatic system at one point it spits out something out that's vile that we have to stop, right? So, back to the other point about thinking about scarcity versus abundance, maybe also that increasingly there will be a different motives for us humans on how we conduct our work every day.

44:46 >> Yeah, I think that's right, but I think Henrik the that where we will insert ourselves is not just to catch the mistakes and solve the problems. I think it's going to be earlier. And I think I want my AI to tell me okay, now be human and inject your opinion and brilliance into this decision. >> Otherwise, Henrik, all these startups are going to be the same. If we're using AI to start all of our companies and do all this work, we're not going to have any differentiation in these products and services. And so, no one's going to win. Like I I I'm not that bullish on AI is going to come up with all these ideas and start all these companies for us without these moments of where the human needs to come in and make a call.

45:27 >> Bring in your humanity. Bring your humanity. >> Yeah, bring a point of view that's differentiated. Otherwise, we see into your point around job searching. Now, there's no friction in in job searches and so people can apply for a thousand jobs and every application looks like it's personalized. And that's what the data says is happening, right? It's And that whole market's basically not no longer functioning. There's so much congestion, right, in that system, in that marketplace of of knowledge work jobs. It's basically not functioning right now. And I think we're going to we'll see the same in entrepreneurship.

45:57 If if all these entrepreneurs are relying on AI to come up with these ideas and execute their go-to-market plans. We're going to be We're going to have a ceaselessness. >> I agree. >> I want the I want the AI Hendrick to say, "Hey, I'm doing all this work for you, but here's five moments where you need to come in and actually steer me." >> You do you, buddy. You do you. >> Yeah, you do you because if you don't, you're going to get like you're going to get my you know, startup idea number 495. You're going to get You're going to get Greg's stuff. You're going to get Greg's stuff if you aren't unless you interject right now.

46:25 And I gave this business idea to a thousand other people last night. >> -huh. >> You know, because they're all using AI cuz So I you know, that's what I'm hopeful or that's what I want from my AI. I want hey, I'll do I'll do a lot of heavy lifting, but we need some brilliancy. We need some human brains. what's the pricing going to be of this product or service? if we're AI do we'll have all the same prices.

46:48 >> We got to hear your short details on how folks can try out Profit AI. So I think we've kind of teed it up nicely if folks want to get in there and start to kind of play and find use cases. How do they find it? >> Yeah, go go to profit.ai and the consumer version is free. If you want a pure team, then you got to you got to call me and we'll charge you for it. no, go to profit.ai. It's great. And we want as many people as possible to join the AI class as fast as possible. As I said, there's no charge for consumers. Companies pay.

47:20 We just announced our agreement with OpenAI or about to announce our agreement with OpenAI this week. So we're one of their service partners to help their enterprise clients drive higher levels of adoption with GPT and Profit. >> Well done. >> Yeah, it's but you know, I think I think we're all seeing the same challenge which these technologies are offer so much upside, but the anxiety and the skills are not where they need you know, anxiety is too high and skills are too low to actually take advantage of these capabilities. As you've talked about before, this is not software. AI is not software. And when it's deployed like software, it fails.

47:59 And this is the key mistake that every leader is making. They're deploying AI like they deploy other software, and this is not software. And so, we've got to really do this differently if it's going to work. >> Brought it home. I mean, what a what a treat to have the first three-peat in Greg's show. I mean, of all people we've talked to, he's a worthy three-peater, right? >> He's such an interesting person. You know, I mean, I think a lot of us we we work a lot with how do we get AI into the organizations, and there is this kind of like almost camaraderie of people that on the front line. And they're like, "Well, you know, what does work, what doesn't work?"

48:36 >> You've touched the orb. Yeah. >> >> Right. Yeah. >> I think we all like I think this is all this race for the holy grail of figuring out how do we best take this new technology and put it to good use, and and we're all just trying all these different ways. And so, I mean, like for me, I'm very honored and joyous talking to a person like him. >> Definitely comrades in arms, that's for sure. I thought I'm just rattling through a bunch of stuff cuz I know we've got a bunch of things on our mind, but I mean, the insights per minute, you know, one of the highest I think of it of any guest cuz Greg just has so much experience, he has so much exposure across so many different areas.

49:12 One of the things that struck me was his comment, "Can you imagine giving one of your children a smart AI and one of your children a dumb AI?" I think it's so obviously wrong, and no parent would do it. And yet, we see it all the time in organizations. This department can have good AI. This department we're not going to give any access at all to. And it's so I I I realize there's dangers to treating companies like families.

49:37 Companies aren't families, but to Greg's point, marketing's working with sales, you know, operations and product development or they're going to know if if one department has dumb AI, they know it. And that will create a class system in the organization, and furthermore, you you almost negate the value of smart AI if a team working with smart AI is collaborating with a team that works with dumb AI. The the cumulative outputs, I don't think are going to rise to the level of the smart AI. I think they're going to be they're going to fall to the level of the dumb AI. What do you think?

50:14 >> Yeah. I think that's true. I think that and I think the other thing that he pointed out that resonated with me is a 10 15% of everybody in an organization are pretty good at it, and they they kind of super users, and and then you have like the rest, and what the rest does not need, which I think you guys both picked on, was they don't need to learn how to prompt better or to learn to do these more generic things. What they need to learn is how to do their job better with AI. Right? And so, they're looking to not kind of go from the abstract understanding of how to use this tool in a generic way, and then apply it myself to my job. They almost need much more specific to say, "Hey, if you sit in this type of job, and you're doing this type of function, here's like a use case that is very useful for a lot of other people. Now, try to, you know, train on that."

51:04 And so, I think a lot of us, you know, we we want to give people a fishing rod and teach them to fish generically. But I think maybe an unlock for a lot of organization is that you just got to go much more specific. I say it as a statement, but I mean it as a question. >> Start to do this here. Yeah, I think that's right. And I really like your comment. I don't know if you realize you're making it, but the comment that the goal isn't to learn AI.

51:28 The goal is to be even more effective in your work. >> Who? >> And I I've started kind of using that language in some of my training programs and things like that. My goal is not to help you learn AI. That's, you know, it's the means. The end is you more effective, more enabled, more quality, more joy, right? And learning how to collaborate with the AI is a means to that end, but the end can't be working with the AI. The end is actually doing whatever you do better.

51:59 >> And then I think the second thing which I'm stuck on and I think you mentioned it also in one of the other episodes, but it is interesting is when people start to become more efficient or even a little bit better doing their job with AI. Where does that extra time, energy where does that go? And you were talking about the Parkinson's Is that the Parkinson's law? >> Parkinson's law. I mean, to me it was very pointed when you said what you asked your bark team. That's incredible.

52:26 That's evidence right there of Parkinson's law. I don't know if it's Parkinson's law necessarily, but to me that was a great story. But anyway, you were saying about Parkinson's law. Please. >> Yeah, but I do think that I think a lot of us are now trying to figure out where does this extra time go? And how do we And I think what he was saying, which I think is an interesting point, is that that is not necessarily the company's time. I realize it's on their dime, but maybe you know, we pushed organizations so hard that what you really need is to give that time back to people to think about how do we get into growth mode again in a lot of these organizations.

53:00 But the time shouldn't just be sucked away and then yielded into kind of like higher e, but it should be how do you become like a better organization, more robust organization, one that grows faster. >> Well, it is it's an age-old problem. I mean, this this was true long before AI that people tend It's a lot easier to make improvements than it is to do something new. >> Hm. >> You know, I like what he said about cut versus create. And you think about cutting as it's six sigma, it's process, it's standardization, and folks will That's a kind of a known area.

53:42 and solving a known problem is a lot harder than identifying a new problem and creating from whole cloth. It's a very different challenge and so this in a way AI is bringing to a fine point the classic challenge of explore and exploit. Organizations, I mean that's like Jim March's theory from the 70s. Organizations are designed to exploit. That's their job. To exploit existing market, existing capabilities, existing resources, etc. And that's at odds with exploration. And the way you deploy resources when you're exploring is different. And I love March's kind of classic line. He says the organization that only exploits generally suffers obsolescence. It's to say they go bankrupt, right? But it's really hard. I mean that's where and but maybe you know, we talked about having experts on the pod. Maybe an interesting expert Henrik for I'm just riffing real time here. Not to make any promises to our audience, but my good friend and hero Charles O'Reilly, he's the author of organizational ambidexterity. I mean it's the kind of the quote-unquote solution to the innovator's dilemma, so to speak. But he's a really interesting thought leader and he and I have created new AI courses at Stanford, which is probably coming out in February-ish. But he would be an interesting expert not because of his knowledge of AI, but because of his deep understanding of this fundamental tension. And the AI moment is bringing that tension to a fine point. And it's exposing, I would say, maybe to come back to our conversation with Greg, it's exposing how little tooling individuals and organizations have around exploration.

55:21 >> I think that's a super interesting point and I think very, very good observation that it's probably the in it's kind of like the thing that people are not seeing yet, which is all these tools makes basically the factory work. And all this factory work that people have been doing can increasingly be done with agent. And so, what is left and that is the exploration, but the exploration doesn't really have a rule book. I I realize there's innovation experts like yourself, but most the time that's not what most of the organizations spend any time on, but now >> Right.

55:59 >> maybe they can, maybe they should. >> And and if you don't, I mean, it it's a race to the bottom, right? You can only exploit become more so much more efficient before you're basically erasing your then margins and like you're and you're cutting into bone. Like there's nothing left. There's no more fat to trim. And so, if your expertise is trimming the fat, eventually that's you're putting yourself out of business. I I would be remiss if we didn't also mention one of the other things Greg mentioned to us, his emphasis on manifestos. he mentioned you super companies. I I wrote down three things.

56:32 They have a manifesto YAI. They give people access to great AI and then 30 said, everyone has access. And on that point around manifestos, I just as a simple data point because I really wanted to validate that statement. I was at a retreat of maybe 50 private equity CEOs couple weeks ago. I did a similar By the way, I've seen these findings I'm about to share replicated. I had a a similar retreat among a bunch of CFOs and COOs. And at each of those three environments, I gave them I call it my 26-point diagnostic. What are the key elements that need to be a part of an effective GAI power transformation. And two of those elements are relevant to this conversation. One is, has the CEO written an AI manifesto?

57:16 And two, has the organization established clear governance guidelines regarding use? >> Okay. >> In each of these environments, both with CEOs and separately CFOs and separately COOs, a very clear pattern emerged, which is most organizations have a clear governance policy and very few organizations have a leader who's crafted a bold ambition. >> Yeah. >> And all organizations are wondering, why are our people stalling out? And I actually put the data on the slide, data from the room, so not like general data, but actual y'all just said this and I put on the room. And I kind of I I mean, I lovingly, lightheartedly make fun of them. I'm like, wait, you guys are wondering why people are stalling out when 90% of you said, we tell people what they're not allowed to do and 10% of you have said, I've told people what I hope they do.

58:08 Is there any wonder why people are stalling out? >> But it's a little bit of the micro version of the same picture, which is, you know, we know how to exploit, we don't know how to explore. >> Mhm. >> And that is going to be such a important part of living in an AI world where exploitation becomes cheaper and cheaper because the machines are very good at doing machine type work. I think as always, if people have a listen to the whole conversation with Greg, we hope they'll share this episode with somebody else.

58:43 >> One thing I will say actually just as we close is folks should listen for the one thing they can do immediately. There is something they can do immediately that I've got written down here. I'm not even going to I'm not going to spoiler it. There is something you can do immediately and it's not use Profit AI, though. Go check it out for sure. I'm not This isn't a product marketing thing. you can actually do something immediately and I'll just say this, it involves your calendar.

59:12 That's all I'll say. Okay. >> And with that >> Bye-bye. >> Bye-bye. >>