The Enterprise AI Gap: Productivity, Agents and Security
At the WSJ Tech Summit, Lattice CEO Sarah Franklin, Grindr CEO George Arison and JLL CTO Yao Morin join WSJ Leadership Institute President Alan Murray to discuss AI’s impact on corporate productivity, exploring the enterprise adoption gap, AI agents and emerging cybersecurity risks.
These 3 people do not belong on stage together. They have, they, so Sarah runs, uh, uh, Lattice, which is an HR company. George runs Grindr, which is responsible for violating all sorts of HR policies.And and Yal is with uh JLL, which is a real estate company. So so like figure out what are we doing.
I was just here trying to drink some wine and then all of a sudden I get pulled in, so it must be a mistake.
So here we are.So that’s why it’s gonna be why it’s gonna be the.panel and the the question I want to start with for all three of you is one that’s kind of been lurking in the background of our conversations for the last 24 hours, which is we all know that the technology is increasing at exponential rates.Um, but the application of the technology within companies is kind of bumping along down here, you know, it’s, it’s moving, but at a much, much slower pace, and we want to dive into why that is. And don’t take offense, George, Sarah, I wanna let you go first.
I love this is a competition, by the way. Oh, it is,
yeah, yeah, no, in fact, they’re gonna vote on the three of you. Yeah,
competition. I still love that. Let’s do that at the end.
Um, yeah, so, so this, you know, orthogonal experience that we’re having is one because we’re, we’re, we’re coming at it from the wrong starting place. We’re starting from the technology more so than the business, and what we need to do is write down.The business objective and really the recipe that we need to apply we’ve bought like we’re here at an incredible resort and I love talking about cooking and food and I think of it like a kitchen where we’ve stocked the kitchen with all the flour and eggs and milk and all the ingredients that you need and we sent everybody to the kitchen and said just make a great meal.They dirtied up every pot and pan and used up every ingredient in the cabinet, but the food that comes out isn’t that good, you know, we’re just hungry and and angry that the kitchen’s dirty.
So goa little deeper there. What do you, what do you mean by that? Yeah, I mean what you’re saying is this is not a simple case of applying the technology. This is really rethinking, use the word rearchitecting the corporation.
Yes, we need to rearchitect it and we need to make the recipe.It’s we have to really take apart what we’ve put together now already and we need to look fundamentally at what the skills of the people are, what work needs to be done, and decide what do we want to make, what is the outcome.
And before I go to Yale, I’ll give you a brief commercial moment. Explain how Lattice helps in that process.
Oh, so Lattice is the platform that really looks at how the people perform.And we’ve got over a decade of experience of knowing what helps people be successful and so what we’re doing is bringing this all together for people plus AI and rearchitecting the work so that you can really have that recipe for success. I want to
come back to that before we get done here because it’s a really important point. But yeah,
so if, if I answer the question correctly, I get an advertisement moment too. You
might, yes, you, I, I will give you an
advertisement I’m gonna try harder. I didn’t know
that.Is that a prize?
Yeah, the, the prize is like if you answer the question correctly, then you get 20 seconds to
talk about, you get to promote your business. Yeah, but, but, so tell us first why the big gap between the potential and theactual application,
um, with all this we have a.A lot of public board members here and CEOs here and then this is something your technology team wouldn’t want me to say, but I’m going to say out loud and because I’m, I’m a trader, um, that, you know, technology is actually the, the easiest part of this whole AI game. Um, the, you know, our models have really gone beyond what, um, you know, what we are using it for. It’s, it’s very powerful. You can do anything as long as you can imagine it.So the dirty work, that really difficult work to, to Sarah’s point is actually the people part, and then the process part. And nobody wants to do the dirty work to actually want to train your people to really build up that AI fluency because it’s a really, really hard job. And second is, everyone is asking, well, how are you going to automatemy processes using AI how much like people you can take out, how much revenue we can do, but you cannot automate or you cannot optimize a process if you haven’t mapped it out. If you don’t even know what processes your companies are, uh, you know, doing in, in a lot of enterprises, I know that you can be invoicing.your clients, you can be billing your clients, but there are 500 different ways to do that. If you are not mapping out those processes, how do you actually use AI to do it? So that’s the challenge is that change management is the hard work.
Yeah, George.
I’m sorry. Yes, OK, go ahead. Wow, promote your company. OK, everyone.
Uh, so, uh, JLL is actually a Fortune 200 company. We do real estate. What does that mean? Um, the resort you are sitting, this beautiful resort is a building, offices, um, warehouses, data centers, they are all the world of JLL is real estate, and then we.Yeah, that’s all I need to say. And it’s,and
I’ll do a commercial for both. Uh, my last company that I built, JLL was our broker for the last building I had, and Lattice was our HR management for feedback.
That’s why you’re here. Sothere you go,
um, I don’t need a commercial for Grindr because 95% blend.Recognition unaided in the US, and a third of all gay relationships started on Grindr, so I sort of agree with you and so I don’t on both sides. That’s fine. On the one hand, I actually don’t think technology, meaning the models themselves, are getting that much better anymore. They might be getting a lot better.The stuff they have not released, meaning the labs are not released, but the stuff that we’re seeing over the last year and a half, has it really been fundamentally that different? I don’t think so. In my, to me at least, there have been like two really fundamental moments when Chad GBD came out. Like that was a really big fun moment 2022. And then secondly, when the coding agents started to actually become really, really good.Uh, and that’s when like work really started to transform. I was at a conference maybe like 2.5 years ago, um, similar to this one, different, uh, different brand, and I said that, that, you know, soon we’ll have synthetic employees and we’ll be managing them, and everyone got really upset with me. I was right that we will have synthetic employees and people don’t get mad at me anymore, but the challenge is that to manage the synthetic employee, and I prefer the word synthetic to.Agent for what it’s worth, you actually need to be a, I don’t know, it just seems more real. I don’t know, uh, synthetic seems more real synthetics, OK, um, uh, and you need to be a much better manager than you are for even humans because they need a lot more direction and, and context and the unfortunate reality is that most companies are terribly run.Like I have this theory that you could take an average big company, fire about a third of the employees that of the company, and nothing would change. But
you’re sort of saying the same thing that Sarah and Ya are saying, which is,
which I’m saying I sort of don’t. I’m just being a little more rude about it.You’re doing I’m being more direct and rude about it, which is that like we have a problem which is that managers need to manage this stuff and most companies are really badly run, and that’s why it’s not happening because in, in our case like Grinder is insanely lean. We make about $3 million in revenue per head, 1.2 million in EBA per employee, um.Our engineering productivity has 3xed in the last year,
so you feel youare getting full. Yeah, I mean,
I have a 200% engineering team now and I actually have 60 people working, um, and, and we have another problem which is that we can’t keep up with how fast engineers in our coding and we can’t give them enough work to do because not everyone else can keep up with.
How about in,is it only in coding or how about in other parts of the.
It’s Mostly in coding. I think design and product management are going to be next, but where I do agree with you is the tools that you need to apply AI to your work are really amazing for coding. They’re not yet amazing for product management, for example, which makes it harder for product managers to become as AI native as you want them to be and to keep up with what engineering is doing.I think product will be next along with design. Finance, I think, is coming along OK too, but they’re slightly behind. But eventually every function is going to have the tools that you need, and then it’s going to be just enforcement. Like a year and a half ago, my engineers were really opposed to this idea of using AI to write code, even though all the tools are available to them. Like we don’t have this problem of like standing in the way. Like everything was available. They didn’t want to do it.We had an all companywide in-person meeting and I said, look, we’re going to do this, and if you don’t, then there’s not going to be a place for you here.And we got them to do it and the outcomes have been totally incredible.
But George, don’t you think that because Grindr is a digital native company, is a technology company, most of your workers are actually, um, you know, tech workers. And we all know that most, um, most of the ROI we are seeing on AI are because of the tech industry are benefiting.And that’s why it’s actually, that’s why I feel like there’s something that needs to be happening in the enterprises level, where there are a lot of enterprises, where, for example, take JLL for example, we have 110,000 employees across 80 countries. Our tech team is only 3000 people versus the 110,000. So I can make the 3000 people very efficient.It’s a drop in the bucket for us. So for enterprises that actually really see value at the bottom line, it needs to a different kind of game. It’s not just coding inefficiency. It’s really changing how we do business, how business processes are being conducted. So it’s not obvious for most of the enterprises out there.
I think that’s a really, yeah, I think,
I think that’s a really great point. I’m hearing this from a lot of, uh, uh, uh, a lot of CEOs. We basically have companies that were built for the 20th century, and we’re now, we’ve now been thrown into the 21st century where the pace of change is so fast and the corp, the way.corporations are structured. I think this is what you were saying, Sarah. It doesn’t allow for the speed of change that you need to take advantage of these technologies.
Itdoesn’t. And also there’s a, there’s a reality that to George’s point is that the workforce has changed. It’s not just people, it’s people and AI.By 2030 that’s gonna be about 5% of the workforce in terms of um about $3.3 trillion of digital labor in the workforce and what we need to do is change the way that we work and how we work with people plus AI together with the most important thing to Yao’s point is that we put the success of people as our primary objective. It doesn’t mean work doesn’t change. It doesn’t mean that.It just doesn’t mean that things don’t change. What it means is that we have to change the way that we work so that we can succeed. But, but Sarah, how, how, well, um, first off, you talked about management.Um, I also, it was about 2.5 years ago we put into our product the concept of a digital worker, and the idea was that this is an entity that’s coming into your organization that not just for, for engineers and coders but for all workers at technical and non-technical roles in companies need to understand the work that’s being done.Imagine if a stranger came into your, you know, home or your work or your job and just started doing stuff and you didn’t know what they were doing, what they were accountable for, how they were being greeted.That doesn’t make you feel good on the other side and it doesn’t help you run effectively as a business. So yes, we really need to fundamentally rethink and stop thinking in silos, more thinking in systems and projects in business function and business outcomes that we want to achieve, and that is every CEO’s fiduciary responsibility is to look at how we drive the business outcomes and to your point, a lot of the inefficiency is in the back.Office work it is in the the systems that we have collaborating and engineers think in systems by their you know work in nature and so we have to get that across the entire organization across HR across finance across operations across marketing across sales across engineering product design legal everything and really be an iterative
system you’re nodding yeah yeah
so definitely I, I can make um.I’m going to give you an example, like, make your point more concrete in the sense how we do it at our company. Um, so there it’s actually really simple as people process technology first, on people, we started 3 years ago and give everyone an AI tool. It’s our own JLGPT, which is a chat GPT but our own version of.Um, the goal is to reskill people, and then increase the AI fluency of the entire company. So we have about 55,000 office workers and 25,000 of them use it uses the AI app every day. It’s, you know, think about this real estate people, not tech people.Second, on the processes, we started at the beginning is like technology people being so excited about what AI can do. Look at this amazing demo that I can do that tells me where to get, you know, lunch. But then we started to work with the the business leaders to really ask, what do you want to solve? Like, what is your business problem? HO like the way you should develop technology.Based on your strategy. The first thing, the third thing is, we are very much focusing on our data. Because at the end of the day, the adoption of AI, the usage of AI, the application of AI is going to get easier and cheaper. And then everyone can do AI. Everyone can use AI as a point. But the data that you have is something that only you have.And then that’s your company’s competitive advantage. So you have to build out the data layer to make sure that you have long lasting competitive advantage, advantage compared to your competitors. So those are the three things I think can be applied pretty broadly.
Can we, can we talk about agents? That’s been a big topic for the last 24 hours, how all three of your companies are thinking about agents.And adapting to agents, George,
you mean to let other agents work with us?
Well, both your own agents, but also are people going to send an agent to find themselves a date on Grindr? Are you,
so we are building our own agents that would help you find the people you should be talking to and kind of function as a wingman inside the op. That’s something we’ve been working on for like 3 years now and.There’s been a lot of infrastructure built for it and it’s moving along really well. It’s complicated because Grindr has a lot of unstructured data. Grindr users send each other 50 messages per day.Um, we produce like hundreds of billions of messages a year and so, um, that, that is the most valuable data we have, but it’s also completely unstructured and so a lot of the work has been on how to like get that into an agent to be able to do the things that you wanted to do that’s coming along well. I’m pretty disinclined to like, is
it in use on the platform?
people, a small number of people, and they go in and have conversations with this agent.
Um, but for, um, third party agents, I’m disinclined to allow that to be used on Grindr because Grindr is ultimately about in-person connections and I don’t know if I want, I’m comfortable with, you know, somebody else’s, you know, muse talking to somebody else’s instinct and that being the outcome of how people meet each other. I’m not sure our users would like that.I’m not saying they wouldn’t like that TBD, but my instinct today would be to not jump into that, um, and create that risk. I think that’s very different from like, I’m not allowed to use an agent to go and shop on Amazon becauseI, I I’m still shopping on Amazon. I think it’s actually hurting Amazon to not letme,
but you, you want to keep it personalconnection,
yeah, and I think our users would want to. I mean, we also have like insanely private data, and we are really careful about the data ever leaving our, we’re,
we’re gonna come back. We’ll come back to that in a minute, but Sarah, how are you thinking of, how does Lattice think about agents versus people and the relationship between the two?
Well, first off, I wanna go back to something you said earlier. There’s been moments, uh, chat GDP chat GBT was a moment. Cloud cowork was a moment, and agents, um, in terms of personal assistance is, is really a new moment and what it means in terms of, um, an exacerbating lack of understanding and education that people have to the capabilities of AI that’s just like becoming very pervasive.Um, but internally for agents we look at agents, um, in terms of how it helps to drive better human connection and so this is what our platform does. I’ll give a very, uh, concrete example that I think everybody here can relate to. You are in an organization where you have to do, uh, an organizational change or a strategy shift, you know, it could be a reorg, it could be a layoff. It could be just a change in direction.And that we think of, OK, here’s the strategy, here’s the information, maybe we have to change some reporting lines in a back end system.There’s tactical things that have to happen. There’s a lot of human things that often go unlooked, and those are the reasons why they go south, meaning you have to have the right cascade of comms. You have to have the right FAQs. You have to have all the I’s dotted and T’s crossed. You have to have met with all the right stakeholders. You have to make sure all the right stakeholders are invited to the meetings and you haven’t forgotten the person that was just contributing.That wasn’t on the team and then after that happens you also need to do the right follow up that’s bespoke to the people and then be able to understand where the problems were and then follow up on those follow ups that’s how success happens in terms of making change happen at a company and it happens I think probably every week or every day at our companies.
And I’m sorry, so where does the agent fit into that? So
the agent does the work for you in terms of making sure you’ve dotted the I’s cross the T’s, so that it identifies, you know what, Sally really needs to talk to Tim, and they need to work this out because this is a problem. So what we do is we have the agent do all the things that our brains just can’t handle the cognitive load of knowing everything that needs to happen.And so when we have the agent be in service of the success of the human and the team and ultimately the business we call this the rhythms. So what we’ve set up is we’ve identified it’s a subset of less than 100 rhythms from a project kickoff to a one on one conversation to an organizational change to a skills analysis, the way a company runs that instruction book, that recipe manual.And by putting that in service of humans then the right people have the right conversations. You don’t have the confusion. You don’t have the miscommunication. You don’t have the misalignment and you have a team which is understanding and aligned and ready so it helps human connection. It helps. That’s a very 3 words better than my
long explanation great. Uh, yeah, how are you using agents at JLL?
Yeah, so JLL as a real estate company, um.The, the, the key of our business is actually people that magical moment that someone was asking is the people interaction.And so we want to leverage agents so that we can make our people to focusing on other people, our clients. And so what we have, what we are doing, for example, um there are a lot of repetitive work, we, um, our annual revenue is $26 billion. And, and, but we spend about $100.billion dollars on behalf of our clients on procuring services, such as fixing the HVAC, um, you know, buying toilet paper, super unsexy. And then we have to processing $100 billion of those bills and invoicing people and getting paid and collecting those. Um, who wants to do that? Nobody.Um, nobody go to college and say, whoa, I want to spend my day to build people. And maybe someone do. I’m not judging. But, um, but those are the things that are not the most inspiring things for people.To do, but I think that’s what agents can do. So we are focusing a lot on really taking out those repetitive work nobody wants to do. And then a very well governed agents can really put that into, we can put them into work.
Yeah, makes sense.A lot of talk here over the last day about cybersecurity and the rising cybersecurity risk. How big a problem is that for Grindr?
It’s probably the thing I’m most scared about right now. Yeah, we had an audit committee meeting this morning because earnings is tomorrow and usually security is like a 10 minute conversation. I think today it took 45 minutes. I think it’s a massive risk and a massive threat for most companies, and I don’t think most executives are at all ready for what is about to come.Um, and so I’ve been, I mean, uh, my team has been on this a lot the last year, and we’re basically having to build a lot of internal defenses that previously you would not have needed. You’re doing it yourself. You’re not, I mean, we’re going to use third parties to help us, but there is not a, there’s no cursor of security, meaning like there’s no company that like has figured out how to use AI.To build a security layer that is fundamentally unique and different, there’s probably like 20 trying, but no one’s yet had a breakout moment. I’ve spent a lot of time talking to VCs about this to see if there’s something I haven’t missed, and there isn’t one, basically. And so you will need to build kind of your own foundations, and I think ultimately it’s about like having your own agents that are almost acting like hackers on your code base constantly trying to find vulnerabilities and fixing them.So that you can fix them before somebody else discovers them and fixes them. That’s probably the path forward. And right now there is not a tool that you can like plug in and deploy. You have to do that yourself.
Uh,we’re, we’re just about out of time here. I’d like all three of you to give your one piece of best advice to people out there who are following, who are going on this journey of trying to figure out how to use this new wave of technology to make great companies, yeah.
Um, I will say, you know, do the hard work to build a foundation before you scale, you actually will find yourself move way faster than actually slower,
move slow to move fast. Yes, Sarah.
I’d say start with the business objective of your company and look at how you’re gonna make the people successful and bring them along through education and alignment and the technology is is there to instrument and make all that successful.
My general sense is that this transition is going to create incredible economic opportunities, so the faster people get on with it, the better off they’ll be, because if not, they’ll be left behind. I mean this is going to be a digital revolution all over again, right? Like companies that went in and build digital capabilities quickly, did really well. Look at Capital One, and those that didn’t are not doing so great.
George,Sarah, Yao, thank you very much. Best panel of the day. Yeah, there you go.
