Subscribe
Share
Share
Embed
AI is changing how we work, but the real breakthroughs come when organizations rethink their entire foundation.
This is AI-First, where Box Chief Customer Officer Jon Herstein talks with the CIOs and tech leaders building smarter, faster, more adaptive organizations. These aren’t surface-level conversations, and AI-first isn’t just hype. This is where customer success meets IT leadership, and where experience, culture, and value converge.
If you’re leading digital strategy, IT, or transformation efforts, this show will help you take meaningful steps from AI-aware to AI-first.
Robert Entin (00:00:00):
Remember in 2000 talking to people about funding our business to be internet capable, and I talked to a lot of VC guys and they said, you can do this in six months, right? Sure, we can take an ERP system that's legacy based in a mini computer world and make it completely web-based in six months, right? But they weren't interested because they said everything's going to be changing in six months. Well, I would tell you, I estimate that that was 10 to 15 years. With ai, it's very different. The infrastructure to allow it to operate already here, it's moving and it's a much faster rate than I think the 2000 revolution.
Jon Herstein (00:00:38):
This is the AI first podcast hosted by me, John Herstein, chief Customer Officer at Box. Join me for real conversations with CIOs and tech leaders about re-imagining work with the power of content and intelligence and putting AI at the core of enterprise transformation. Hello everyone and welcome to the AI first podcast. I'm your host, John Herstein. On this show, we talk with leaders who are putting AI and intelligent content at the center of how their businesses operate. Today's guest is one of the most respected technology leaders in commercial real estate. Robert Enton is the executive vice president and chief information officer of Vernado Realty Trust. A 20 billion read behind some of the New York's most iconic office towers, including the transformation of the Penn District. Robert has led Vernado's technology strategy for nearly two decades, evolving it from a legacy infrastructure to a cloud first AI ready enterprise.
(00:01:34):
Before joining Veneto, he founded a real estate software company giving him the rare perspective of both innovator and operator. In our conversation, Robert opens up about the real world wins and challenges of using AI from automating rent invoicing to exploring how AI could unlock insights buried in complex lease documents. We'll also talk about the future of self-managing buildings, the evolving role of knowledge workers and what it really takes to drive digital transformation in commercial real estate. This is going to be a candid, sharp and future-focused conversation with a leader who's not just watching the AI wave. He's helping shape it. Robert, welcome to the AI first podcast.
Robert Entin (00:02:16):
Thank you very much, John. Great to be here, and that's quite a weighty introduction. I'll try and live up to some of that.
Jon Herstein (00:02:22):
I know you will. Now, for our listeners who may not know Vene well, can you briefly introduce yourself, your role and the scope that you oversee as EVP and CIO?
Robert Entin (00:02:31):
Okay, so vdo is a large commercial real estate investment trust and having evolved over the years through a number of stages. At one point we were much larger than we are today. We had the largest portfolio in Washington dc We had a large retail portfolio. Obviously New York office is our prime asset base. We do own the merchandise, Martin Chicago and 5 5 5 California Street in San Francisco. Two very large great assets. They're just outside the portfolio. But after a number of years, I think the analysts were, what do you want to be when you grow up? Are you an off the street? Are you in Washington Read, et cetera. Management has over the years, spun off two public companies out of vdo. One is JBG Smith down in Washington DC which is now one of the largest if not developer in that area, and Urban Edge Properties, which was our old retail division.
(00:03:24):
So then now Vedo ends up being certainly not small with really the best assets. Our big assets are obviously in New York City multi buildings and the focus was always when we looked at New York, the focus was always with Hudson Yards being developed over the past decade into that really beautiful space west of the Penn Station, we were sitting on about 10 million square feet of office space right in and around Penn Station, which as you may know is 600,000 people a day travel through Penn Station every day making it the busiest transportation hub in North America. And so the management has always been focused on when is it time to invest a massive amount of capital in the Penn District. So really where we are today is, that's the biggest thrust of our initiative is that, and we redeveloped the Monaghan train station several years ago actually during COVID, which if you've come to New York, go to that train station, it's absolutely magnificent.
(00:04:22):
The old post office building. And by the way, Facebook leased the entire back end of the building, 530,000 square feet for office space. So it's a very, very hot area. And then we spent an awful lot of money in both renovating Penn one and Penn two and they came out. They're just absolutely beautiful and we also knocked down Hotel Pennsylvania getting ready to build Penn 15, which will be another very large beautiful office building. So our big bed has obviously been on the Penn District and New York in general. You come to New York, you definitely want to be around Madison Square Garden because you wouldn't recognize it if you haven't been there in a few years, you absolutely wouldn't recognize it. Says a little bit about vdo. I started my career as I'm a software engineer. I'm a computer science engineer by education and had gone into the ERP space for real estate property management at a company for many, many years and through an odd story tornado and IBS crossed paths.
(00:05:20):
I certainly very friendly with senior management. They ultimately acquired my company to bring us in house to help run the enterprise. I've been through a lot in terms of seeing the company come from the old legacy technologies where all we were doing was accounting, building automation was not much of a budding young silo in the industry to today obviously where we're today with the internet for web services, microservices, everything about the way we all do business is very, very different today. So it's been a very interesting journey, continues with the stuff that's ahead of us is probably more exciting stuff that's behind even
Jon Herstein (00:06:01):
More exciting than anything that's already happened and still lots to do. How would you say that that background of having done a startup and then being acquired and then kind of coming to tornado becoming CIO, how has that influenced how you operate as a CIO today?
Robert Entin (00:06:15):
That's an interesting question, Sean, because a lot of people when this happened, which is nearly 20 years ago, everyone said, I don't understand. You started as an entrepreneur and now you're going into corporate America. It's usually the other way around. You start in corporate America and then go the other way. So it's been a little unique in the sense that I was always a vendor dealing with customers and when the customers closed their doors at night, I was at best an advisor. I wasn't dictating what they do or how they do it. I was advising we were selling services. But what was interesting is now being inside the four walls and being the person who's actually responsible, you have to actually eat the dog food you make. That was a very interesting facet of it because we would write all this software, we would ask users to use it, this is the business you're in, and sometimes they use it the way you think they should and sometimes they don't.
(00:07:04):
So when I got inside the four walls, I would say, wait a minute, the stuff that we thought was great, some of it wasn't so great. Other stuff that we didn't think was okay, they really like, it's almost like if you're building, we were building these large international harvester machines that could harvest 17 acres in an hour and you build it and design it and what do you do? You take it out in the test field behind the factory and you plow an acre, this is a great machine. Now you sell it to farmers where they have to plow thousands of acres and they come back and they tell you it doesn't really work the way you thought it would. That was a lot of what I experienced now being on the inside. So something from something that was smaller or we really could move fast and we could change things whether you want it to change them to being inside of an ecosystem where bringing about change and getting people to adopt the things you want 'em to adopt. And of course now being able to interact much more closely with the business, the end users directly to understand what their problems are and were I found, I think that my experience prior to vdo dealing with lots of different companies and lots of different user groups really helped me once I got inside vdo.
Jon Herstein (00:08:16):
So when you look at everything that's on your plate now in this kind of massive organization, what gets you most excited about leading technology at vdo today?
Robert Entin (00:08:25):
I think our senior management is extremely, I don't want to say irreverent, but senior management doesn't necessarily go by anyone's playbook. They're very innovative and when they have ideas they want to act on them quickly. So I think what's been exciting for me at vdo is that there is something about being in a big corporation that is definitely a little different than a smaller company, but as a big company we're in when they make decisions to do things, they want things to be done quickly. They don't put up a lot of bureaucratic boundaries to get things done. If the answer is we want to get this thing done, then don't figure out a way to do it. So that's been one of the things that's been fun and exciting and the idea is the diversity of areas where management thinks about innovation is also very refreshing
Jon Herstein (00:09:12):
And we're definitely going to talk about some of those areas and I actually want to zoom out a bit and talk about's broader digital transformation journey and when you all use the term transforming real estate through technology, what does that really mean in practice for you? So to kind of dig into that, let me start by this. When you think about verna's digital transformation, how have you define that journey and what does it really mean to say transforming real estate through technology?
Robert Entin (00:09:36):
I mean that could be a lot of different areas. I look at my job, which is divided into multiple sectors, right? There's the accounting software and what's the software that runs the corporation? What's the operational software that runs the buildings? What's the stuff inside the buildings, the real top tech, the IOT stuff inside the buildings? So when you think about transformation on the basics of you want to automate more and more of the stuff that runs the corporation, largely accounting, property management, job cost, construction, tenant, FedEx, that sort of thing. That's kind of a slug it out in the swamps like every organization faces. How do you get the administrative part of the organization to have better workflows, to do things more automated? But that's really the much more mature side of the business. We've all had those systems for years. It's really been the last 10 to 15 years where with the advent of the internet and the advent of converging networks inside the building, and I guess again it's the microservices, it's the internet, it's handheld devices, really changed a lot about the way buildings can operate.
(00:10:48):
So the transformation, we could talk a little bit about the transformation of the administrative area of the business or property operations or soon to be the ai. What are we doing with all the content we have? How can we make the content that we have allow us to operate more efficiently or give us better answers? Things that would take people hours to compile theoretically could take minutes to compile. And that's I guess the whole general AI overview on all the contents in our organization. The buildings are probably the most exciting area, or at least the area that's had the most inertia for change because if you go back again, 15 years ago you had elevator systems, fire and life safety systems, turnstile systems, camera systems, everything in the building were on different networks. They weren't even really networks. Some of them were just hardwired pots lines to fire and safety and elevator systems and such.
(00:11:41):
But as the internet started to invade the building and all of a sudden the answer is we could have a lot of the things inside the building really communicate with each other. Converged networks inside the buildings became the thing. We were pretty far ahead of that and the idea is if you build the highway where everything inside the building is on a converged network, then the answer is the software applications that can run the building, whether it's automated visitor management, you swipe your badge or facial recognition of the turnstile. Turnstile opens, you go to an elevator, the elevator ought know what floor you're going to because it knows where you are as a visitor. All those things in the PropTech space which have really been exploding over the last 10 years have probably been the area where you've seen the most inertia. Still today I would say that the engagement and the takeup has been somewhat slow, but that's always the case. The technology is way ahead of the implementation in general, but that's where in terms of transformation, you're seeing by far the biggest up to date you're seeing right now where all the momentum is. I think that might shift inside the office with the administrative burdens being taken away by AI coming forward, but that'll be, we'll get to that in a little while. That's an interesting thought.
Jon Herstein (00:12:57):
I'm curious, when you have the capability of these converged networks, does that mean that you can begin to move faster as new technology comes out? Does that influence your overall strategy, having that kind of capability?
Robert Entin (00:13:10):
Well, it's an enabler, right? Having a converged network allows us that it doesn't matter and we manage the converged network. We don't actually typically select or put this stuff inside that. What I say to the people in, we have a technology group in New York that's really focused strictly on the buildings and our agreement is we'll manage the network and protect the network from a cyber and you put all the stuff inside the network you want and then we'll vet it and make sure it operates. And I think our job is to really be enablers. So when they say, look, we've got a new application, we have a tenant amenity app that our tenants walk around with and more and more of the things that are going on inside the buildings, they can affect from their phone service and that requires two things. It requires the right third party vendors who are providing the software and the right connectivity and security to enable that stuff to happen.
Jon Herstein (00:14:05):
Got it. Makes sense. So pivot a little bit. You've alluded to content a couple of times we moved from maybe networking to content and then we'll certainly get onto ai, but you all started the move to at least boxes cloud almost a decade ago and it was pretty bold move then particularly in commercial real estate that I think was maybe a little bit slower to adopt cloud. So I'm sort of curious what drove that decision and how did you build support for it across the organization? Was it controversial? Was everyone fully on board or what did it take to actually make a move like that?
Robert Entin (00:14:37):
It wasn't that much of a democratic decision, but the reason it works is because the why at the time ransomware was in its early stages, it's very different today it's evolved to very different type of threat, but in its early stages where it's the attacks were typically pretty easy and pretty successful, but they were quick, they encrypted a lot of stuff. If you had good backups, which we always did, and you restore the backup, you get rid of the node. They weren't taking over your whole network, they were hitting a node and then you encrypting a bunch of stuff. But we got to the point where we said, number one, we don't know what's leaving our network on a typical Windows NTFS file system. The answer is, yeah, there are products you can put in place for DLP and for monitoring who's doing what. But it seemed to me that was going backwards.
(00:15:23):
You're already in a spot that's going to go away. All your content's going to migrate to the cloud over time anyway, so here we were with a ground-based network with various security problems, both in terms of protecting the data from ransomware as well as protecting the data from data loss and understanding where the data's going. That's pretty much why was the real driver for us to say the ground networks got to go. We've got to get everything that's on the ground into a content management system. We spent a fair amount of time looking at Box as well as its competitors, and the thing that impressed me the most was Box had the best vision for Box being at the center of a wheel and the wheel had lots of folks of functionality for how work gets done, where work gets done. So whether it was, whether it's metadata, whether it's box relay, whether it was the protection shield you had a really good vision of, it wasn't just, wasn't just OneDrive in the cloud in the sky because obviously Microsoft was a compelling argument.
(00:16:26):
You already licensed the OneDrive in the sky or Google Drive. I don't think anyone had quite the vision that boxed it. So we said this is where we want to go with respect to getting them there. It's probably before your time. It took us years, it took us a lot of years. There was a lot of resistance, but the only good news was again, they didn't really have a choice. The answer is we're moving. So some of the difficulties are that permissioning in a Windows NTFS world is really different than permissioning in a box world or any content management system. So you really have to start to look at how are you storing things, what your folder structures work, how the department chairs were what of home for those look like. We spent a lot of time on strategy as to how we were going to configure it and then we said, look, department by department, we're going to migrate.
(00:17:13):
We used a replication product to kept the cloud and the network in sync until the users felt comfortable. Then we cut off the replication and then they were in box alone and I guess it probably took us, I mean it's a solid three years. It was three or four years before we finally got the last bit of data to evaporate off the ground into boxing. The unlimited versioning. The odd thing is it's easier to protect your data in box in the cloud than it is to protect it inside your own network. That was the compelling argument for me that said, you got to get this stuff off the ground
Jon Herstein (00:17:48):
And in the early days of cloud, not everyone fully bought into that idea. So we took a while
Robert Entin (00:17:53):
To get you're, I used to run into that a lot was, wait a minute, all your stuff's going to be up there and you can get to it. I said, you don't know that They protect that for a living. Everything is encrypted. Not that we're not doing a good job too, but with the unlimited versioning and with the tracking of every single access to every file when it was open, when it was modified, again, your ground-based networks, that's not standard cable station. You have to buy a lot of layered products on your network
Jon Herstein (00:18:21):
On the ground to get that done. I mean, this was pretty prescient back then and now you're there. It took you a while to get there and I think we will have some time to talk about change of management, how you lead an organization through that kind of a change. But I'm curious, in what ways did that early, relatively early shift to the cloud lay the groundwork for the stuff we're going to talk about, which was the AI initiatives that you're driving today and in the
Robert Entin (00:18:42):
Future? I mean that's right on point. Not that we knew then, but now that all of our unstructured data, all of it is in a place where it's easily leveraged for tomorrow's AI technology, we're really in good shape. It's a combination of now listen, whether it was Box or Ignite or Google Drive or OneDrive, almost any of them at this point, you're certainly better off than it's on the ground. I think the advantage for me to looking at the box environment again is the type that it's not just content management. There's lots of layers of functionality that are hovering around box to provide increased functionality, whether it's, again, whether it's the workflow or box shield or box hubs, all of these things that are leveraging the content you have there are UR to our benefit at
Jon Herstein (00:19:32):
This point. So you've painted, I think a great picture for us of the journey, the foundations that you've put in place, and now I want to fast forward to the present and talk a little bit about where you're leveraging automation. Then we'll start to get into ai and I'm curious how this is actually showing up inside of, so some of the real use cases that your teams are taking advantage of, and let me start with this. Can you share with us a very concrete example of automation of tornado? One of the things we've talked about is automation of the creation of monthly folders and emailing tenant statements. Can you take us through one or two of those use cases and how that's changed day-to-day operations for the teams at vernado?
Robert Entin (00:20:10):
Well, I mean you're saying the rent built production, but that goes back a long way. That's kind of mundane at this point. But when you think of how rentals used to be produced for tenants, right? Once a month they get printed, whether we printed them internally, you go to a printing service, sometimes the printing and stuffing service lock boxes are in effect PO boxes for rent collections. But it was a very paper laden process and years ago people started to say, well, can you email, can we do this by email? We wanted to both email things ERP system producers a rent bill. You want to email it to the, but you also want it stored somewhere so that property managers, tenants, others can reference any rental bill they want at any time. That plus a number of other initiatives, we said, look, box is our platform.
(00:20:58):
Let's build a box agent. So what we did was we built a box agent internally, which is a fairly general purpose agent that we have that will allow us to, it's really a gateway basically to pull and push information in and out of box. So one of our first applications was the rent build process. What we can do is we can say that the rent build process, in addition to emailing rent build over tenants on a building by building basis, month by month, it will take all of those PDFs and store them well, but you can't just put them in one big folder. So it's smart enough to know that for each building, there's a folder under the building, there's a month in your folder and it automatically creates the folders, puts the rent bills in another property managers. When a tenant says, I either lost my rental bill or last February there was a charge I don't understand.
(00:21:48):
Instead of a long-winded piece of research, they go to a box folder, find the invoice, share it with the tenant. Having that sort of infrastructure just allows for automating it's call it simple workflows like that around the organization. It's been good. We're now actually almost done. All of our invoices have been scanned for years and it was a gause, which is an old document management, definitely a legacy document management system. OpenText bought it like they buy a lot of these and running it, but it's no place to store ear invoices. But it was easy because our whole scan center was hooked up to GAO and it was fed into our workflow. But we looked at this a year or so and we said, the problem is that these invoices are landlocked, they're dead now. We have about 2 million invoices scanned invoices going back many, many years and you don't want to lose.
(00:22:39):
So we have no paper. The paper goes, we've been without paper for years. That way by the end of January we'll be done with a complete migration of all 2 million invoices plus supporting documentation into a box structure and all of our internal programs that would allow you to retrieve those invoices. Now talk to the box agent to say, let me see the invoice. So again, having that agent on-prem, which is kind of a gateway and made enabling things like that much easier. And for us, one of the reasons that we actually did it, you probably are aware of this, is that with box apps, which is just another one of those things that floats around the box wheel with box apps and the ability to have the correct metadata attached to each invoice and then be able to do kind of intelligent searches, that was one of the compelling final organs to say, let's decommission guse and let's move everything up into box,
Jon Herstein (00:23:35):
Right? Because with metadata on those 2 million invoices, you could know what's the invoice number, what's the invoice date, the amount who the counterparty was, all that can be then served
Robert Entin (00:23:43):
What building was charged gl. So if somebody said, I want to see all repairs, bills, want to see copies of all repair and maintenance bills on pen one with box ops, the answer is it's pretty easy. Now with a good document management system and good metadata, you should be able to get that. With Gouss, we didn't really quite have that much because again, it was old. We're one of the proponents. He push you guys on metadata all the time. Unstructured data is great, but the more metadata and more layers of metadata we can put against that unstructured data, the much more powerful than it becomes.
Jon Herstein (00:24:15):
And one of these we'll talk about is how AI helps make that even more possible, right? Than when you had to manually go into that metadata or have some system try and do it with OCR or some other technology. So it's come a long way. So you mentioned a couple of examples where you've used technology to enhance the tenant engagement and services. Are there any examples where you've implemented some of these improvements and been able to measurably improve customer satisfaction or occupant satisfaction or the tenant experience?
Robert Entin (00:24:42):
We started getting involved in a tenant experience app before COVID, thankfully, and we were actually, so it's a very customized app for we own very high-end class, a commercial office building, highly amenitized, and it turns out the back to work argument. How do you get people back to work? Well in class A buildings, you have to provide a great environment with great amenities. We started our journey on a tenant amenity app a few years before COVID, which was very helpful because of course post COVID that became very important. But that I would say is measurable. So the tenant amenity app, because it's again, New York, one of the things that New York buildings are very heavily turns stock. One of the interesting arguments about tenant amenity apps in the industry is if you're in suburban office buildings in Nashville or some or Atlanta or other cities, why does the tenant need to walk around with a tenant, a landlord app on the phone to order food?
(00:25:40):
They got Uber Eats and they got plenty of places to order food, transportation. There are too many apps on their phones already. It was always the art. The good news about being in an urban area like New York where turnstiles are really at all of our buildings, you need something to get through turnt stuff and it was just typically a badge. Well, the badges are going away for Apple Wallet for NFC credentials for facial recognition, and if all they're doing is opening signal, but now that they have to use it for that, ordering food within the building is good. Reserving conference rooms and coworking spaces is good. Getting your car is really good. So the more amenities you have in the building, the more important it's for a tenants to be able to access those amenities in a clean and efficient way. The VNE app, which is called live work do, it's a branded app.
(00:26:29):
It grows all the time and has a lot of capability in it, largely because our buildings, our campus around the Penn district has tons of amenity and those amenities and those amenities are really important to tenants. So can we measure the satisfaction in a dollars and cents point of view? Yeah, probably not. We can sure measure the results based on the demand for our space. The leasing activity is really is great. Both the capital we put into the buildings, the buildings are great, the amenities are great, but I think the tenant experience clearly is also a factor.
Jon Herstein (00:27:04):
So it's great because you use the phone as a badge as essentially a key to the building in a sense, and then once you have that app on there to perform that function, you have a platform that you can provide a lot more service to your tenants and drive business and engagement within the property.
Robert Entin (00:27:21):
You invite visitors, you put in work orders, you can get your car, you can get food. We have conference centers today in a large office building, you really have to provide conferencing. Coworking is a different business, but certainly conference centers and being able to reserve conference rooms and conference facilities right from the phone as opposed to logging somewhere else is good.
Jon Herstein (00:27:45):
So we touched a little bit on this idea of adding tructure or getting structured metadata from unstructured content. We talked about invoices, lease documents and so forth. Maybe let's pivot now to AI because we really see this as the great enabler of getting structure or getting insight from these unstructured documents. So how are you looking to use box ai? What are you doing already? You've been in a very early adopter of what we call intelligent content management. What does it look like in the early days of where you're going to be able to leverage these capabilities?
Robert Entin (00:28:16):
Well, we're saying with metadata over documents or just generally the unstructured data?
Jon Herstein (00:28:20):
Well, specifically being able to pull both insights and structured metadata from your unstructured contents using ai. So I
Robert Entin (00:28:29):
Think there are two pools, right? One is you've, we've got just tons of unstructured data in the, like everybody, all your content's in one place and varies across so many different areas of the business. So regardless of metadata, the question is how can you leverage AI to make the business more productive? So we have really great senior sponsorship right now. Senior management is manic over. We must deploy this. We must figure out how to make our business more efficient, and that's been a great godsend to us. That means that we won't have any trouble when we find a use case as we're finding them, we'll have senior management backing, but on the one side of the world, which is your not metadata driven, just all the unstructured data. So I'll give you a few examples. We have massive numbers of loans on all of our portfolio, literally thousands of documents documenting our loans.
(00:29:26):
People ask questions about these loans all the time. We created a box hub for loans. The good news is our loan data was in a really well organized folder structure, a great name. So it was really AI ready because AI is going to be probably one of the terms that's hugely important in trying to get end users to understand. So we create a box up for loan documents. We go to the people in debt and acquisitions and we say, give me the top 10 questions you want to ask and make 'em as hard as you want. Every single question it knocked out of the park, where's the Bank of America in our loans? What's our exposure? What's this property? What's the amortization? What's the interest only how many fixed versus variable? You name the question, AI was able to produce really good answers, really active answers.
(00:30:16):
Think about 10 Qs in 10 Ks. You guys file 10 Qs, 10 ks all the time. So we have a hub for a world over of 10 Ks and 10 Qs. So somebody can walk up to it and say, give me a table of our EBITDA for New York office segments over the last eight quarters and give me a graph. Or there's just a billion questions you can ask that are asked all the time. And today when those questions are asked, it means somebody in our accounting group or someone in the SEC group or the fp a groups, one of those groups will go, and today all the documents are there, but they manually go and do stuff like that. There are just so many applications like that around the business. Here's an interesting one that we did. You have earnings calls, Aaron does his thing.
(00:31:00):
I don't want to admit that I don't listen to them, but so we took all the transcripts and all the analyst reports from all of our earnings calls in the last two years till we built a hoe because oftentimes somebody's going to say, John Smith, the analyst at Smith Barney had this, what did we ask the last three quarters? Who answered it? Would they answer? Those are things that people will typically, when you're preparing the earnings call script for the next earnings call, you're going to ask stuff like that, right? And what did we say about the Penn District or what did we say about a refinancing at this property? And now to just have a hub that you can walk up to and just ask a question and you get a perfect answer. Like I said, the applications like that, having nothing to do with metadata, just saying you've got some of these documents, board memos, deals sheets, offering memorandums.
(00:31:50):
They're just so many important documents that are sitting in your unstructured data that being able to have them in a place that you can organize them, create a hub, and then put a good AI engine on top of it, will give back many, many hours to workers, knowledge workers in the business. No question about as you're working with, we're a beta customer on this and we can't eat the dog food. You give us fast enough and the work with box hubs and stuff, because we're really in the middle of, I don't want to say a war, but we're in the middle of making these hubs, making training these agents and getting them out into the end user's hands. And so we're really in the beginning stages of that, where the end users are starting to play with it. We've been playing with it for a while and we're very happy with the answers, but when the end users start to really work with it and say, yep, it really helped me.
(00:32:39):
We have one of the guys in fp and a after every earnings call, the analyst reports come out right after and overnight he has to put together a summary of who said what about on the earnest call and where we're going. And it's a very short timeframe because management wants the data very quickly, and that was being done manually. We put all the analyst reports into a hub. We trained it on what his output was supposed to look like, what his comment was, it doesn't make a finished product, but it gives him basically the data in a form that he can easily take it. And safe syn talk, we're looking at the tip of the iceberg. November, 2022 was a watershed moment for the world because open eyes said, Hey, look at what this thing could do. And everybody said, oh wow, the world's going to change in a minute.
(00:33:31):
Well, it's already 2025. We're three years later, and it's really starting to change much faster now, but I think people didn't understand because they just spent five years and 30 billion trading this thing on the internet. So yeah, it does great things and LLMs really are cool, but going from there to give us something in our business that's going to really help leverage and run our business, that's a different thing. So things like box AI and in every product we own and every product you own, AI is being introduced into the product at this point and some more effectively than others, but you guys are on an arms race right now, like everybody to make that work as seamlessly as possible. And I think three years, fast forward three years to where we are today, it's pretty fast. The expectations were just silly like they were in 2000 by the way.
(00:34:25):
And I would say that transformation, when you think about 2000 to, I remember in 2000 talking to people about funding our business to be internet capable, and I talked to a lot of VC guys and they said, you can do this in six months. Sure, we can take an ERP system that's legacy based in a mini computer world and make it completely web-based in six months, but they weren't interested. They said everything's going to be changing in six months. Well, I would tell you, I estimate that that was 10 to 15 years, probably about between 2010 and 2015. And a lot of it was this, right? The guy who's not alive to see it anymore, but where you could say, I want a car to pick me up at this corner and take me there, even though it has a driver in it. It didn't happen until all the components came together to make it work. With ai, it's very different. The infrastructure to allow it to operate sort of here, it's moving and it's a much faster rate than I think the 2000 revolution.
Jon Herstein (00:35:30):
Totally agree. I mean certainly the pace of change is just unlike anything we've ever seen and you've already taken advantage of some of those capabilities. But I am curious, beyond documents, where else do you see AI and value?
Robert Entin (00:35:43):
Let's go to the metadata for a second. So my dream has been, because talked about it and I've talked to Ben about it a lot is lease data leases. We have thousands of leases. These are not boilerplate leases. Some are, but some of our leases are hundreds of pages. Very complicated. If you're renting three or 400,000 square feet and it's a 20 year lease or 10 year with two five option, whatever, a lot of terms and conditions, our job in running real estate is take these leases and first extract all the financial data out of them, put 'em into our ERP systems and let's run the buildings, let's build the tenants, collect the rent, do the escalations, et cetera, right? I don't want to say it's the mundane part of the business, an important part of the business, but that's stuff we've been doing for years.
(00:36:27):
You have run the business. Then there's all the other unstructured data in the lease, renewal information, expansion information, termination information, signage rights, insurance requirements and stuff that you really need to know about. We've been a little ahead of the game. We've had a web-based abstract system which relied on human extraction. So we get a lease that it would take humans, it would take a person a day or two to get through the whole thing, and we would extract about a thousand fields of information about the lease, and we extract a thousand fields of information about the lease into a relational database model, Oracle database with probably 25 tables, relational tables. So throughout the years, if somebody said, tell me who in Penn one has options to expand on the 15th floor, or people in lease administration could run a query and get an answer and just questions like that are every single day in a large commercial real estate organization.
(00:37:32):
And the question is, how do you answer those questions for us who again, we've been kind of a forerunner. They said we've been abstracting this stuff for years, but why can't we take all the leases, throw 'em into a bucket and go ask the question, ask that question. How about this one? Who in the Penn district has an option on this? Or who has this renewal, right? Or who has this cancellation, right? Because now you're not asking about a single lease. You're asking about a collection of leases across a collection of buildings and the ai. How does an AI agent really understand all this relationships? We've done a lot of experimenting with this, taking a single document in any one of these AI engines saying, summarize this document. Sure, the real pot at the end of the rainbow for real estate is going to be the whole lease bucket of leases, parsed, vectorized in a way that you can ask complex questions and get really good answers.
(00:38:28):
The only way that's going to work in my opinion, is if you have the extracted metadata already out of the leases. So yes, if I have the extracted metadata attached to the leases and I have the leases, then I can envision an AI engine on top of that. Being able to ask a complex question, be able to figure out I'm where the AI agent's going to get its answer from the metadata referring back to the lease to give a really cogent answer and then bring up the lease language that shows that. We did a couple of experiments, one with AWS one with Google where we sit where they said, oh no, don't worry. We could just take the leases, throw it in. This is even before box hubs, right? And the results were spectacularly disappointing because we said to them right up front, this isn't going to work unless somehow you train this thing to know certain things.
(00:39:23):
Like you can't look at a LEAs and get a lease start date. It's not in the lease. When we were doing some of these things, we said, Hey, hey, knock, knock. We actually have the answers to all these questions. We have all these Oracle tables, so we could give you in some structured format, we could give you the tables and the lease ID and we could show you the documents. And if you could marry those two things, you can get great answers. It never came to pass with Google. What happen was they took the structured data, the structured version of the unstructured data and put looker on top of it. Great product Looker was really cool, but it's just a bi tool, not just, but it's a BI tool that you can get stuff out. And then the other side of the pod was the LMS looking at the leases, never joined.
(00:40:07):
I always argued with Aaron over the years, I don't like your will because you're building it. One spoke at a time. So until it's all done, I'm really interested in this spoke out here, so why don't you just get this one really? But the answer is no. You're pressured by a lot of people and you're building out that way. But when you think about box apps, so box apps comes out, it's not a full blooded document management system, but it's a pretty cool way to put metadata on top of your unstructured data. You can really retrieve things well. And then lo and behold, if you then get box apps type AI in it, which you're doing in box apps, I'm going to let the AI look at both the metadata and the documents. You've got the answer. So what we're looking at right now, and we're going to be working with Ben and others to do this, is the typical metadata template on the document might be, I dunno, 10, 15, 20 fields, something like that, division subdivision, 10 i, stuff like that.
(00:41:03):
So we're saying, all right, if we're really going to let AI unleash AI on this stuff, well, so why can't we build a least metadata template where we'll put like 200 fields and then we can take 200 fields from our structured version of the unstructured data and automatically populate that metadata, which I would expect in the next three to six months to be able to do. When we could do that, it'll be really interesting to see that second part of the equation. The first part was just what can you do with AI over lots of stuff? The second part is how much more powerful than AI be? If you can put AI over that unstructured stuff with really good metadata on top of it. In our industry, it's definitely the least example is the one and all of our colleagues are the same on this topic.
(00:41:55):
Now the final thought on that is, right, well, how do you get this stuff out of lease? Because extracting so extract is now you guys are building extract. We're going to look at it. It's an interesting product. Remember, you're already competing with several purpose-built solutions in the real estate industry that do this. Now, the advantage they have is they understand the lease way better than let's say box would. The disadvantage is they don't have the kind of resources the box does. So it's going to be whether you extract using one of these industry standard products that are out there today that are pretty good at extracting stuff out of leases, or you do it with box extract. I don't see that as the pivotal feature because not to diminish the value, but all you're doing is you're saving a little g and a time. This is not like we digest 25 leases a day. You don't that many commercial leases. So yeah, there's a time and effort savings and extraction, but it's not like a human. You have to have a human over the extraction anyway, because no matter how good extraction is, the language is complicated. Somebody's going to have to look at it and say, yeah, you got it right, or you didn't get it right.
Jon Herstein (00:43:06):
Yeah, and I think the structured metadata plus the capabilities, the LLMs just means that when the human does get involved, they've already narrowed down quite a bit to the thing that they really need to focus on as opposed to the needle in the haystack. So I think it's going to help make both the human part of the process more efficient and also likely more accurate would be my guess, because you get
Robert Entin (00:43:27):
Way more accurate. I'm thinking of the one that they ask all the time about expansion options because expansion options are complicated in these buildings because you sign a 300,000 squares to a lease, you have to negotiate that as you grow. I have other spaces in the building I can take, but there are already obligations on those spaces. So are you first in line, second in line, third in line? What are the terms of the people in front of you? They get a one shot option, they not. So asking an l, LM to figure that out, forget it. But imagine if the metadata you have on top of the lease has an option, yes or no. So there's gatekeeper feel to say yes or no. So it can immediate eliminate start to narrow down, narrow down, and then if it's a yes, it's got a list of the spaces. That's going to be a really interesting challenge for you guys and probably the rest of the world is when I ask an agent a question like that, how is it going to no, what to use the metadata for and what to use, where to use the document and be able to go back and forth fluid.
Jon Herstein (00:44:26):
And I think given the challenge of context rot that we're hearing about more and more, the idea that if you're just throwing more context at the LLMs doesn't actually give you a better answer because it actually winds up getting overwhelmed by all the context you provided. So being able to use the structured data to narrow down and then just provide the context that you know is relevant for that particular query means you're going to get a better answer.
Robert Entin (00:44:48):
Absolutely. Yeah, no doubt. When you say throw context is really important to these engines, I had a conversation with one of our senior executives and we said, look, here's hubs. You got a hub for loan documents, you got a hub for 10 Qs and TED Ks and analyst reports and really important documents. The answer was, I don't want that. I want to just be able to ask the question why? Because there used to chat tape. If you want to vectorize the entire mass of documents and it will be done, but there's no way an AI agent is going to answer a question better over that then if you could focus its context to where it should be looking. And I think so I think AI ready is something that we're trying to educate our users to say, look, you had a network for a hundred years.
(00:45:39):
There's folders all over the place. You got data. That's some data that's so old that you don't want an ai, it's so old and outdated. You want an AI agent giving you an answer on that. It's going to be a wrong answer, it's going to be a wrong answer because the data's wrong. So you've had this network for all these years, your data's really not in great shape as it relates to being AI ready. One of the toughest challenges I think for almost all organizations is going to be particularly on this content management piece, unless there's some exceptions who are doing really well at structuring their folder structuring where they store things. I think most are going to be dealing with a fairly major reorganization effort to get stuff into the right place so that the AI is really effective.
Jon Herstein (00:46:22):
Agree. And I think abstracting all of that for the end users can be really important too, because at the end of the day, to your point, your senior executives, they don't care. They just want to ask a question, get the right answer. And so all that's going to have to happen behind the scenes with agents fanning out, figuring out what's the thing they're really trying to get at what context would be useful for the model to have to be able to answer that specific question and then just going after that specifically is going to result in a better response.
Robert Entin (00:46:50):
So I think hubs is a good starting point, right? Because hubs lets us create these little focus groups,
Jon Herstein (00:46:55):
Create the context,
Robert Entin (00:46:57):
And I know you're doing a lot on hubs and we look forward to seeing the next generation, but the right generation of hubs will be when I could walk up and ask a general question and it's going to figure out which hub to go to. So I don't actually have to worry about that.
Jon Herstein (00:47:11):
Well, we talked a lot about the technology aspect of all this, but obviously technology alone, and we gave some examples that doesn't completely solve these problems. So you've got to also be thinking about people and process. So I always like in these conversations to talk a little bit about change and culture and delivering value and how you really get an organization as complex and large ASI to adopt and benefit from these changes. So I'm curious, when you implement a new technology, whether it's box or cloud generally, or an AI pilot, what is the critical path to realizing value? How do you think about that?
Robert Entin (00:47:46):
There are two swim lanes for me when I'm thinking about that, I'm thinking about implementing kind of a system versus rolling out a piece of technology. So rolling out a piece of technology like, okay, so how do we get people to buy into the box AI or any AI initiative? First thing is, so we poke around the organization and we did this with box hubs. We'd poke into a department and say, well, what if you could do this? What if you could take all of your engineering manuals on all the HVAC and BMS systems in the buildings and have 'em in one place? So any building engineer could walk up to something and say, I'm on a Siemens chiller and I want to know how to replace this gizmo. It sounds good, but somehow we don't get any takeup. We don't get any takeup because it sounds good and it is a good idea, but oh boy, what do I have to do to do that? Geez, I got to get all these documents. I dunno where they are, update. It's that mountain. How do I get there? So
(00:48:42):
We would poke around at various corners until we find one that becomes a high ready. And so we found our high ready in the m and a world in our acquisitions group, very analytical, dealing with lots of really thick and rich content, whether they're offering memorandums or board memos or deal sheets and stuff. So they seem to be amongst in our user group would be kind of the high readies. So I think the answer is you got to, and this is the same with I think almost any implementation project, a little different in the other dimension, but you got to find your group that's the high ready and if you can get a group or two or three going and they can show results, then the rest of the organization starts to see those results. And then maybe you get some traction. But getting people to do things the different way that they didn't do the day before, even though there's an obvious benefit, it's just somehow not always so straightforward.
(00:49:39):
Now, when it comes to implementing, let's say systems automation in some of the administrative systems or doing full system replacements, we just kind of finished one in Chicago and it went, look, from my point of view, it went, okay, you ask the end users, they say maybe not so good. And I think it's a couple of things. One is, well, you're really changing the way they're going to do things because it's a different implementation of a different type of system. This was for trade show management. It turns out, first thing is they really have to be involved in seeing the new product before they select it. And they always are. But end users, when they look at a demo not being in the data processing business, they can look at demos and say, Ooh, that's nice, but somehow they can't really picture how it's going to be when their hands are on the keyboard and that's what they're doing every day.
(00:50:29):
So I think along with that, getting them to really see and buy in. I think the other thing is too, don't sell them a bill of goods. Tell them it's going to suck. Tell 'em don't lie to them. Set expectations. Yeah, you got to set. I mean, they would say they exceed expectations, but with implementations it's hugely important. You have to tell them upfront that some things are going to work well, but many things are not, and it's going to take a while to settle down because if you're not honest with 'em, it's going to happen one way or the other. Once the thing goes live, you're going to have a bunch of things that just didn't work the way you thought they were going to work, or they're more inconvenient. And at that point, if you've given them the right preparation and then when it happens, you don't lie to them then either. You don't say, we'll fix this tomorrow. You say, yeah, you're right. This isn't good. This is going to take us a month. It's six weeks and here's what the workaround is or what have you. I think if you do that, you're not an adversary to the end user. You're actually a play.
Jon Herstein (00:51:31):
When you're pitching those benefits and you're thinking about ultimately realizing value from these sort of initiatives and technology adoption, what metrics do you focus on for yourself and for the people you're pitching to? Is it mostly cost savings? Is it speed, tenant experience, risk reduction, or anything else? What are the key metrics that
Robert Entin (00:51:51):
Matter most? Sometimes there's costs, right? But you know what? I don't think it's usually cost because usually you're placing one cost with another, and you might even replacing it with a higher cost, a more modern
(00:52:01):
Tool. So I don't think it's cost more. It's your work. Ultimately, you're going to be able to do things in this system that you were just unable to do in the old one, and indoor took you much longer in the old one. And that's why when it gets implemented and it doesn't take less time, it might even take more time, at least initially, that's why they're so disappointed. But it's usually the reason for doing it is usually that you're going to get more automation, more of the work you're doing is automated and the information you can get out of it is
Jon Herstein (00:52:35):
Better.
Robert Entin (00:52:36):
So
Jon Herstein (00:52:36):
Sort of capability, efficiency and maybe better quality data at the end?
Robert Entin (00:52:41):
Yeah, I mean that's more than the cost I think is typically why. And sometimes the answer is the thing you're using, it's a 1954 Chevy and it doesn't work anymore. So the answer is you have more choice. You have to do it.
Jon Herstein (00:52:56):
And have you found specific programs or tactics that work well to educate teams as you're rolling out something like new AI or other digital technologies, things that have worked better than others in bringing the end user population along?
Robert Entin (00:53:11):
I don't know. It is something that works better. I mean, getting their hands on the keyboard, getting management to say that when we have to do this, and then having their managers be able to say, well, you have to carve out a amount of time per day per every other day per week. Without that happening, many users won't gravitate to it on their own. It's not like many people have tons of extra time during the day. The jobs are oftentimes filled from the morning. There's always something to do
Jon Herstein (00:53:44):
And you know how to do your job with the tools you have today. That's your own.
Robert Entin (00:53:49):
And no matter what, no matter good, the new tool is when you first start using it, you're not fast enough.
Jon Herstein (00:53:54):
So when you think about the user experience, whether it's the tenant or an employee, how do you decide when a solution, whether it's AI or anything else, is really ready for the audience? Is there a set of go no-go criteria for you that, okay, this is good, we're ready to roll it out knowing there's always going to be some resistance, but it's good enough? Or how do you think about that decision point?
Robert Entin (00:54:16):
No, simple answer to that. I mean, sometimes it's ready because there's a hard date. Oftentimes you plan, you say, well, this is our target date and if we miss it, we have another target date. But sometimes you can't slip. So sometimes it's driven by that. But if it's not driven by that, which is hopefully the case because some things are not so date dependent, then it's a judgment call. You can't wait until it's a hundred percent polished because it won't be a hundred percent polished until they get it and they don't like it. So you have to say, we have to make a decision based on prototyping. When someone's building furniture, you have a rough finish and then you have a final finish. We have to put it out there when it has kind of a rough finish on it, but it's functional because we won't really know what the finishing fts are going to be until they get their hands on it anyway. So you don't want to put it up to soon. If it's too soon, then it really isn't going to work. If you wait too long, it just takes too long. And then you didn't accomplish anything anyway because you waited too long. You got it in their hands. It's not like it's going to be done anyway. You thought you finished the sanding and you final coating and you've got three coats of lacquer on it and they say, well, I don't like that finish anyway, you've got to
Jon Herstein (00:55:22):
Do it. Or they don't like the shape of the chair that you built. Well, you did our
Robert Entin (00:55:26):
Exactly. So you really have to get it to them where you still have time to apply finishing touches, but it's not so rough, but they'll reject it.
Jon Herstein (00:55:36):
Yeah, I was going to say, there's probably just an element of the experience of having done this for a couple decades and knowing the business. There's got to be some judgment applied to that as opposed to here's just a framework that we use to tick the boxes.
Robert Entin (00:55:47):
And the other thing is, and I always hate to say this, but it's not really a democratic process. If it is, it'll never get done. At some point you have to put your foot down and say, it's too bad we have to go now. Now sometimes that's a fight because we're not all powerful. We can't make all
Jon Herstein (00:56:05):
The rules. So maybe pivoting a little bit from user experience and value and how you make those decisions to maybe thinking about the fact that you've got some pretty sensitive data. You talked about things like obviously invoices that probably have somewhat sensitive data, but your lease agreements probably have some very sensitive elements to them. And how do you think about setting guidelines for what AI can look at, what it can't, where you require human oversight? How have you codified that or have you yet? At Veneto
Robert Entin (00:56:35):
We have a little bit. So when the L LMS came out, obviously check cheap pt, first, perplexity Niro, you've got Gemini. Now we put in place pretty quickly a procedure that we block all that at the corporate firewall level, not permanently because we block it and then there's a procedure. We encourage it, but we want you to use it. But the answer is that there's a form and an acknowledgement form and a conversation that has to take place for you to get authorization to use the platforms. And it basically says these are public models. They train on data. We don't want them training on any of our data, so please use them for anything you want to generate whatever you want. As long as you're not putting sensitive information into these models. Which again is the beginning. Look, that's the beginning of the AI race where we all have access to these public mugs.
(00:57:30):
So like box ai, and I think many of the colleagues around you in the space are doing the same thing. And chat. GBT has the corporate model where they can give you your container. They don't train on your container. That's the most important thing in implementing these technologies is to do something that you know can take your most sensitive data, get it containerized, get it vectorized, but models aren't contain on and it's private. So for us, box hubs, when you first introduced hubs and we start to look at it, we said, great at the time, we are not going to let people do things in chat, GPT, putting sensitive documents in. But that's really, I don't think it's that complicated, really, is that you have to containerize all your sensitive information in the container that whether it's, I'm want to say it truly shouldn't be on-prem, it should be in somebody's cloud, but it should be containerized in such a way that the security is guaranteed.
Jon Herstein (00:58:27):
And that's been a very core principle for us, is no training on your proprietary proprietary.
Robert Entin (00:58:31):
Absolutely. Yeah, absolutely. Yeah. Otherwise what's going to happen is somebody's going to ask a general question down the road and it's going to come with an it's answer out of one of our documents. We can't have that
Jon Herstein (00:58:45):
Cannot happen.
Robert Entin (00:58:46):
Yeah.
Jon Herstein (00:58:46):
So we talked a lot about kind of things up until now. I want to pivot a little bit to the future, and I found a quote where you said AI is the most significant thing in our future, the biggest thing that could change the way people work. So I'm curious, you pull out your crystal ball, what do you see coming and what do you see becoming mainstream specifically over the next, again, choose your timeframe, but let's say two to three, five years?
Robert Entin (00:59:10):
Well, I mean, no revelations here because I'm not going to say something that I think a lot of people aren't saying. But we talked a little bit earlier about, I looked at the 2000 revolution and it took four x longer than everybody thought it would, where the internet really, you could do almost anything you want on the internet. And we said that this one's going to happen much faster because the whole infrastructure to support this is already there. So I look at things and if you look at just from, and I don't want to say from 2022 to 2025, really from 2024 to 2025, we were two years gawking at how cool open AI was. All of a sudden in the last year in all of our products, it's creeping in, whether it's you guys or Zoom, it's virtually everything on Smartsheet, everything you own AI is creeping in.
(00:59:55):
So in one year, look at the progress that's been made already and what we're at a point now where we have really good technology that we can have great conversational content rich stuff over all of our content and get really good answers. So that's going to give us a lot of time. But we're just at the beginning of training agents to talk to other agents and MPC connections and stuff. I mean, we're at the very beginning of that. So if in the next call it two years, it becomes commonplace that my agent in box can talk to my agent in Slack and my agent in the workflow product and we can train them both and they can connect and they can really interact, then the automation that's going to take place inside the administrative areas in all these businesses is going to be really profound, I think.
(01:00:48):
So what I see is generally like at 20,000 feet, we have millions of workers that work in administrative areas inside companies. An enormous percentage of those jobs are going to be disrupted through AI agents. And what's interesting to me is that, I mean, look, in a way it's going to be tragic because it's going to be a big disruption if you go the disruption that took place, call it from maybe 2000 to 2010, both with the advent and the internet, with globalization, with large shipping containers and the ability to produce in China and consume here for less money than you could build here. That ruined the deal. It created the rust belt, it gutted manufacturing in the United States and many other modern nations. It kind of ruined the deal. If I worked in a steel factory for 20 years, the deal was I work here, I'll be able to retire, I can take care of my family. That deal was shattered by that disruption tragically shattered by that disruption. And it created, I could argue not to get into politics here, but it created a geopolitical atmosphere that probably is responsible for where we are today in our politics.
(01:02:09):
The workers have changed parties, which is really weird. It's happened a few times in our history where the Democrats and Republicans switched, but this is one where they've somewhat, and I think it's because of that disruption. Well, let's think about the next five to 10 years. The disruption is not going to take, it'll take place a little bit in the factory. What factories are left. It's going to really take place in the office, and I don't think it's going to create more jobs than it destroys. I just don't.
Jon Herstein (01:02:39):
So what do you think that means for you in terms of your IT organization and the skills and mindsets that you'll want going forward as AI and data just become more and more important and how do you think about evolving your teams?
Robert Entin (01:02:56):
There's always going to be a need for programmers in the real technical side of it, although that, Hey, listen, we know that these engines write code. One of the things they do best is write code because it's a very structured language, unlike English and stuff. So I still think there's going to be a place for that, but those will be people watching bots write programs. I think the number one skill is going to be people who really know how to work with ai, meaning they understand where the data has to be, they understand how it's got to get structured. They know how to train agents. They know how to connect agents, they understand they have to understand workflow more than almost anything else. Now you're going to have to understand what the business does, which oftentimes it really didn't understand the business that well. But I think needing to understand what the business does so that you can take this new tool set that we have and apply the tool set to automation, that's probably the pretty important skill to have.
Jon Herstein (01:03:57):
So it sounds like your advice to someone in IT now, or thinking about it would be get more familiar with business process, the business side of things, and not just that technology side changing so quickly?
Robert Entin (01:04:08):
Yeah, I mean, the traditional role of the business analyst over time used to be talk to the liaison between the business and the programmers. Well, the programmers aren't going to be as important anymore because a lot of the code's going to be, first of all, there's a lot more off the shelf stuff. There'll be less bespoke stuff and you're going to have bots writing code. So I think the IT professional can't just be the translator. They have to be kind of the implementer. They've got to be the one that says, look, I know how you process invoices and they come from five different ways and they're approved differently. And why are they approved differently? Oh yeah, because the capital is approved this way. Is operatings approved that way? Oh yeah, leasing commission happens this way or that way. And understanding those workflows and what are the important parts of them and therefore how to automate them. I think that's tomorrow's world is going to be a lot of automation.
Jon Herstein (01:05:04):
So you have obviously spent a lot of time thinking about not just the technology, but the implications of the technology and what it means for your organization and people and so forth. And so as we wrap up, I want to ask you advice for other CIOs, technology leaders, maybe those who aren't quite as far along as you are, who are thinking about how to navigate ai, the digital revolution in their organizations, their people. What are one or two or three tips or thoughts of advice you would give them
Robert Entin (01:05:35):
If you're not using it in your daily life? That's a mistake, and I actually had a town hall a couple months ago and I said to them, if you don't go use this in your daily life, your jobs are going to be at risk in a couple of years. A lot of this AI ready and dealing with agents and how do you work with it? It's not rocket science. The number one piece of advice to somebody I give is talk to the agent, whether it's a chat GPT general agent or whether it's a box agent we just trained, or whether it's a chat g PT specific agent, we put over a corporate bucket of data. You got to talk to it as if it's smarter than you and don't. Yeah, we're used to searches. So searches like you curate the sentences, you don't even write sentences, you curate the phrases.
(01:06:20):
So the search engine can give you a bunch of links and then you poke around the links to get an answer. And it's exactly the opposite with engines. You need to be more explicit. You almost need to bloviate in your question, give it way more information than you think it needs. It says you can, right? Yeah. And you'll always be surprised that the more you give it, the better it does. So I don't think it's necessarily rocket science. It's a lot of fundamentals. It's if you don't use it every day, so you get good at it. I would tell you that a year ago, I am to be a perplexity fan, so I subscribe, but it's one of a lot of great engines. I use chat pizza, but perplexity is my go-to for the moment. Maybe a year and a half ago I was 15% perplexity, 85% Google.
(01:07:10):
Then Google's Gemini is creeping into Google now they realize search is dead the way it is. I would tell you today, John, I'm 90% perplexity at 10% Google. Using your daily life is usually important. The other thing I would say is, do you understand what the business is doing? Many CIOs, I was always, I'm a computer science engineer, but oddly, I had this love of accounting. My grandfather said to me when I was a kid, if you want to be an entrepreneur, you got to do two things. You got to go get a job as a salesman at some point because if you can't sell anything, you're dead. And number two, when you go to college, you must take accounting because if you understand accounting, you'll understand the way all businesses operate. I didn't just want to build this software. I want to understand why you're doing it, why you're doing it the way you're doing it, what function in the business does it accomplish.
(01:07:57):
That made me a little bit unusual in CIOs because again, there's usually a lot of liaisons and VAs between the IT guys in the business, but I think in tomorrow's world, I think IT guys are going to have guys and gals. I think they're going to have to understand the fundamentals of the business at a much deeper level than they used to be really effective. So I think it's understand the way data has to be organized, understand how these things work, and most importantly, understand why the business is doing what they do when they do it. So
Jon Herstein (01:08:29):
If you're a technologist, don't just be a technologist. You also need to understand the business.
Robert Entin (01:08:33):
I got lucky. I had a grandfather who prompted, I have four children, three of which are engineers. I said the same thing to all of them. Take accounting.
Jon Herstein (01:08:43):
Well, Robert, thank you so much. This has been an incredible conversation. I want to just thank you for obviously your partnership as a customer box who's pushed us on product, helped make our product better, but also thank you for your time and your willingness to share your vision, your candor, and your leadership with us today. It's clear your ability to bridge real estate expertise with technology and the people side is rare. And I think your insights on AI automation and the transformation that our workforces are going through certainly am me a lot to think about. And I think our audience as well. So to our audience, if you found this episode valuable, interesting, please be sure to follow our series. We've got a lot more conversations ahead with many, many more leaders who are redefining how enterprises approach ai, just like Robert. And so until next time, I'm John Stein. Thanks for joining us on the AI first podcast. And again, Robert, thank you and goodbye. Thanks for tuning into the AI first podcast, where we go beyond the buzz and into the real conversations shaping the future of work. If today's discussion helped you rethink how your organization can lead with ai, be sure to subscribe and share this episode with fellow tech leaders. Until next time, keep challenging assumptions, stay curious and lead boldly into the AI first era.