Subscribe
Share
Share
Embed
Welcome to Chuck Yates Got A Job with Chuck Yates. You've now found your dysfunctional life coach, the Investor Formerly known as Prominent Businessman Chuck Yates. What's not to learn from the self-proclaimed Galactic Viceroy, who was publicly canned from a prominent private equity firm, has had enough therapy to quote Brene Brown chapter and verse and spends most days embarrassing himself on Energy Finance Twitter as @Nimblephatty.
0:00 Good afternoon, everyone. My name is Kinesia Cesario. I'm the CTO for Collide started about six months ago. It's been quite an incredible journey. My background is software. I've been in
0:11 software tech for about 20 years. Different companies in Austin. So doing a lot of the up and down between Austin and Houston. I'm excited to be joined by Bill Aaron. He's the managing director at
0:24 Microsoft, overseeing a lot of the strategic teams that interface with Within the energy sector, they enterprise companies. And over two decades of experience, that spans Workday, Oracle,
0:37 DataGumbo. So been around the block a long time for just great companies. But Bill, maybe if you could share a couple of things and I can kick us off into the conversation. Yeah, I appreciate the
0:47 opportunity. Thanks for the invite. I have some cold sweats. My last startup was in the canon over there. And ironically enough, where's Bobby? He's not here. I gotta tell him that the
0:58 beginning of the end Grayson Mill acquired the assets from Equinor. Equinor is our major funder
1:07 and that we're about a month from go live when that acquisition happens. Buroodle. But anyway, 31 years in this enterprise software space seen a few things. This gen AI story is something that
1:20 will all it back at in our careers as one of the most generational impacts we've had
1:27 Awesome, thank you, and to kick off the conversation, we're going to focus a lot more on, as it says, the on-prem and cloud that is still the thing across compute data security. AI is everywhere,
1:40 right? And Microsoft is also at the forefront. And just for context, Collide Enterprise is built on Azure technology. That's the full stack that we use. A lot of really amazing Microsoft tools,
1:52 Azure, that's Doc Intelligence, AI service So we are very much embedded within the Microsoft ecosystem. Part of that is just because, so my background is a lot more AWS, but this was the first
2:04 time that I was moving over into Azure, because pretty much every oil and gas company is a Microsoft shop. So by default, that gives us a lot more enterprise entry in this integration in. But what
2:18 I found was that while everybody does talk about AWS, the Azure ecosystem is super strong now, and especially for the developers, so you can actually get up and running very quickly It was very
2:29 impressive, actually, to see how far it's come. So, Bill, to kick us off, in oil and gas companies are still spread between cloud and on-prem, like a lot of the enterprises are. What are, in
2:41 your opinion, the key factors that should guide that decision today? And how do you see that migration path for a lot of companies? Yeah, no, it's certainly a journey. And when I boomerang back
2:52 to Microsoft, it was before this AI story broke and when you look at. what's happened with AI. It's been around for 50 years, 1950. Artificial intelligence was in, people sometimes forget that.
3:08 Well, why is it all of a sudden the front page news? And it's really because of generative AI. And so what has happened is the technology that is out there from this generative gen AI story and
3:21 open AI got to 100 million users in three months The internet took nine years, cell phones 14 and a half years to get to 100 million. So all of a sudden this explosion happened of something that's
3:36 making impact in how people interact with this solution called chat GPT. And so immediately to be able to have that kind of compute and be able to have that kind of capability, cloud became a
3:49 requirement. And so we don't have the discussions. I lead a team focused on our large enterprise space for the energy sector.
3:58 Midstream still has a lot more on-prem than cloud, but most of the ENPs are all in, they've made a migration to the cloud. So it's a situation now where the decision criteria isn't so much winning
4:12 or losing on the cloud or not. It's a strategy of hybrid and at the edge and being able to develop a strategy that's both, right? And I think that is what's playing out at a theater near you One of
4:25 the up streams that I worked with, hadn't met with Microsoft in many, many years, and he told me, Lenden, I have to talk to you now because you have what I need which is compute and being able to
4:39 take advantage of something that is delivered in a cloud at scale. So that's the big change is what's taking place. That's been the, I guess I'd say the driver to this movement towards the cloud at
4:52 a huge scale.
4:55 And on that topic, if we dig into compute a little bit, since that's come up multiple times, and these models require a lot of compute power, how is Microsoft and Azure scaling to that? Because
5:06 the demands a lot, I've heard some people talk about, let's just kind of put our own data center and our own racks and our own compute, just so that it's cheaper. We can access, like, what are
5:19 your comments on something like that? I mean, the sheer buying scale is just unbelievable with the three hyperscalers, right? So we'll announce earnings here and we'll talk about the capital
5:32 expenditures and what we're doing to build out our data centers and what we're doing. The volume and the size, and it's massive, I think it's80 billion of investment so far this year, this
5:44 calendar year on building infrastructure to deliver compute at scale If a small company or a medium-sized company or even a large enterprise decides to do that on their own. They can, but they can't
5:57 do it at the kind of scale we can in any of the hyperscalers. So I think there's a reality check there. Yeah, awesome. So the next thing that I've definitely gone into a lot when we have talked to
6:09 our clients, Focalight, Joke in a test to it when we were talking to his IT team is a lot around data security, right? Compliance data security. People are very sensitive about that this is my
6:22 data
6:24 I don't want to give it to anybody else. And how do you make sure that you're not training on my data, which is a very common question and keeping customer data segregated?
6:35 We have built on Azure with the trust that Microsoft provides a lot of that. But being an insider, could you shed some light to how Microsoft and Azure goes through that process? Yeah, it's table
6:46 stakes. It's built on trust. I mean, this is a story here where people are taking
6:52 intellectual property, they're taking their crown jewels and putting it in the cloud, and they want to make sure that it's not going places that shouldn't be, where people have access to it. So a
7:02 true multi-tenant SaaS story is effectively sharing infrastructure. Well, the data is encrypted, the data is separated. There's strategies here on how our customers are protecting their data.
7:17 Under zero circumstances, are we using our customers' data to train the models? The exception is if they opt in. So there are things that it's very important. And we put in our contracts, it's a
7:31 very part, a big part of our secure first initiative, which is in order for us to build on trust, we have to make sure that our LLM and the training that we're doing is not taking and commingling
7:43 other customers' data. It's an opt-in motion and we have a copyright in a data protection clause in our agreements that is. effectively saying, we are putting our money where our mouth is, and
7:55 it's very important. If we tray that trust, then bad consequences. So to me, this world of building out models and being able to take and building a competitive advantage with what AI can bring.
8:08 We were talking earlier about some of the examples in your world, and the last thing we want to do is be able to take a small medium or large customer and be able to take their data where another
8:15 customer is using their data to help do something with
8:25 it, right? So the story here for us is built on encryption, it's built on segmentation, and it's built on a contractual obligation. So for us, it's table stakes. And when you look at how our
8:38 customers are architecting at the highest end with what we're doing with ExxonMobil, what we're doing with Conoco Phillips, what we're doing with Chevron, every one of our customers is doing
8:47 unbelievably sophisticated things. each of them are architecting it differently based on where they're at, and we have to meet them where they're at. And that's part of the story. And the
8:58 underpinning and the setup for your obsession was perfect 'cause the data is the foundation of it all. And so we have a data set and a data story that we have to meet customers where they're at. We
9:11 have to build on that data in order to train the models and be able to help our customers build these stories where they can solve some complex business problems and do it in a way that they can trust
9:23 that we're not doing that improperly with fixing and mingling data.
9:28 And that's something that's very cool to collide as well because as we, as I've kind of seen a lot of customers across SMB all the way to enterprise, everybody has a little bit different deployment
9:40 model that they want to support in the cloud. Some people want to support,
9:45 The traditional SaaS, just make sure that there's data protection, and we just want to log into aicollideio. On the other spectrum, people have Azure set up already within their environments, and
9:56 they want the collide enterprise application to be deployed, fully managed. And then there's a whole spectrum that goes in. Well, is that kind of what you also see happening across the industry,
10:06 that especially with these SMB to enterprise scaling? What kind of deployment models have you normally seen? About the companies and on the partner school too? Yeah, I think it's, depending on
10:17 size, really, size and scale. Our large NOCs are putting their data in their own tenant, and then the ISVs have to meet them where they're at and do things different to their configuration.
10:33 There's sovereignty rules with data, and you'd be able to have data in country. So there's rules that dictate where things are gonna play, But once again, it goes back to a hybrid as a strategy.
10:44 not a stopgap. So there's bad connectivity. You need to have stuff at the edge. You have sovereignty rules. You've got to meet the government's responsibilities and the goals and not the goals,
10:54 but the restrictions that the government's put on data sovereignty. So architecting for that is going to be hybrid story. The SMBs, the small to medium businesses who don't have that luxury of
11:07 having a deep bench of people. This is where the multi-tenancy and the ability to leverage a SaaS model makes a lot more sense. Pure play, right? But I think we see it all over the map. And the
11:21 trouble that it presents is for ISVs like Collide, you have to meet your customer where you're at. Even at the highest level, some of my customers in my portfolio are Halliburton and the SLB. I
11:34 mean, they have the same exact problem. So their story, they have have to architect their solution at scale. and be able to meet the customers where they're at. And what we're seeing is the SMBs
11:46 are agreeing with and are able to lean in with how it's constructed out of the box with their story. And then the large NOCs are saying, No, no, no, no, you need to put it in my tenant. And so
11:58 there needs to be adjusted in configurations and architecture to meet them and meet those requirements, which is complex. Yeah, no, and definitely faced a lot of that across And I think Azure also
12:12 definitely helps with having a lot of dev tools that actually allow us to move. But at the same point for the enterprise clients, they do require a lot more customizations, which is where then we
12:26 have to budget for that property to, right? Like it can't be the same solution that's sold in a complete SaaS environment that's then sold in an enterprise. I'm assuming that's very similar to what
12:35 you have seen as well. Absolutely think there's the art of this is monetizing it and putting a pricing engagement that makes sense. There should be a premium on something that's put on a customer's
12:46 tenant because the way ISVs are set up, I mean, you look at service now, they're charging a 20 premium if you want it in the customer's tenant. You look at salesforce, you look at everybody who's
12:57 a SaaS solution. They're multi-tenant nature, the
13:02 ability to support that customer and be able to have innovation pushed in a very, of a quick way Those stories, there's a story that's playing out and how to monetize all this and make it reasonable
13:17 and it's business outcome based. So I think some of the use cases you all were talking about, the power of the story is it's problems that companies have been facing for years that haven't been able
13:29 to really put technology to it. So this is really a land grab for taking and you look at the headlines of all the, all the layoffs and the things that have happened. It's lean and people are being
13:41 asked to do more. It's all business outcome driven and solutions like collide and what you can do to really help scale up and go execute on things where it was very people oriented. This is where
13:56 the agents in the AI world is shifting how workflows are executed. So it's gonna be a fun run, we're just beginning. Yeah, that's awesome My next question is, this is the more the platform and
14:11 the innovation behind Microsoft and Azure and what we are using as our Rails. Like what in your opinion, what makes Azure
14:19 particularly well positioned at this point for the energy sector? And which innovations we talked about AI Foundry, we talked about block intelligence. I know Microsoft has a lot of things brewing
14:30 right now. Like, do you, What do you think will have the biggest impact in the near term as well as we are kind of shifting towards using more Microsoft build technologies? Yeah, I think we've
14:42 been in the energy sector from the get-go here. We haven't left it and come back. We're very laser focused on serving our customers. The unfair advantage that I think we have is when you look at
14:53 the initial use cases of how co-pilot, how many in here actually use co-pilot just out of curiosity So when that first came out, it was sort of a nice, man, this is pretty neat, right? It's a
15:06 toy. And that toy was sold and given to our customers. And it was a productivity assistant for your traditional productivity tools, your office, your teams, your outlook. As things have evolved,
15:22 it's moved into a world where people are extending those co-pilots our ISVs are embedding co-pilots into their solutions. And that becomes the new user interface of what we expect. We expect to
15:35 interrogate data using natural language queries. And interrogating that data assumes the data is right, which we all know is it's a journey. You're not gonna have perfect data out of the gates. So
15:47 the unfair advantage we have is we have these productivity tools 400 to up are believe I which,
15:52 million users of our productivity tools with these co-pilots that are attached to them, where our customers are starting to use them in novel, interesting ways, and now they're extending those
16:04 co-pilots. And now we're in this new generation called a Gentic AI, where it's actually sophisticated to the next turn of the wheel here, is how you can take a workflow and have agents as part of
16:17 your day-to-day activities. That's unbelievable to me, but it's real. Every day I'm in meetings, I have a facilitator agent in my meeting, I have a project manager agent in my meeting.
16:29 Those things are helping me be more productive and pick that from frontline operations, pick that from legal, HR, finance, pick up the room of the house within these enterprises. Our opportunity
16:41 to attach solutions on our stack that are taking last mile functionality in the oil and gas segment, that's not what we are. We are a platform and we want ISVs and companies like Coli to develop to
16:55 their heart's content, very sophisticated solutions on top of our stack that's driven by security and by
17:03 scale. And our story of co-pilot studio plus the orchestration of agents and using our AI Foundry. I think the old Microsoft was much more proprietary. I think today our AI Foundry has what? 4,
17:18 000 models in it. Pick your model, build your own model And the story of building that out is tremendous opportunity for every one of our customers. able to take advantage of. So I think we have a
17:31 first mover and the term we're using is we want to help our customers be frontier firms. So getting out in front of your competition and I think small to medium customers get a chance to play above
17:43 their weight class. I think you can do a lot, a lot with less. Yeah, no make sense. And actually building onto something you already mentioned about how Azure and Microsoft is the platform,
17:57 right, for ISPs like us to build on top of. I'd like to dig in on that a little bit because some of the questions we have to feel is, I already have from our clients, right, our customers that I
18:09 already have Microsoft, I already have Azure, I already have SharePoint, I can, I have co-pilot, why can't I just do this myself, right? Why can't I just give my agent all these files? What is
18:21 special about collide, right? Like why should we go for collides? I know you know a little bit about collide as well. I would love that answer from somebody sitting in that can actually help me
18:31 answer the question for our clients too. Yeah. Well, I think that there's a, in the software business, there's always the build versus buy discussion, right? Do I build something myself or do I
18:43 buy something that's off the shelf and it's already been built for the purpose I'm looking for, right? You know, I think the ability for, for companies to look in the mirror and recognize that
18:56 they're not software companies, they're operators, they're
19:02 definitely not maybe self admitting to that. I think that's part of the challenge. But I think what I look at collide and I look at Microsoft, it's an end. It's not an either or. I think people
19:15 are recognizing that what you all do and frankly, every ISV that's developed a last mile piece of functionality on top of our platform, it's designed to go solve problem it takes time to develop a
19:28 solution to that problem. And if it's already been done and dusted where you've had repeat in its at scale and you've proven it with stories, that the time to value is unbelievably powerful 'cause
19:40 now you can plug something in and just think about your example with the railroad commission filings. I mean, that in real time saved the company a lot of time and heartache and money. Okay, so
19:53 building that yourself I mean, you just in your session there, Bobby, you talked about stories of the data and the challenges of getting there. You all have already done it. You've gone through
20:03 the
20:05 parsing and I'm not gonna be able to nail all the terms. I've learned terms that I've heard before but I'm not intimately knowledgeable about all of that but that stuff takes time, effort, it's
20:15 hard. I look at SLB and they're huge, they're a monster. They're just, they're Lumi product and their story is brand new, right? It's a lot harder than people think. And that's where the
20:30 Collide Plus Microsoft story resonates really well, especially the Microsoft, 'cause we're not gonna develop that last mile functionality. We're gonna enable our customers to use our platform. We
20:42 wanna support our partners like Collide to help you go scale
20:46 it. Yeah, that's awesome. And I think that's kind of the secret sauce, right? That we are building here as well, which is very specific to oil and gas You make sure that we can build the search
20:58 in that manner that is very context aware and the workflows, the agent-tech workflows has been a big piece of what we are building that's been. That's resonated a lot with our clients as well. But
21:10 I'd also like to open up the floor for questions as well. I know we have kind of discussed a couple of things, but we'd love to take some questions. Oh, Victoria and the engineer on IT got it. So
21:20 I asked a data security question, which is weird for me to ask
21:24 So as I think about - this increase in AI, continuing to get smarter and smarter. How does, what are your thoughts around the risk of that encryption? With one executing or whatever it increases
21:35 computing power, do we get to a point where
21:40 it could break encryption such that people are going back on-prem so that it's not connected, or is that not a risk? Not a risk, I don't think that's a risk. I think reality here for that, and I
21:40 was gonna actually comment on your question last earlier about the
21:57 use case you were describing. Think about
22:01 a law firm. When a law firm comes speak to us, they have 60 to 80 retirees every year. What they did is they took the legal findings, all the proceedings that that lawyer took over their career,
22:12 they scanned, they did everything you were talking about, Bobby, with taking and imaging, whether it was a napkin, or whether it was a court proceeding. They loaded it and created the story
22:25 individual, so they can always interrogate and ask questions of that lawyer that's leaving. Think about that for operations, think about that for landmen, any critical role that's aging out that's
22:37 leaving, being able to pull that knowledge into a platform that can help you ask it questions, there's tremendous value in that. By default, all that data is encrypted and already in there, so
22:53 everything we do and everything that happens will be in a safe, secure environment where mixing in mingling, whether that's on-prem or at the edge or in a hyperscaler, that's all protected beyond
23:05 what you can do yourself. So I just don't view that as a risk. So whether last week, AWS, USC's one went down, right, and then our internet shut down, basically. I think that brought up a lot
23:18 of more people talking about multi-clouds, and let me talk of on-prem and cloud and hybrid, but. You see we got multi-cloud and probably some of the cautionary tales there. Yeah, look at your
23:29 headlines today. Everybody in this room is down, but 30 of the internet was down today. So I've been getting buzzed left and right with situations, but it's back and we're in good shape. But I
23:41 think a single cloud story is more likely than a hybrid or a multi-cloud story. I think five years ago, the skill sets and the ability to hedge your bets was real because of, it's all new. Let's
23:57 see where this goes. I think people are now in a situation where the skill sets, it's a different world. Each flavor has a different set of skills and that takes overhead people. And I think what
24:11 you're gonna see more of is resiliency built into a single cloud environment. They're gonna pick a horse. They can still have failover They can still have stories that are allowing some resiliency
24:22 built into their enterprise. but I think you'll see less of a multi-cloud strategy than you will a multi-cloud strategy. And I think that is largely economies of scale. That's not necessarily a
24:36 resiliency strategy or oh my, if I just pick one, what happens if it goes down? Today, all three were down. So it's not a matter of one versus the other in that situation. I think it's more of a
24:49 building a skill set and a army of folks that can support and have the skill sets around each of the stories 'cause they're different. And I'm not saying one, I like Microsoft better, but there are
25:00 certainly capabilities in each of these hyperscales that are fantastic. And so now it becomes, how can we, at economies of scale with cost and headcount and people? So having two clouds becomes
25:15 more expensive on the people side and it doesn't necessarily play out as a negotiation to lower your costs, right? Resiliency is a tricky thing. We've become very dependent on technology. I mean,
25:28 it's just a natural part of our day-to-day. We expect it on our personal lives. So when things go down, which will happen, it's inevitable we have to have a backup plan, right? So that's part of
25:39 the resiliency strategies that companies have to deal with. Midstream in particular with the data and the stuff they're doing, most of the midstream customers have SCADA on-prem. And we're pulling
25:51 that data for invoicing for things. So if it goes down, it's a stored forward kind of thing. So there's strategies that can be built in with resiliency by not going to multiple cloud as your
26:03 strategy, if that makes sense. Yeah, I think while they kind of come up there too, similar to your point about on-prem, almost thinking on-prem is safer, but it's like, do you want to manage
26:14 your own exchange server? Sure, absolutely. Over here is probably the now left-right and it's being managed something even if you're
26:22 fully on the ground, you're still probably holding to SharePoint or from the other, or if you're on the Google Cloud, you can easily and all that stuff. Yeah, absolutely. It's interesting.
26:33 Security is part of this story, and it's a critically important part of the story. And if you think about, I hate to bash on it, but no one wants to be the colonial pipeline. That was one of the
26:44 energy specific examples If you double click on that, that was an on-premise server that was an ERP story that was not in the cloud. So it was really interesting how that was used as an excuse for a
26:60 lot of companies not to go to the cloud because of that situation. And when you look at that situation, it was actually an unpached on-premise server that was the weak link. So I just don't see
27:14 when I - I mean, we're managing - I know the number of people who or maybe wrong, but it's not far off. Four trillion signals. day of security and we're working with every one of the security
27:26 providers out there to help them be better at catching the bad guy. So I think I like our chances of helping at scale versus an individual enterprise being at themselves. It takes a village. It's
27:40 not going to be one company that's going to keep people safe. So it's an important part. Thank you. Probably have time for one or two more questions. So I'm an attorney. So this may not be in the
27:54 right type language, but someone earlier talked about making sure if you're on the cloud that you retain access to your data, you talked about how everybody's looking to monetize this. Our company
28:06 is now like cloud company is now looking for how they're going to charge you to bulk attract your data to do this Yeah, great question. So from a legal point of view, you own your data to your data.
28:20 The access to it is egress and ingress and egress charges, where people are paying for taking data away and going to. So I think part of the story is people put handcuffs on that there's a charge
28:34 for that, right? So that's probably one of the, I think the context of one of that, but being an on, if you have an on-premise portfolio software, those customers or those ISVs or software
28:47 partners are developing cloud solutions And those cloud solutions are built on a history of 30 years of data of you using that software on-prem. So monetizing it, companies want the software
29:02 companies and want to find ways they can take it and help learn on so that take that data that you have that history and help make good use of that for you at a price. 'Cause they're building the
29:15 models, they're doing something as a service, right? So where you get into trouble is when you don't have access to that data. And I think that was the call out was making sure in your legal
29:26 agreements that you have access to the data that you have created with the software that you're using. And it's just like Word, we don't own your licensing Microsoft to use something that creates
29:39 data. That data is yours where it's stored as the debate. It could be in your hard drive or one drive or it could be in a SharePoint site or it could be on a third party server somewhere, right?
29:52 Legal in the AI world was hard to get people on board at first. Now they're one of the leading cheerleaders for it, which is great. So if you're in the legal department, it's the example of the
30:07 documentation of the contracts, phenomenal use case. Being able to use document intelligence to track with MA in particular, the discrepancies you have in agreements. And instead of having to
30:20 peruse through all of that, the due diligence spent on just evaluating your contracts and your risks, you can automate that in a just unbelievable fashion that can give you access to a summary of
30:33 where your risks are. So you're spending your time on the risks, not figuring out where they are. So anyway, that's sort of two answers to your one question, but legal has been a huge beginning
30:46 stage two years ago as hey, hold your role, we're not gonna be able to do this. And then the other question that you'll have is data governance is such an important part. And AI exposes data. And
31:01 if you don't have governance around your data, there is absolutely a risk that things can get exposed. It shouldn't have been. So customers have to think about putting a governance plan in motion
31:12 for their data. And they have to put some tagging, some form of protections in the United States. because it'll surface information. And the worst thing that happens that we see, and I still
31:23 believe it or not, still happens, is shadow IT, where customers haven't embraced AI. Every single one of our customers has an AI strategy and AI policy, but there are people who are bringing in
31:33 their phones, their own version, their uploading documents. And all of a sudden, that document is not protected under the enterprise world. It's up, and that's the part that we see you're far
31:33 better developing a strategy that embraces this than prevents it. And small to medium businesses are probably more susceptible to it now. Most of the large enterprises have made very bold claims and
31:33 done a good job of measuring it and managing it. But it is absolutely table stakes. You have to have good data.
32:11 You have to have governance of that data,
32:16 a whole slew of tools that you can use to unleash it to help drive productivity. So, there's a lot there. Hey, and Phil, just one quick question. What are one or two things when we're doing this
32:32 again in five years that we're talking about, but nobody's thinking about today? So this is the crystal ball question
32:52 and it goes crazy if you want for a long time around AI. Like I mentioned, AI's been of is, you've got this continuum here heading think that where this is all I what's it Well,. Machine learning,
32:54 this thing has been, it's evolved over the last 50, 60 years. We're in this agentic AI world where I think the most radical thing we're gonna see is agents managing agents So you have an agentic
33:13 world now where there's agents and you have co-pilot as an assistant. You have agents that are invited to your events and your meetings. You're gonna have agents managing other agents instead of
33:24 people managing people. That's what's happening. I haven't seen it at scale, seen talked about, seen we've had discussions about it, but some companies are taking that and that is a real story.
33:37 And I think the other part of this is, I was asked to look, where are we in the journey? It's the first inning. I mean, we haven't even at scale There's so many trials of technology going on
33:48 right now. I can't say that any one customer has gone far out where they're way ahead of anybody else. There's definitely some really good production examples where people are using AI that are
34:02 affecting their bottom line. But it's not at scale. So five years from now, I fully expect to have agents managing agents at scale I expect that we're going to see Gen. of AI. in large
34:16 enterprises and small businesses in a ubiquitous fashion. You know, that's what I think is the five-year window.
34:25 Awesome. Well, Bill, I think we had time, but thank you so much for all the thoughts here, and I appreciate you taking the time to make us.