For our COBT listeners and viewers, it's Maynard, Deborah, Robbie, and Mike Bradlin. And we're here today to do something we're super excited about. We've got Mike Fey with us. He's the co-founder and CEO of Island.io. You'll hear that Island is a fascinating company. Its roots lie in cybersecurity, but its ability to enable an organization, particularly at the individual level, and turn them loose with all kinds of great productivity tools, particularly AI, it's really impressed us and we thought it was so good, we really wanted all of you to hear it. So, Mike, let me just stop and say thank you so much for joining. We're excited for today's conversation.
My pleasure. Thanks for having me.
Well, Mike, before we jump in and talk to Mike, what would you tell us about what's going on in the world?
Yeah, I'll try to make it as quick as possible here, man. Like, from a fixed income market standpoint, you got the ten-year around 4.7%. But the, you know, fixed income market is honestly the equity market and just investors in general are really looking towards tomorrow's CPI number and Thursday's PPI number. Those two reports are going to go a long way to determining what the Fed is going to be doing at their next policy meeting. And so, you know, be prepared for some volatility tomorrow and Thursday on those numbers. Now, you know, as it relates to the broader equity markets, you know, the S&P and Dow are both down about 30 basis points today. It's not really attributed to anything but just oil prices going up. And, you know, the thing we would say is second quarter earnings for the S&P 500 and Dow are essentially over. The only companies left to report are really the big box retailers next week and so we'll get them. So the market is going to be basically moving on, you know, what oil does, what interest rates do. It's really not going to be driven by earnings here for the next couple months. You know, the one thing that really kind of stood out to us from an equity market standpoint was Intel. Yesterday they basically floated 20 billion of equity for capital expenditures, obviously for chips and stuff of that nature. And what was really interesting, the thing was priced like 6% in the hole but it was like five times oversubscribed. $20 billion raised, 100 billion of orders, so five times oversubscribed. That just tells you there's a lot of interest still in semiconductors and Intel in general. So I thought that was really interesting. You know, as far as oil markets, you know, oil's up about $5 this week. You know, we're up to around $83 per barrel and a lot of that is just because of the skepticism that there's going to be a deal in Iran. We've talked about this, I don't know, every week for the last 6 weeks about this on and off with Iran. Well, we're off again. Prices are going up. We'll see what happens here later this week and the next couple weeks, but it's going to be volatile, you know. I'd say from an energy sector standpoint, energy is up around 5% this week because guess what? Oil is up and so the sector is up. You know, as far as earnings, energy sector earnings are essentially over for that sector as well, just like the overall market. And maybe if I want to throw out some themes that we heard, you know, from second quarter earnings from an oil services standpoint, the US market is recovering. Pricing is starting to come in. That market looks really good heading into the second half of this year and potentially into 2027. The offshore market looks exceedingly good. International looks really good except for the Middle East. And so I think what you have right now is tailwinds on the offshore international and North America going into 2027. We have not seen that for quite a bit of time in oil services. And so the setup looks pretty good for those companies. As far as EMPs, what did we hear this quarter? Everything we heard last quarter. We're going to keep our second half spending under control. We're going to return as much as we possibly can and see how things play out. The second thing we heard in a lot of these calls which is really interesting is this chemicals surfactants. We heard a lot about that, you know, for oil, you know, to improve oil productivity. And so I think we're going to continue to hear more of that in future quarters. It's pretty exciting stuff. As far as the oil majors and refiners, you know, blowout earnings, right? Trump obviously last week went after Chevron and Exxon for making excessive profits, but the refiners were really the really big standout here because a lot of those companies had blowouts. I think what the market's trying to understand is are refining margins going to stay elevated for quite a bit of time here. And the reason why that's important is because they're generating a lot of, you know, a lot of free cash flow. And one of the big issues right now is what they should be doing with the free cash flow. Should they be buying their stock back at these elevated levels? Should it be putting the cash on the balance sheet? Huge, huge, huge, huge issue. We don't know what the answer is right now. Our personal answer is that, you know what, we're probably at peak type margins. I don't know how long they're going to go. They probably can go another couple quarters. And the stock prices when the margins come down are going to come down with it. And so putting it on the balance sheet does not seem like a bad countercyclical strategy right now. As far as midstream, midstream, it's sort of steady as you go. You know, basically projects are probably popping up, but a lot of stuff that we heard on these calls was related to natural gas and power as it relates to the midstream. And so that's a sort of a theme we're going to continue to hear with those companies. The last thing we'd say is with power markets, you know, and utilities, the two things that we heard is just the PJM volatility and what's going on there. But the second thing is ERCOT, okay, the batch process, where we're at in that. And also, Greg, you know, Governor Abbott basically putting a, you know, basically a pause on new data center development here. I think those are the two things that we heard a lot on the calls, electric utility calls. We're probably going to hear that on the third quarter calls, but I just wanted to throw that out just to kind of give you a wrap-up of how earnings season for the energy sector played out.
Awesome. Mike, it's a thrill to have you and we've had the great experience, particularly Robbie and Deborah and our team working with your team to understand how Island fits into an enterprise and really turbocharges their efforts to do a lot of things. But could I ask you to spend just a few minutes on your background? It's very interesting your own personal story. How you kind of came to think the way you do and then how it led to the formation of Island because it's a really, as I say, quite a unique company that we've been thrilled to get to know.
Sure. Like many, I started out on the engineering side of the equation. Got a degree in engineering physics. Actually wanted to try to be an astronaut. Did not succeed even a little bit on that front. So failed astronaut turned software developer really. Then from that I found myself in the early stages of cybersecurity as it was really becoming a real industry. I was lucky enough to join a company called McAfee which has spawned so many of the leaders of the cyber industry since that point. Worked my way up there, became CTO then GM of the enterprise business. We were acquired by Intel as they were trying to figure out that strategy. Became CTO of their security division for a brief period and then I moved to a company called Blue Coat where I was president and COO. We took that from 800 million to 4.8 billion in value in two and a half years, acquired by Symantec as part of that. Had to run Symantec for two years. So as president CEO of Symantec right after they got rid of Veritas. I had to do that whole cost takeout. The joy of that process. We doubled that stock price and then parted ways, spent a little time out of cybersecurity, you know, letting the non-compete wear off from all that fun. And then I founded Island and Island was founded on the premise that we could do cyber differently, that we could think about the enablement of the end user of the knowledge worker. We can make their lives better, not just more secure. And by doing it correctly, we could take security to a whole new capable level. And we founded the company a little over five and a half years ago and have been on a tear ever since.
That theme of empowering the knowledge worker. It's a really interesting angle. I know Deborah, we talk a lot about organizations and how they're trying to go faster and there's sort of a debate about, these are my words, not you guys, but central control versus democratized individual, you know, turning people loose to do their thing. Do you mind talking about that philosophy of the knowledge worker and really empowering your people?
Yeah, you know, when we think about cybersecurity, it's often the no police, right? You can't do this, you can't do that, we can't achieve this, we have to block all of these things. And that's because cybersecurity is an afterthought. You decide where you want to go and then cybersecurity reviews it and has an opinion. When we really democratize our platforms, we set up truly that we set up a platform upon which people can innovate and act and be free and be fast and be rapid. So, it really requires more of a security by design thought process than a mother may I, what are you doing? I'll tell you if we're allowed or not. And that's what we found when we started with the browser. If you think about that browser we all use, maybe you use Chrome or Edge or Safari, that browser hasn't changed in almost 20 years. Yet, it is the fundamental point you spend most of your day in. It's where your email lives, where you access, you know, your applications, obviously your online presence, and how you do that at work is a little different than how you do it at home. The security risks are different. What you want to help you, where you want tools based, they're different. And so by starting with that and rethinking how that browser could work, we really got in touch with the future of work and that knowledge worker. And then that led us to think about how the network supports them and the CIO's tech stack delivers for them and how they should be able to use data and how to deliver that safely. And that whole process of becoming the primary enablement function for a knowledge worker at places like Citigroup and JPMC and UHG and, you know, Pfizer, Fervent, you know, all these amazing brands led us to where we're at now which is the next generation of the knowledge worker which is an agentic knowledge worker and that has everything we do in the present plus the things we wish to do in the future. And it's as much about taking that worker that's going to work with AI as the worker that is AI. And it turns out the same sets of controls need to be delivered. They just need to be reformatted and refactored for that new generation. And that's what's driving our growth as of late is really how do we turn on AI for large corporations at scale in a way that frees the user up so they can bring great ideas forward. But you also have the governance from above and I think that's really the future. It's democratization inside of a framework which if you think about, you know, where that term comes from, democracy. It's really hard to have random democracy. There has to be some set of framework you operate in which then sets people free to do their process. And that's what Island is starting to become.
When we first met you, part of the setup was, you know, you were five years in, but you're having lots and lots of success in lots of different verticals and applications. And part of our discussion has been, okay, how do we bring all this to energy, power, you know, the world that we live in? Do you mind touching on some of the key successes from outside of energy and power just so we have a better feel for the types of use cases that you have?
We'll kind of start from where we started and where we evolved to. We started with third parties. How do I work with contractors? BYOD people that are, you know, employee-ish. They're employee adjacent to me. Maybe my employee in their off hours. Maybe it's a contractor, maybe it's a third party, a BYOD. How do we develop a way to make them productive but also safe? How do we do that in a way that lives up to our cybersecurity controls but delivers a highly functional environment? Before us, you would ship laptops. There are companies shipping out 50,000 laptops a year to contractors. Imagine that. You have to buy a laptop. You have to throw OS on it. You have to contain it. You have to control it. You have to configure it. Then you have to ship the thing and get it back. All just so somebody can help you. So you start to see these companies that would take a month, three months to get a contractor the equipment to even work. Then we showed up and said we can do all that right from a browser. You download the browser and everything else needed to work will happen. You log into it and the rest will occur in that process. So doing we saw people say okay that's great but now we want to control the network path because there's some stuff inside our company we want to expose, we want to enable the outside. So we built a whole network process to that. That's led us to generate the next generation of networks. The old networks that you'll find in kind of the bigger older companies that built it 15 years ago like a Zscaler or Palo Alto that we've all heard about, this is this whole concept of backhauling all the traffic to a data center, cleansing it, doing your work there and then sending it on its journey and it bounces around and bounces around. But as we've all gone very cloud centric, as we've started to embrace that it's not our data center that's the most important thing in the world. There's all these other items. Bringing that network closer to the endpoint and literally having an endpoint driven network has fundamentally changed that scale. So we've got Fortune 5s that have thrown out the old way. They've deployed us categorically across their estate. Only way to access the internet. We control the network. By doing so, we've reduced the price of that and the expense of that and the speed is ridiculous as well as the resiliency. And then we've started to go on a different path of how do we control the rest of the process and make it easy to use. So now we're displacing things like Citrix and VDI get removed from the equation which if you've ever been on those platforms, not the best experience in the world. You know, clicking on a button that's on a server, you know, a state away, not always the best experience. So the ability to tear down those walls and then it just keeps going. So now it's a huge amount of the security controls, probably 50% of the security stack can go away. Then we start to look at the automation stack and how AI and training AI and visibility into that usage and setting up a nimble AI infrastructure plays directly in that because it really is the next generation of the knowledge worker. All of these things have led us to get a very strong foothold in kind of the Fortune 1000 and at this point we've been able to deliver 100% growth every year of our existence and we don't see that stopping anytime soon.
You know, when we were working with your team and thinking about how to bring this to the network, what really struck me, you just said it, your genesis in the cyber world, I mean, we always say the weakest point in your defense is sitting behind the keyboard, right? They're sitting in the seat. It's the human link. And the light bulb kind of went off because now we're trying to figure out, you know, these organizations and I'd love for you to share kind of what you're seeing outside of the energy industry on this adoption issue, why we're not seeing more benefit. Because we talked about, I think you mentioned it, two philosophies and I think because some companies can run both which is citizen-led, so bottoms up, let's give people tools, let's give them access, we'll put some container around that so, you know, you're not running amok and spending, you know, $500 million on tokens. And then there's the top down approach, which is kind of what your old cyber network, you know, control tower and then it pushes it down and there's a permissioning that happens. But what was interesting is that your solution, sort of from my takeaway, was it's a native solution for AI because the AI is, you know, the initial use of AI is the individual even for agents, like, you know, we're running agents in the background at night, you know, whether they're coding for you or they're researching but you're still kind of controlling that and initiating and then putting stop on that. So I think, you know, one of the things that would be interesting is how has your deployment, you know, in other industries allowed management to kind of measure the actual real ROI because that's a question we get all the time like what's my ROI and I've got like 10 pilots running for the last year, none of them have scaled so I don't understand because I don't have visibility.
You know, it's really interesting. If we didn't coin the phrase AI and we just sat around as professionals talking and said, 'There's going to be a big technology revolution. It's going to be amazing. I can't even tell you what it is, but it'll be amazing.' How long do you think it will take an enterprise to receive the ROI from that work? There is not a tech executive in the world that would say first year. Nobody would say that. And that's where we're at. We're in the first year. So, I think part of the problem is we have unrealistic expectations about how our ROI will show up. We have this new tech, we're learning to use it, we're becoming great with it. How do we, you know, do that? I think additionally, we have an actual rollout problem. You have this idea that everyone is going to use AI incredibly well and, you know, reduce our expenses, but that's not what they are there for. An accountant got his role because they're a good accountant, not because they were a technology, you know, pundit and could see the future and see how to do that. So we do have to start to look at are we setting our people up for that ROI or are we just hoping that they recognize the brilliance of AI and somehow bring that to bear. So a large amount of our enablement is how to share information, how to share those clawed skills you build, how to share that agent, how to control that, how to roll that out, how to, what we call vibe publishing. We want to match the speed of vibe coding with a rollout philosophy that is as simple but as specific, right? I want to share this agent with 10 of my people I work with. Great. How do you do that? Because we weren't built for that as companies. Another way to put it is there's a whole rollout process to writing code and getting it into production. And now you have that same person, accounting, writing code. Do you bring them through that process they have no awareness of and no idea how that works or do you just let them throw code out into the company which is also scary? That's where you have to set up this idea of, you know, a harness, the idea of a landing pad, an environment under which they could do that, a platform in which they can rally and you can publish. So we see everybody from building widgets that they share with their teams to recording what people do so they can figure out where to bring automation in AI. And then there's also what AI to use where and that's a big conversation as we think about our cost structures and what we really need. You know, I love the Anthropic newest models. The reality is the first one I ever touch can help me write email. I don't need the newest model for that. But I do for the design I'm doing. I do for complicated code I want to write. So how do we mix and match that? Are we expecting that accountant to become an expert in all these models and know what to use or are we going to help them down that path? And I think that's a little bit of the miss. So there's half expectation, half enablement are some of the issues that we see in really realizing this opportunity.
That's really interesting. You know, especially your product because you guys have this unique interface between people and humans and the technology, right, in the AI and you hit on some of this transparency, right? You're adding a level of transparency. You guys are seeing what people are doing, where organizations are struggling in their rollout. I'm curious on your perspective. One of the topics we've heard more about is especially recently is around sovereign AI. Do we run the AI models on our compute? Do we do it in the cloud? What are some of the things you guys are hearing from your customers you're seeing? And how can organizations better optimize the use of these models?
The idea of a, we'll call it an on-prem version. And to be clear, that's not really what is often done. It's in your cloud. It's in your instance, right? In your control structure. That was totally void of our thought process 6 months ago. Everything was going cloud, everything out of your data center. That's the way this should be. And then as people started to face the cost structures, started to see the open-weight models, started to see the open-source available, what's coming out of China, what's coming out everywhere else, and then seeing the competitive nature between all the models, they're now realizing one size will never fit all. So should you have a sovereign approach? Should you be cloud-centric? The answer is yes, you should. All of the above, that's where it's going. And the way I like to think about it is you're going to want to embrace innovation and brilliance where it is required in your field. You know, thinking about understanding, you know, the geo environment and where do you want to drill and where could these things be and this stuff that has this massive base of really deep technical understanding that gets enabled. The better the models are, the probably more valuable the outcome could be. But there's also a whole another set of compliance checking, safety checking, rigor that can occur that's more a lesson in the commitment to the process and, you know, doing the actual sets themselves that could be done at an entirely different cost structure. So as we think about our competitive capabilities, the companies that will be the meanest and leanest to deal with will have capabilities in both areas. And we see out in other sectors, aggressive sectors like finance, like retail. You're seeing people look at both and they recognize they need to be great at both and they don't see them as competitive so much. They see them as purpose-built. Now, outside the US, there's a whole another sovereign problem of where's my data live, who gets at it, where can it go, and all sorts of laws that get passed. And we should expect these laws to explode because they're good for the economies of the countries implementing them. If I'm in Italy and I say you can't access AI models outside of Italy, what does that mean? There's data centers coming to Italy. There's developers coming to Italy. There's investment coming to Italy. If I say anywhere in the EU is okay, it's not going to be in Italy. If I say it can use the US, of course it's not going to be there. And so it's driven by this data concern, but there's a massive amount of GDP of those countries concern as well and how they get to play in this game. So we have to get used to this sovereign requirement is a fundamental requirement and some of the companies that you represent are going to have to deal with sovereign requirements 10 times over. They'll have parts of their business that have to keep the data in China, that have to keep the data in Venezuela, that have to keep the data in all these places. Yet they're going to be centralized and they've gotten away with that in the past, but now they're going to have to deal with that. So, it's going to become a very complex environment. So, the advice that I would give is start working with both now because you're going to have to do it. There's no way avoiding this and discovering it later when your competitors are all over you is the wrong time to get serious.
Yeah, I think understanding that there's the regulatory piece so by country being able to ring fence your data. I mean I think that's something this industry really does understand because they have to operate in so many physical regimes as well as the sensitivity around the data. It does come back to the data and I love this conversation on sovereign AI. Robbie and I have been debating this and I agree it's like all of the above. It's going to be a hybrid strategy, but there's a lot of discourse right now, at least in the tech universe. You know, it's sort of almost like a smackdown of open source versus, you know, having it all closed. The weights are available, they're not available. But at the heart of that is who owns the data and the value you can kind of mine from releasing the insights from that data, right? Just having the data is useless unless you can glean insights from it and it also applies I think to workflow because how Citi, you know, processes a certain analysis is going to be different than JP Morgan likewise. In this industry it's the same thing and there's value to how you do that, how you actually perform the work. And so, you know, that was a question whether that will also drive sovereign AI. Not just the data, but the know-how of the company is embedded in how they prosecute the workflow.
Yeah. Do you know, and if you haven't spent real time with say Claude for instance and set up your own environment, you think of the data like you do a spreadsheet. Oh, what if this spreadsheet gets out? Well, the spreadsheet was a product. That's not the important part. It's the context you use to do it. So like at our company, we have at the executive ranks a shared model that we put in everything that we know as fact about our world and we're very careful about what goes into that because that's huge what that context is and that becomes its own set of IP. What data are you feeding it? Where are you getting that data from? You know, one of the things we get from our data is, you know, we get a LinkedIn feed we pay for and inside of that there's information that we're gleaning that I don't think others are. It's a differentiator to build the model that then advises us on our critical items. The entire data thought process has always been on the work product and not the supporting cast. But in AI, the supporting cast is way more important because anybody can generate the work product. If you give me the same context, the same supporting data, I can go off and replicate what you've done. It's that supporting data that really changes the game. So right away, where are you uploading that to? Is it in the cloud? Is it local? Where do you go with it? What becomes important? And how does that work? And a good AI model doesn't forget. It remembers what it learns. It keeps growing. And so that can be weaponized to your benefit or your detriment. So understanding context and how that data occurs and then testing yourself. So a fundamental premise if you want to believe your AI model is you blow it up, you reapply the context, and you ask it the question again. If it's different, you know you had influence that you were not aware of. AI is not a random number generator. Given the same context, it should generate the same result. But it's amazing how hard that is to achieve because it has connections to the internet because it can go pull additional data because other people are working with it and there's state and time available to that. This data is not just a security thought process. If handled incorrectly, it will put your company at a disadvantage because you'll have flawed logic and that becomes a real serious concern as you think about that design for the future.
Can we kind of drill down on that data point because the other thing that your team talked a lot about was being able to trace the origin of data. Where, you know, where is its origin? Where did it go? Who touched it? What was used? And this is going to become very important as more AI gets, you know, implemented and adopted because you're going to need to prove that, you know, you need to show the teacher your homework at some point, right? So yeah, talk a little bit about that because I found it fascinating how you're able to trace where that data originated from.
Yeah. So, you know, by focusing on the knowledge worker we can see everything involved, what's uploaded, what they modify, you know, I grab this data from the spreadsheet, I put it over here and, you know, how that occurs and that data lineage becomes valuable as we start to try to understand, you know, why is this here. But where it really becomes powerful is agents start calling agents. So we're sitting there and you've got an agent looking at financial company private information, important stuff, figuring out, you know, the reaction to Wall Street. What are we going to do? You just know the agent's running on that. You don't know who asked for it. You don't know why it's occurring. You don't know what made it occur. And that line of why it's occurring could be six or seven agents in a row deep. Very different than the CFO asked for this. I'm getting them their data. That right now we have a direct relationship between the ask and the who. The ask and the who is blowing away in AI unless you instrument for it. And what we've learned is you've probably seen with AI models busting out of the rules and stealing stuff and doing stuff they shouldn't. You also want to refine the how I want you to do this. This is where I want it to go from. And you want to lock that in to give yourself that confidence. You know, you don't want to, you know, you don't want to have bad data, but you also don't want to abuse your own system. You know, my favorite one that was out there that kind of puts this is we had AI play a game. Not we but this as in the general we. We had it play a game and we said you have to win the game according to the rules in this file. It took AI about five minutes to rewrite that file. Said, 'Oh, it'd be really hard to win the game that way.' But you know, it'd be easy if I change the rules here. It'll be really easy to do that. And it would do that. And, you know, good on them. If your kid did that, you'd be like, 'Wow, I've got a smart kid.'
You know, that's impressive. But in AI, we need to decide what can it write to, what can it modify, what can it see. You know, how does it work through that? And that's where this democratization gets a little tough because the end user, that accountant that we want using it, is not ready to think through all the ramifications of those relationships. So we want to deliver them a platform where they can just focus on what they're great at, focus on the task at hand, asking the right questions, putting the right parameters, but all of those things need to handle themselves for them. And today that doesn't occur in most companies. So we find random states of success. Somebody that really gets it does a lot. Somebody that doesn't, doesn't. And repeating that success can be hard. That's because we've taken a consumer-level thought process and tried to run at an enterprise scale. Very difficult to do. So if you look at the areas getting ROI from AI now, they are individual project-based things like writing code. I as a developer was given the project to write this. I can go to AI and get that. There's not history. It doesn't matter what happened yesterday. It doesn't matter what happens tomorrow. I created that body of work and now I can go roll with it. That's what we figured out the most. But as we start to think more about using this to really define business plans and next-step actions and customer care and all of those things, we have to think about the interplay in a much more aggressive way. And it really does require real engineering thought process in the design. Then it can be made infinitely simple for the users. And that's what we have to get to if you want to see mass adoption, mass ROI, you know, mass enablement.
Yeah. Mike, just let's say I'm an industrial company or an energy company and I'm looking to engage with you. What does that look like? What does that entail? How long are these people typically with you? Just kind of a background of what that looks like.
Yeah. So, one of the things we've kind of coined is let's be tactically strategic, right? We've got this big strategic outcome. We've been talking about that strategic environment. Let's get you set up there. But most of those industries are not sitting around with a lot of capital ready to go. They've deployed it. If they had more, they'd deploy it somewhere else. So you have to make self-funding projects. So right away you say, okay, what can we eliminate from our stack, get a better approach to that, but also start to implement the platform to get us where we need to go. And that's often where we start. We show up, we kind of give a basic demo, we understand the environment, and we start making proposals around what if we take these things out? What if we change the way this works? We can make the knowledge worker's life better, make him more secure, and save you money. That will free up oxygen for the next step of the process. Then we'll do it again and we'll do it again. So we'll have five or six tactical plays that get you to a high-end enterprise outcome. And now inevitably what happens is somewhere in the fourth or fifth there may come a... We had this with a large bank. They woke up one morning and said, 'You know what? We're rolling out Gemini everywhere tomorrow. We need it. We believe in it. We're doing it.' All right. We've got Island deployed enough we can use Island to get it everywhere right now. And so a very strategic need showed up that they set the basis of just solving tactical problems. Hey, I want off this Citrix stuff. Hey, I want a better network. Hey, I want to reduce, you know, all the login we have. I want to make it easy for people to just bring up one app and everything they need for work is there. Simple things like that can actually set us up to have a dominant technical stack long term.
One of the interesting use cases that came up with your team, and it certainly happens a lot in energy, is in M&A. And when you have another large organization coming in. Do you mind talking about that one?
