Matt Brown Show - Conversations That Power The Business World.

Simon Taylor on AI as a cyber risk multiplier and the need for human-centric leadership frameworks

40 min Listen to the original
CybersecuritySaaSAI RegulationData ProtectionLeadershipEnterprise Software

Summary

Simon Taylor, CEO of HYCU, discusses how agentic AI exponentially increases cyber risks by enabling automated attacks at scale. He highlights the fragmentation of data across hundreds of silos and the critical gap in identity management. Taylor introduces the HER framework (Humanity, Exceptionalism, Resourcefulness) for leaders navigating AI adoption. The conversation also covers the erosion of traditional apprenticeship models and the ethical implications of AI transparency and deepfakes.

Key takeaways

  1. 1
    Simon Taylor asserts agentic AI allows attackers to scale operations, making 100 trolls act like 100,000.

    Simon Taylor · 2:38

  2. 2
    Look at CrowdStrike. You know, so a lot of the things that we predicted in the book around this sort of SaaS apocalypse have already happened.

    Simon Taylor · 5:30

  3. 3
    Simon Taylor states HYCU reduced integration build time from three years to three months, targeting three weeks.

    Simon Taylor · 13:00

  4. 4
    Simon Taylor warns AI automation is nullifying the value of junior employees by destroying apprenticeship models.

    Simon Taylor · 25:44

Episode sections

  1. 0:05 – 1:39

    Introduction and HYCU Overview

    Matt Brown introduces Simon Taylor, CEO of HYCU, describing it as the world's fastest-growing SaaS data protection company. Taylor explains HYCU protects data across on-prem, public cloud, and SaaS environments, citing 86 integrations, 4,600 customers, and a 91 MPS score.

  2. 1:39 – 3:58

    AI as a Cybersecurity Risk Multiplier

    Taylor argues AI exponentially increases cyber risk by enabling attacks at unprecedented scales via agentic AI. He disputes the notion that AI aids defense equally, stating attackers can now simulate thousands of threats with minimal effort.

  3. 3:58 – 10:21

    The SaaS Data Apocalypse and Supply Chain Risks

    Taylor discusses his book 'Averting the SaaS Data Apocalypse,' noting that AI exacerbates data silo sprawl. He cites the CrowdStrike outage as an example of how SaaS dependencies can take the world offline, validating his earlier predictions.

  4. 10:21 – 13:21

    HYCU's AI-Driven Marketplace Strategy

    Taylor explains HYCU's shift from building integrations manually to using an AI-powered low-code marketplace. Partnering with Anthropic, they allow SaaS vendors to build integrations in weeks rather than years, leveraging models like Claude 2 and 3.

  5. 13:21 – 16:17

    Data Access and Identity Management Risks

    Taylor identifies access management as the primary data gap, not volume. He warns that tools like Okta and Google BigQuery create risks if source and target data aren't protected, leading to potential leaks when employees input confidential info into LLMs.

  6. 16:17 – 21:07

    Regulatory Landscape and the HER Framework

    Taylor criticizes current AI regulation as insufficient ('underboard') and proposes the HER framework: Humanity, Exceptionalism, and Resourcefulness. He advocates for a human-first regulatory approach to protect societal well-being.

  7. 21:07 – 24:39

    Speed of AI Adoption and Societal Impact

    Taylor expresses surprise at the speed of AI adoption, comparing it to the rapid shift to ChatGPT versus the slower cloud migration. He notes AI has moved beyond technology into the core functioning of society and business operations.

  8. 24:39 – 31:40

    Loss of Apprenticeship Models and Leadership Skills

    Taylor warns that AI automation threatens traditional apprenticeship models in fields like law and coding, creating a skill gap for junior employees. He advises CEOs to cultivate empathy and purpose, arguing that humanity remains critical in an AI-driven economy.

  9. 31:40 – 32:48

    Leadership Framework: Humanity, Exceptionalism, Resourcefulness

    Simon Taylor outlines a leadership framework for CEOs navigating AI, consisting of three pillars: Humanity, Exceptionalism, and Resourcefulness. He argues that integrating AI into corporate goals is now essential for success and that AI should be viewed as a tool to increase team capacity and efficiency.

  10. 32:48 – 34:40

    Ethical Considerations and AI Transparency

    Matt Brown discusses the ethical implications of AI deployment, citing a test where labeling a product as 'powered by AI' caused a 60% drop in engagement. He questions the necessity of transparency and the potential for AI-generated content to deceive customers in B2C markets.

  11. 34:40 – 36:17

    Perception of AI Quality and Generational Shifts

    Simon Taylor explains that early AI experiences were likened to talking to a high schooler, but modern LLMs are comparable to graduate-level students. He predicts that younger generations will care less about the origin of content and more about speed, shifting the primary concern to ethics and integrity.

  12. 36:17 – 38:09

    Regulation and Deepfake Risks

    The speakers discuss the rapid evolution of AI, noting that AI can now create realistic videos of people, raising concerns about identity and consent. Taylor advocates for slowing down to study regulation and understand the impact on society, particularly regarding the use of likenesses in training data.

  13. 38:09 – 39:20

    AI in Elder Care and Loneliness

    Simon Taylor predicts that AI will significantly transform elder care by addressing the epidemic of loneliness among the elderly. He suggests that AI companions could provide meaningful interaction, contrasting this with potential negative impacts on human relationships.

  14. 39:20 – 40:12

    Market Growth in Elder Care and Robotics

    Matt Brown mentions speaking with a franchise owner of 'Elder Kid Network' who reported explosive growth due to the elder care epidemic. He speculates on the potential role of robotics, specifically referencing Tesla's Optimus robot, in this sector.

Assets discussed

AssetKindStanceContext
Anthropicprivate companyneutralSo we called up our friends in Enthrame. Anthropic and we built in a partnership with them. We leveraged their their current most their latest model. It started with Cloud Two and then three and et cetera, et cetera. And we leveraged their their AI model so that we could enrich it and then create an automated AI framework so that you can now easily within two weeks build a software integration between your SaaS service and Haiku. And I will tell you that you know, it takes most data protection companies upwards of a year to build one of these kinds of integrations. So let's say you're a Salesforce, you're a 365, you know, you just sort of name it. You can literally call up Haiku, get on the website, and enter our low-code development platform. You'll be leveraging Anthropic to do it. And very, very quickly, you'll be able to spin up your own integration and add that money. module into the Haiku marketplace, providing your customers with end-to-end data protection as a service.

“Stance” is how the speakers talked about the asset, not a recommendation.

Claims worth verifying

HYCU is the world's fastest-growing data protection as a service business and the number one leader in SaaS data protection.

Simon Taylor ·0:53

How to verify: Check HYCU's official website or press releases for growth metrics and market position claims; verify customer count and MPS score via HYCU's public reports or third-party review sites like Gartner or Forrester. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

HYCU protects over 86 SaaS integrations, has 4,600 customers in 78 countries, and has a 91 MPS score.

Matt Brown ·1:27

How to verify: Verify via HYCU's official website, investor presentations, or press releases detailing their customer base and integration list. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

AI allows for attacks at scale where 100 trolls can act like 100,000 using agentic AI to build bots.

Simon Taylor ·2:38

How to verify: Review cybersecurity industry reports from firms like Mandiant, CrowdStrike, or Microsoft on the impact of generative AI on attack volume and automation. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

The average company has data in over 300 different data silos.

Simon Taylor ·4:44

How to verify: Search for industry studies or white papers from HYCU or data protection firms citing the average number of SaaS applications per enterprise. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

The CrowdStrike outage took the world offline.

Simon Taylor ·5:30

How to verify: Verify via news archives from major outlets (e.g., Reuters, Bloomberg) covering the July 2026 CrowdStrike incident and its global impact. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

HYCU protects 86 SaaS integrations, while the rest of the Gartner Magic Quadrant at the time only protected five.

Simon Taylor ·6:10

How to verify: Check HYCU's integration list on their website; compare with Gartner Magic Quadrant reports for Data Protection software from the relevant period. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

HYCU partnered with Anthropic to build an AI framework using Claude 2 and Claude 3 models.

Simon Taylor ·11:56

How to verify: Search for press releases or partnership announcements from HYCU and Anthropic regarding their collaboration on SaaS integration tools. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

HYCU reduced integration build time from over a year to two weeks, and from three years to three months.

Simon Taylor ·12:36

How to verify: Verify via HYCU case studies or technical documentation describing their low-code marketplace capabilities and development timelines. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

The CrowdStrike outage is an example of SaaS services taking the world offline.

Simon Taylor ·5:30

How to verify: Review news reports from July 2026 detailing the scope of the CrowdStrike outage affecting global airlines, hospitals, and banks. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Elon Musk built companies using a tough leadership style, contrasting with the need for empathy.

Simon Taylor ·30:00

How to verify: This is a subjective opinion/characterization; verify via biographies or leadership analyses of Elon Musk's management style at Tesla and SpaceX. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Simon Taylor states that HYCU is the world's fastest-growing SaaS data protection company.

Simon Taylor ·31:40

How to verify: Check HYCU's marketing materials or third-party SaaS growth rankings (e.g., SaaSQuotient, Gartner) for claims of 'fastest-growing' during the relevant period. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

A test showed a 60% drop off in engagement when a product label included 'powered by AI'.

Matt Brown ·33:36

How to verify: Request the specific study or focus group data from the best-selling author mentioned by Matt Brown; verify via the author's published work or press releases. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Simon Taylor's company, HYCU, uses private LLMs (referred to as 'Haiku' in transcript) internally.

Unidentified speaker (UNKNOWN) ·34:43

How to verify: Check HYCU's technical blog posts, engineering case studies, or press releases for mentions of internal AI models or partnerships with Haiku AI. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Simon Taylor claims modern LLMs are comparable to talking to a graduate or doctoral student.

Simon Taylor ·35:37

How to verify: This is a subjective analogy; verify by reviewing Simon Taylor's public interviews or articles where he elaborates on this comparison. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Simon Taylor references the actors strike occurring 'a couple of years ago' regarding AI image usage.

Simon Taylor ·36:37

How to verify: Check news archives for the SAG-AFTRA strike dates (July-November 2026) and verify if the podcast date (2026-11-19) aligns with 'a couple of years ago'. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Matt Brown mentions speaking with a franchise owner of 'Elder Kid Network' who reported explosive growth.

Matt Brown ·39:20

How to verify: Search for 'Elder Kid Network' franchise news or press releases from late 2026 to verify growth claims. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Matt Brown speculates on Tesla's Optimus robot being helpful in elder care.

Unidentified speaker (UNKNOWN) ·39:35

How to verify: Check Tesla's official announcements or investor presentations regarding Optimus robot applications in healthcare or elder care. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Do your own research

  • Verify HYCU's customer count and MPS score against their latest investor deck or SEC filings.
  • Check Gartner Magic Quadrant reports for SaaS Data Protection to compare HYCU's integration count with competitors.
  • Review the CrowdStrike outage root cause analysis to assess SaaS dependency risks.
  • Research the 'HER Framework' (Humanity, Exceptionalism, Resourcefulness) for additional leadership resources.
  • Investigate the impact of AI on junior-level hiring trends in law and engineering via industry labor reports.
  • Check for recent deepfake legislation and compliance requirements in target markets.
  • Evaluate the ethical implications of AI-generated content disclosure using consumer trust studies.
  • Review Anthropic's partnership announcements with HYCU for technical details on the AI framework used.
Full transcriptexpand

0:05Unidentified speaker (UNKNOWN): Hi

0:06Matt Brown: everyone and welcome back to yet another cracking edition of the Map Round Show. This is the AI Advantage series where we are talking to CEOs and business leaders all about the impacts of AI on their industries. And in with us in the hot seat today uh is none other than the myth, the man, the legend, uh Simon Taylor.

0:26Simon Taylor: Hey Matt. how are you?

0:27Matt Brown: Good and new, man. Good to see your your uh your beautiful face again. Uh but Feynman is uh the CEO of the world's fastest growing SaaS data protection company. Um and so Simon, welcome back.

0:39Simon Taylor: Matt's great to be here. Thanks so much for having me on. Always a pleasure.

0:42Matt Brown: Yeah man so obviously uh I know a lot about Haiku and what uh you guys are doing but for our audience around the world that doesn't know anything about haiku uh, what's the elevator pitch? What are you guys up to?

0:53Simon Taylor: Sure. So you said it right. I mean, so haiku stands for hybrid cloud uptime, and we are the world's fastest growing data protection as a service business. And we focus on protecting data no matter where it resides, on-prem, public cloud, and SaaS. Uh, I think our claim to fame is that we're the world's number one leader in SaaS data protection. So everything from Office 365 to Salesforce to Okta and everything in between. we protect over 86 SaaS integrations, uh 4,600 customers in 78 countries to date and a 91 MPS score. So customers like it, they use it, and they stay protected using Haiku.

1:27Matt Brown: Brilliant. Brilliant stuff. So um let's uh start with uh potentially a blunt question here. But obviously AI is I call it the AI echo chamber because everyone talks talk it's there's so much noise about AI and obviously you are at the cutting edge of cybersecurity. So um from your perspective with all the customers that you have four thousand six hundred seventy six countries and so on, do you feel that AI is more of a cybersecurity asset or liability right now?

1:58Simon Taylor: So anything that is digital is a risk. Anything that is connected to the internet and is online is more of a risk. Anything that is technically advanced increases the risk. And AI exponentially increases the cyber risk. I think what a lot of people will tell you is that you know there's two c two sides to the cyber coin when it comes to AI. One is the defense and one is the offense, uh, meaning that it is easy, it is much easier. for an attacker to attack you, but it's also easier to defend. I think that's completely untrue. Um, the reality is that AI allows for attacks at scale in ways that are unprecedented. Meaning that, you know, instead of it being a hundred trolls in a troll farm that are hitting you, uh, now those hundred are acting like they're a hundred thousand because they're able to use agentic AI to go after you, to build bots, to really, really disrupt you in different ways. I also think that the the way that we tend to defend is we look at all the different places, the different threat vectors, and then we try to defend each since every one. The reality is that with AI, it's full throttle. I

3:04Unidentified speaker (UNKNOWN): mean

3:04Simon Taylor: they're gonna def they're gonna be hitting you from all sides. I ultimately think that we're gonna get back to the point where we are almost entirely air gapped as a civilization. And the only fun the only truth. true safe thing is going to be sending a letter to your relatives. And I mean old fashioned posts. Now, you might say, I'm a cyber guy. Why am I saying this? But the truth is that I think we need to understand that that is the safest, right? But it's also completely impractical. And so I think, you know, the world today is all about studying and figuring out what are the steps we can take acknowledging the AI threat and the and cyber the implications. of AI or the AI implications of cyber, and then sort of, you know, turning that around on itself to say, well, what where what are my most valuable assets? How do I make sure they're protected? Uh, and how do I make sure that, you know, because AI is going to be so prevalent, that I've got the right tools and the right setup to actually keep myself safe.

3:58Matt Brown: So you wrote a best-selling book called Diverting Averting the Nass SaaS Data Apocalypse a couple years ago. Where are we now with the apocalypse in the context of AI? Do you feel like the apocalypse is going to be accelerating, you know, towards us from the future to the present day? Or do you feel like, you know, with AI and the advancements of AI from a defender's perspective, do you feel like that apocalypse hasn't, you know, it's being pushed back?

4:24Simon Taylor: Yeah. So so averting the SaaS data. apocalypse, as you know, Matt, was all about how the explosion of SaaS, you know, software as a service applications and databases, was going to make it virtually impossible to get the kind of data protection, the kind of backup and recovery capabilities you were going to need in the future. And you know, we mentioned in the book that, you know, the average company today has their data in over 300 different data silos, meaning different SaaS applications, clouds, et cetera.

4:50Unidentified speaker (UNKNOWN): Um,

4:51Simon Taylor: that only gets bigger with AI. You know, every agent uh ultimately becomes a new data silo. Every vendor you add becomes a new data silo. And I think that the trouble is that because of AI, the again, the attackers can go after all those different data silos in a lot of different ways. You know, you you asked me i if AI is is is propelling us towards the apocalypse or preventing it. Um I think it is propelling us towards it. And I and I think that, you know, certainly when we look at, you know, even what's happened in the last year. You know, I would say, you know, three, four years ago, let's let's go back to COVID, all anybody talked about were ransomware attacks, right? Somebody's gonna send you an email, you know, spear phishing attack, get in there, charge you a bunch of money. Now we're seeing vicious supply chain attacks. We're seeing problems where whole SaaS services are taking the world offline. Look at CrowdStrike. You know, so a lot of the things that we predicted in the book around this sort of SaaS apocalypse have already happened. They're happening every day. And so I think the good news about that, it's not all, you know, gloom and doom. Uh the good news about it, Matt, is that the world is starting to wake up. You know, um, haiku today has 86. eighty-six different SAS integrations that we back up and recover. You know, the rest of the the the entirety of the Gardener Magic quadrant at the time when we built the platform only

6:10protected five. Only protected five. Um we've got eighty-six. We think there's about two hundred mission critical apps and databases that have to be protected. And frankly, we believe that legislation should force that because it's true that, you know, the more protected we are as a society, the safer we're all going to be. The reason these attacks increased in frequency is because they worked and because people were getting paid. You know, if nobody ever paid the ransom, right, these attacks wouldn't be happening because it would be unprofitable. But these attacks work and they continue to work. And until people have all their data recoverable, they will always work. Um, and I think the AI A AI is only gonna going to spur that on, make it easier and make it faster. And we just need to be prepared.

6:53Matt Brown: You mentioned the AgenTech AI uh earlier on. And um it seems to me that that is a very hot commodity in terms of the AI echo chamber as I like to call it. Uh because you can start to automate, integrate with different SaaS applications or any application really. Multiple agents, you pulling data from one system, pushing it to another. It's meant to uh really accelerate organizational capacity as an example. Um, do you find that SaaS, traditional SaaS uh sprawl

7:25Unidentified speaker (UNKNOWN): um

7:26Matt Brown: versus say a Gentec AI? Do you feel like one of those is gonna be compound this apocalypse uh more than the other?

7:34Simon Taylor: Yeah. I think there's an even bigger question in there, which is this sort of I I think um sort of nonsense that's out there that SaaS is going to disappear and these companies will all disappear. You know, technology evolves. It's the nature of technology that it doesn't stand still and will continue to evolve. And so you've got, you know, 30,000 different SaaS services, SaaS companies out there. Those companies aren't going to all wither and die. They're going gonna to transform. And the way they're going gonna to transform is by building agents, by looking at how they can automate their processes, how they can become smarter for the customer, how they can add more value by transforming themselves. Right. So SaaS will transform into sort of this AI revolution phase. And in so doing, we're gonna provide more value to the customers in all sorts of different shapes and sizes. And I think what that's gonna going to do is two things. It's gonna, going to, it's gonna going to add to the sprawl right? The data silo sprawl and the data fragmentation problem. And it's again going to make it more and more difficult for people to understand where is my data, right? What am I putting my data in into? Um, you know, who's got control of it? Uh and and I and and you know, one of the best examples of this is, you know, let let's take, let's take the LLMs, like let's take open OpenAI, AI, for example. Um

8:51Unidentified speaker (UNKNOWN): I

8:51Simon Taylor: can almost guarantee you that in every every company of every every listener out there, there is at least one, if not many, employees that are feeding uh information to these LLMs that they shouldn't. Right? Everybody knows if I've got a contract sitting on my desk and it's paper and it's got confidential information that it belongs in a filing cabinet. Everybody knows if I've got a contract, contra, I've got a uh uh private, you know, financial document sitting in my Excel that I probably shouldn't put that on the internet. These are kind of obvious things. But it's amazing how even a very smart individual might be sitting at home going, you know what? I don't know if I really want to write this myself. I'll just copy and paste this into the LLM Them and ask it to do it for me. Boom, all of a sudden, you know, all of your information is out there in the training models. Um and so I think, I think, and you could we could argue back and forth about whether or not that's really anonymized or it's not, but at the end of the day, it's been shared and it's been shared with an outside source. And I think that's not a good thing. Um, so I think I think the extent to which we as a society need to understand and create societal norms that will protect uh organizations Organization from AI, um, I think that's really important. I think the extent to which regulation can be used to ensure that AI is managed effectively and that there are rules, right? I think that's very important.

10:11So I think we've got a lot of work to do as a society to sort of work through this challenge and make sure that we can figure out a way to protect ourselves.

10:21Matt Brown: Absolutely. So there's a lot of lip service to AI, you know, and it there's this sort of GT. GPT, you know, paradigm where it's like, yeah, we're AI enabled, or you might be a SaaS uh company and you've just integrated with AI, you're not actually building your own models or training any model to solve a specific use case or or uh or problem like that. To what extent is Haiku um developing its own AI, for example, to uh accelerate the the capability. to detect anomalies or breaches faster than say traditional cybersecurity firms.

10:56Simon Taylor: That's great. You know, Matt, you're absolutely right. There are a lot of examples that we're seeing in the marketplace of companies that want to get out there with an AI banner ad, an AI PR announcement, and say, look, we've got AI as well. We've we built an agent, we've done this. We saw examples of some some folks who had built uh agents that literally just let you crawl their own website as though you couldn't do that yourself. Um, and and so so what we really took a studied approach uh and a slower approach in terms of understanding where AI could add value to our customers. And what we realized is that, you know, we do cover 86 of the world's SaaS applications, which is great, but there's 30,000 of them. And so we sort of decided to turn the paradigm of development. on its head. And we said instead of us trying to build backup and recovery for 30,000 SaaS vendors, why don't we let the SaaS vendors build on top of our platform? And we said, to do that, we need to make it really simple for them to actually code these modules because they're not data protection experts, nor should they be. They need an easy way to build into Haiku's RCloud R Cloud platform. So we called up our friends in Enthrame. Anthropic and we built in a partnership with them. We leveraged their their current most their latest model. It started with Cloud Two and then three and et cetera, et cetera. And we leveraged their their AI model so that we

12:16could enrich it and then create an automated AI framework so that you can now easily within two weeks build a software integration between your SaaS service and Haiku. And I will tell you that you know, it takes most data protection companies upwards of a year to build one of these kinds of integrations. So let's say you're a Salesforce, you're a 365, you know, you just sort of name it. You can literally call up Haiku, get on the website, and enter our low-code development platform. You'll be leveraging Anthropic to do it. And very, very quickly, you'll be able to spin up your own integration and add that money. module into the Haiku marketplace, providing your customers with end-to-end data protection as a service.

12:56Matt Brown: That's badass. And I imagine it's not going to take them a year to do that.

13:00Simon Taylor: That's right. We've we we you know, we like to say that we've brought it from uh uh three years to three months, and it's now three weeks. So, so we continuously scale this down. I want to get it to the point where it's three minutes. Uh my engineering team tells me that might. be a little bit much, but you know, who knows where this thing is going and how quickly we can get that done.

13:21Matt Brown: We shall see. We shall see. So obviously this is about data protection. And uh, you know, most of these models are only as good as the the data that you train them on. So what do you see as being the biggest data gap um currently within cybersecurity? in order to develop these uh, you know, these uh protection uh tools?

13:44Simon Taylor: You know, I think there's an access problem. You know, uh uh if if you think about identity access management, which has been getting a lot of press recently, you know, things like enter ID, Okta, et cetera. Um the And at the end of the day, uh we're using so many tools as a society now. and as companies and organizations, even schools, governments, et cetera, we're so dependent on so many different tools that we had to make it easier as a collective society to gain access to them. And so organizations like Okta have really figured out nice, jet nice sort of elegant ways of allowing you to access data at scale, you know, carving out profiling and figuring out you how know, you're going to assign different identities, you know, to different apps. The challenge, of course, is that you then have potential risk because if you can access that, you can access everything else. And so, so I think that, you know, it's not the it's not the the the volume of data anymore that's important.

14:44It's To making sure that you're protecting the access. And the better able you are to protect the access. to the data, the more able you're going to be to leverage the data in all sorts of new and unique, sort of AI-driven ways. I think we've seen this over again and again, where people are turning to things like Google BigQuery. And you can think about Google BigQuery as a way of sort of taking all of your data and allowing you to formulate all of the calculations that are required to sort of uh analyze. it in new and different ways. Um, and what people don't think about though is is well, are am I protecting it at the source? Am I protecting it at the target? Where is the data protected? And how can I sort of you know keep it safe? And so I think I I suspect that as AI, you know, sort of continues to grow, we're we're we're so much in its infancy right now, where you know, people are. aren't really thinking about these things. They're sort of like, you know, just give it to me. Just give it to me now. I want to get answers now. I want to build on it now. But I think down the road where this will go is people will say, well, who's got access to what? And who's allowed to enter data into the AI? And then, you know, how are we going to sort of go from there?

15:57Matt Brown: You mentioned um uh the role of uh regulation in the space. And obviously as the this tech I mean, I don't even think we're on like an exponential curve, it's kind of just going straight up like this. Um and so what that creates is uh is a landscape that's completely unregulated because that we've never been there. So how do you see uh the regulatory landscape evolving around AI and cyber security? And are uh companies ready for that kind of security? scrutiny?

16:26Simon Taylor: I think the regulatory question is such a key one. Uh, I I typically believe that regulation goes overboard in most cases. With AI, I think it's gone way underboard, um, actually. You know, we talked about this a lot at Haiku because we're always thinking about how are we going to leverage AI for ourselves, for our customers, our are partners, but always with an eye towards protecting data. Because as a data protection company, having something that's fancy and looks really cool doesn't really matter if you can't protect the data and make it recoverable. And so we developed a framework we call her. And it's humanity, exceptionalism, and resourcefulness. And, you know, we let we think about humanity when it comes to AI. You know, is what we're doing actually good for the human race? We're all human beings. Um, you know, the exceptionalism piece is we have to leverage AI to be exceptional. exceptional. Uh, we have to leverage AI in order to be better than our peer group. And then resourcefulness, you know, we need to be thinking about AI in terms of efficiency. And so I always like that model because I think it allows it allows us to ground the business in terms of what's really important. Are we getting being good people? Are we being the best in our craft? And are we actually being efficient in terms of an organization? And I would really encourage governments of the world to follow that principle as well. I think when it comes to you know crafting regulations

17:45around AI, we need to take a human-first approach. We have to. Um, yes, science is incredibly important. Yes, technology is incredibly important, and we all are capitalists that want to make a lot of money. You know, but at the end of the day, we are playing with human existence. We are playing with the future of the human race to some extent. And we're certainly playing with how our children and the and our grandchildren are going to interact with the world. And I sometimes wish that we would take a step back and be a little bit more thoughtful about that. You know, when we're discussing, you know, with global leaders and so forth around you know, the way that AI, you know, needs to be thought about. You know, there's such a rush to win. And I think I I I again I believe in American exceptionalism. I believe that we need to be the best in the world. I do. I believe that. Um, but I also believe that AI is so new and so powerful that we do need to take a minute to study it, understand it, and slow down just a little bit once in a while to ensure that we protect the human race.

18:51Matt Brown: Yeah, that's a that's a big uh question. You know, what role does, you know, I mean, if you think about it, if AI can detect and respond to threats faster than humans, what role does something like human intuition play uh in a modern cybersecurity defense?

19:08Unidentified speaker (UNKNOWN): Oh

19:08Simon Taylor: gosh, yeah. I mean, I have a 12 12-year-old year old son. I think about this all the time. You know, right? I mean, the the you're starting to hear stories about people who are saying, I'm not going to have children until the Neuralink you know is advanced enough to put the chip in their brain. And it's amazing how the, you know, the pendulum is swung. It used to be conspiracy theorists talking about, you know, technologists putting chips in the brain. Now we are putting chips in the brain and people are waiting for it. And I think I I I think that uh again, there's just so much excitement. And there, and rightly so. I mean, the power is incredible. It's a immense. You know, but you know, as Superman used to say, with great power comes great responsibility, right? So I think I think we've got to be really, really thoughtful about, you know, sort of how do we approach this. And it's no different than than what I tell people about data protection and regulation. You know, um, the Europeans who sometimes do have a tendency to overregulate for sure. Uh, they did a great job with Dora and with the NIST framework saying you know, hey, all you all this data in SaaS has still got to be yours. You got to have a local copy. And if you don't have a local copy, that's not good because it's not in your control. These are simple things. It's not a huge requirement. It's not a huge burden to have a local copy. You just back up your data. But it's a good thing. It's a healthy thing for society. It's a healthy thing for people who are counting on that data for things like hospitals, things like schools. right? Things like the

20:28military. Uh I think there are so many places where data and uh the protection of it are critical to the life is life of humans uh and are critical to modern day society. We can't forget that. And AI, if anything, is is a far more extended and advanced version of that.

20:47Matt Brown: What surprised you? I mean, I haven't spoken to you for a couple of years, but and obviously your space is evolving probably quicker than most, I would suggest, given how important cybersecurity is to um to most companies. Um so what has surprised you the most uh or has there been an unintended thing that you've seen uh AI detect in the cybersecurity context?

21:13Unidentified speaker (UNKNOWN): So

21:13Simon Taylor: so it's it's not something so much the detection for me that's a surprise. I think that, you know, these things build on each other. What really surprised me was the speed mat. Um I gave a talk at the United Nations about five years ago um uh on the subject of AI. I was a guest speaker along with a bunch of far, far more intelligent people than me. And uh and we had a a really nice talk there and we discussed sort of the future of AI and these things. And I'll tell you, at the time, maybe it was six years ago, but it was a very theoretical event. I mean, it was very theoretical. It was when this happens, then this will happen. If this happens, then this will happen. Um and of course, w everybody's saying, yes, it's going to happen. Uh but I think what stunned me was,

22:00Unidentified speaker (UNKNOWN): you

22:01Simon Taylor: know, if I if I look at even how we run our business today versus how we ran it a year ago, it's almost completely different. When I think about, you know, how I interact with our with our investors and our board, it's completely different. When I think about the decisions and the trade-offs we're making as a business, it's completely different. All as a result of AI. You know, it's entered the f it's entered not just the lexicon, it's entered the way that we function as a society. at such a deep level. Um it's almost moved beyond technology to into into into uh a key part of our life. Um and and I think I think I always thought that that would happen. I don't think that I realized or recognized quite how fast that adoption curve would be. You know, it took people from I don't know what, two 2000 thousand and

22:49Unidentified speaker (UNKNOWN): uh

22:50Simon Taylor: maybe 2005 or 2006 to 2020 to get to cloud, you know, right really at scale. And it took people sort of like three months to all change the way they live and use chat GPT every day. And and and so so I I I just I I think it's it's incredible um the herd effect of this. And I think it's incredible the speed with which a AI. is just building on itself and it's not going to stop. It's not going to slow down. If anything, it's going to get faster. Um, and we just need to be ready for that. We need to think through that. And again, I do think that, you know, not putting on the brakes, but just slowing down a little bit uh to to sort of take stock of what we want this world to look like and then applying a little bit of light regulation around the sides would be a very, very healthy endeavor.

23:38Matt Brown: Yeah. couldn't agree with you more. I think when, you know, when things are moving at such a pace, um, it's easy to get overwhelmed. I mean, even I mean, you know, uh, we're in the AI space uh for with the number of value engines and what have you, but even just, you know, like if you think about things like AI voice models, you know, um, and six months ago we wouldn't even be talking about voice agents, right?

24:02Simon Taylor: No.

24:02Matt Brown: Um and now those voice models, you know, eleven labs, uh, even open AIs isn't that bad. But these things don't suck anymore.

24:11Simon Taylor: Yeah.

24:11Matt Brown: So the I mean, just in that one space alone, there's been an explosion of startups who are now offering voice like I mean, it's just it's ridiculous.

24:22Unidentified speaker (UNKNOWN): And

24:22Matt Brown: that's just such a narrow that's just like one use case as an example. And then when you extend the impact of AI on all industries, like I talked to a lot uh to a lot of law firms for instance. And uh as an example, if you have I spoke to this one fifty year old law firm

24:38Unidentified speaker (UNKNOWN): um

24:39Matt Brown: and I said, you know, how many associates do you guys have? And they're like fifty. So I said, Okay, so when you have a medical malpractice suit, like, you know, how many days do they spend on average doing, you know, research or discovery?

24:50Simon Taylor: It's

24:50Matt Brown: like you know, uh probably three days. So I'm like, great. So that's three days times fifty associates. That's all billable time. So then but now, like we were talking earlier on about Agencyc AI, now if you can create uh a s a a workflow as an example that can enable all those fifty associates to do all that discovery in eight minutes, how then does you how does a law firm recoup the billable time of those 50 associates? So it's not just about like adoption for the sake of adopting. One actually has to think about the business model impacts of that adoption itself.

25:24Simon Taylor: I'll do you one better, which is I really am concerned about the concept of apprenticeship and the loss of sort of uh enablement that happens with on-the-job training. You know, I think I think there's a there's a real run right now. at sort of um uh removing or nullifying the value of you know junior employees. Let's say you're an engineer and you've got you're one year out of school, right? Um, well, those were critical folks for a very long time, for a very, very long time, since the dawn of coding. Uh you come out of school, you learn, you have an apprentice sort of an apprenticeship model, you join a scrum team, you sort of move up the ranks. You maybe you started as a tester or QA and you sort of worked your way up. Um, and you're learning, you're learning, you're learning on the job. Well, now that we're starting to automate all of those jobs, how do you start? How do you get started? How do you get the wisdom and the knowledge that comes with being on the job for 30 years? You can't throw someone at the deep end of the pool. right? And so all of a sudden, you, you know, prompt engineering starts to displace uh, you know, you know, actual coding. Um, and you know, there's an argument to be made, sort of that, you know, machine learning isn't what coding is today, right? Uh, you know, punch cards, you know, my father was a

26:43punch card programmer in the 70s. Um, so, so I mean, I I of course these things do evolve. Of course. they do evolve. But I think that there is going to be a skill gap. You know, and I think that the sort of new entrance into the career, until we figure this out, are really going to be in trouble to some degree in a lot of these different jobs. And you know, you use law as a great example of that. What do associates do, right? They fill out the the the re repeatable templates that they all the customers, all the clients get. Um, well, now that can be done with the touch of a button. Am I gonna hire 10 associates from Harvard and pay them hundreds of thousands of dollars a year and give them expense accounts so they can go use Chat GPT? Probably not. Um and and so again, where does the wisdom come from? Where does the knowledge? And I think these are the unaddressed issues that we need to really be thinking about as a so s as a society.

27:37Matt Brown: Who's leading the the charge on that then? Um this is why I say it's this there's this uh there's this massive gap, right? In terms of education and that. I mean, my son's ten and I encourage I mean, I tell bedtime stories to Franklin using OpenAI.

27:53Simon Taylor: Yeah. Yeah. Yeah.

27:55Matt Brown: Expose him to the technology because uh I think one is obviously exposure to the technology, and then obviously as they get older, it's about well, how do you use this? Practically. And so for instance, it's not necess it's not only just about the prompt uh engineering itself, it's about looking at the output and then observing objectively what is this you know output, is it actually what is correct or not? And it hallucinates. And I think you know, just with that one thing alone, um, you know, I think that it's a big opportunity for from an education perspective, right? I mean, the question I actually want to ask you is, you know, in this new context that we're all moving into, this AI uh enabled economy, what sort of leadership skills does a CEO now cultivate and develop? You know, is it about creativity? Is it about empathy? Like what is it now? Because if you're going to build a company to scale, especially if you know, again, recognizing that I think today there's going to be more, or certainly in the next three to five years, there's going to be more valuable companies on the market, but with smaller teams. And so I'm a great believer in

29:04constraints. I I my my whole thing is, you know, if I can't have my whole team around my house for lunch, there's is too many people in the business. But then I can think about AI systems. and workflows and how to automate things so that I can make that a reality, right? And so as

29:20Unidentified speaker (UNKNOWN): a

29:20Matt Brown: CEO, um then you have to say ask yourself, well, what what kind of culture can do we now build? You know,

29:26Simon Taylor: right.

29:27Matt Brown: Um, and what skills do I need to develop or other CEOs or founders need to develop in order to be effective in this new AI economy?

29:34Simon Taylor: So that's the model we talked about earlier, right? That's the her model. Humanity, exceptionalism, resource. resourcefulness. Think about this. Um, where did that model come from at haiku? It came from this idea that we do need to evolve our thinking. We need to evolve our leadership style. And, you know, in the past, I always talked about authenticity, grit, and empathy. I still think those are very important. In fact, you know, you mentioned empathy just now, and I think that empathy uh is is is somehow uh negotiated. negotiable in today's world. And it's really not. You know, but when you hear people say, well, they're so tough and they're so this and you know, they have no empathy at all, and and this is the best way to run a business. Look at Elon Musk and how he's built these companies and so on and so forth. At the end of the day, humanity is the most important thing. You know, um, I think that people have started to realize in a post-COVID post COVID world, this sort of new generation. they want different things. They want a really good experience from their founders, from their, from their, from their businesses. They want to feel like they're showing up with purpose. Um, and and you know, so for me, empathy and humanity are critical. Uh and

30:40I think that the more that AI bleeds into things, the more that you have to make tough decisions, the more that you have to say, well, you know, um, I'm gonna use AI for this. I'm not gonna hire for this job. I I'm gonna use AI instead, right? These are very, very hard decisions. They're decisions CEOs have to make. Okay, don't get me wrong. We have to make those decisions, but we have to do it with humanity. We need to understand the impact it's gonna have on the business and on the human beings that work there. Um, the second piece, exceptionalism. You have to embrace it, right? So every CEO today is embracing AI. This is nothing new. I think that, you know, setting real metrics and creating a real framework for what you intend to do with it is really critical. You know, so at Haiku, you know, we sort of have a team looking at this. We've sat down and we've we've laid out, you know, five KPIs that unfortunately I will not share uh on the podcast, but five KPIs, confidential. KPIs that we are we are driving the business towards in order to ensure that we are exceptional. And we are we are the most exceptional company in the data protection space. And so I think building a framework and a model and driving your business to that that's focused on AI, just the way you would with your corporate goals, uh, is now part of the game. It's

32:00what you have to do to be successful. And then I think the third one. resourcefulness, right? So when you look at AI, you need to look at it as a way to get more from your team. You need to look at it as a way to uh be able to take away the time, money, and resources dynamics so that you can almost have infinite capacity to do anything. And I think when you approach it with that lens, uh the ability to resource compl com changes drastically. And I think so so I think, you know, keeping those three constructs of humanity, exceptionalism, and resourcefulness in your mind as you're going through every day, as you're going through every meeting, as you're building your models, as you're, you know, working through your your plan or your AOP, um, I think that's what really matters today to a CEO who's sort of grappling with this.

32:48Matt Brown: Yeah. And I think for me, I couldn't agree with you more. I love that. uh framework. You're gonna have to tell me off air about what goes into this. But um but I think uh also from my side I think it raises a lot of ethical considerations. Like if you're gonna adopt AI, deploy AI to a thousand customers, four thousand plus customers as an example, you know, what role do you do things like transparency play, you know, in the deployment of AI? Because yes, you have that acceleration of organizational capacity, but your customers might not like that. And here's a true story. I was chatting to uh an a best selling author of a book who wrote about uh the whole role of AI and custom experience. And they did this uh test where they had a product name, uh like whatever it was. Um so they tested this against uh you know a focus group. and all they did was with the second uh label was they said powered by AI and they had a sixty percent drop off in engagement simply because of the association of a product with AI.

33:51Unidentified speaker (UNKNOWN): Do

33:51Matt Brown: you see what I mean?

33:52Simon Taylor: Yeah

33:52Matt Brown: and and it was fascinating right because then this then the question becomes well do you how transparent do you need to be? I mean if you think about VO3 as an example I mean

34:00Unidentified speaker (UNKNOWN): that's

34:01Matt Brown: ridiculous. I mean I can put you on stage as an example.

34:04Simon Taylor: Yeah.

34:05Matt Brown: And you can be like, Well, we we really really struggled with whatever. And Matt Brown from Showworks called me and now it changed my life. You know what I mean? And it's completely fake. So at at the end of every piece of content, I mean, I'm not saying we would ever do

34:17Unidentified speaker (UNKNOWN): that,

34:17Simon Taylor: but

34:18Unidentified speaker (UNKNOWN): do

34:18Matt Brown: you know do you put a disclaimer there now that says, Well, this was created with AI? Because, you know, customers sometimes are not necessarily in B2B, but certainly in B2C. they can't tell the difference. Right. So you can

34:31Unidentified speaker (UNKNOWN): also

34:31Matt Brown: create your entire marketing function, right? With V3 putting out plethora of content and it's all fake. It's not it's not real.

34:40Simon Taylor: You know, I I think this is this is part of

34:43Unidentified speaker (UNKNOWN): this

34:43Simon Taylor: is part of the nuance around this being so new. Um you know I I I find this with my own team, you know we have we have some uh private, you know, LLMs that we we leverage, you know, here at Haiku. And if I generate something via that and I say to my team, hey, read this, okay, this is this is a model I think we should think about. And they know it comes from it comes from the GPT, it comes from the LLM, immediately it's sort of dismissed, right? Whereas if I send them a memo and I say or an email and I say, hey guys, I want to do the following 10 things. I'm going to get immediate responses and these kinds of things. So you see this even on the management layer, you see this in the company. And I think it's because people's early experiences with with the with Chat GPT or OpenAI Opening or whatever they're using, um, were uh that of you know talking to an 18 year old, right? Uh talking to somebody who's fresh out of high school. And, you know, people who tried it a little bit later on, it was sort of like a college, you know, college student. And um what they don't realize is it's become so good. It's now like talking to a graduate level, even a doctoral, you know, uh student. Uh it's really changed

35:57and it's continuing to evolve. I think that what will happen is the new the next generation that's come up with it. AI are gonna be far less compassionate or concerned about whether it was written it was written by this one or this one. I think they're gonna they're gonna expect that all content um is delivered fast and furiously from thousands of different sources. And I think it's gonna become less less of an issue of people not getting the eyeballs and more of an issue of ethics. Uh and integrity. Because I think, I think, you know, you know, the the the way that we're building even the LLMs today, it's going so fast and so furiously that people are grappling with this issue of ethics and integrity and trying to figure out, you know, what's what what is legal to put into the LLM for training, what is not. Um and and you know, I think we're we've we saw the actors strike a couple of years ago where people saying, I don't want my image, you know, used by AI. Well, the problem now

36:57is that AI is creating its own pretty realistic videos that look pretty real. Um, and, you know, all of a sudden, you know, we'll be in a world where maybe our kids, their favorite actress or or actor won't be Brad Pitt. It'll be some AI generated model. Um, and it won't really matter. And so so I think I think I think that, you know, again, it comes down to understand taking it taking it not a break, not a not a step on the brake, just slowing down the car a little bit, you know, figuring out what regulation will allow us to study this and truly understand the impact so we can make the right decisions for our children and the right decisions for how we are actually interacting with the world.

37:35Matt Brown: Yeah. There's a lot of that uh avatar stuff happening on uh only fans.

37:40Simon Taylor: Yeah. My God, you caught me off

37:43Unidentified speaker (UNKNOWN): screen

37:44Simon Taylor: there. Thanks, Matt. Oh,

37:47Matt Brown: anyway, but let's wrap this up, Simon. One uh fun question for you, um, and uh we'll let you go. If you had to predict like one surprising way that AI is going to influence everyday life in the future, not necessarily cybersecurity. and B2B or what have you. What would you say that is?

38:09Simon Taylor: I think that this is going to sound very out of out of left field, but I think the concept of elder care is going to be entirely different. You know, I think that we've never done a good job of taking care of the elderly or the infirm. And I think that AI, if done correctly, and with integrity and with high morals, provides an opportunity to decrease loneliness in people at just such a great rate. I think you know, you know, elder the elderly community in particular suffer massively from loneliness. There's an epidemic. And I think that I was talking to somebody the other day who runs a whole bunch of elder care homes across the world, really large company. So it's the biggest challenge they have is how to peop make people feel less lonely in their sort of, you know, uh their the end of their life. And I think that that that's a place where, you know, AI could be really helpful. I think there's some other, probably far more damaging ways that

39:11they that AI could, you know, uh hurt relationships, you know, and other in other area spectrums of life. But I think that the elder care piece could be really, really interesting.

39:20Matt Brown: Hm. It's funny because I actually spoke to a franchise owner of Elder K of Elder Kid network uh this morning and he's just seen he's do he's seen explosive growth here just as a franchise and let alone the the whole like group itself and and 'cause because

39:35Unidentified speaker (UNKNOWN): of the

39:35Matt Brown: elder ultimate care epidemic as you as you describe it. So I mean I do want wonder whether the Optimus Robot's gonna be helpful.

39:45Simon Taylor: We'll have to see. We'll have

39:46Unidentified speaker (UNKNOWN): to

39:47Matt Brown: see. Well Simon, so great to chat to you again, bud. And uh yeah, obviously uh congrats again on all your continued success. I know you had this

39:55Unidentified speaker (UNKNOWN): big

39:55Matt Brown: dull announcement. Um so that's a big one. You guys just seem to be taking over the world. So obviously wishing you and all the team there at Haiku all the very best.

40:02Unidentified speaker (UNKNOWN): Matt

40:02Simon Taylor: you're the best Great to see you.

40:04Matt Brown: Likewise. Everybody else, we'll see you again soon.

This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.
All insights