Mandy Andress (CISO Elastic)

We sit down with Mandy Andres, CISO of Elastic, to map how AI is reshaping security programs and why business fluency matters as much as technical depth. We dig into visibility, identity sprawl, agent behavior, and what CISOs can do right now to manage risk at machine speed. 
• Mandy’s path from accounting and auditing to security leadership 
• Elastic’s security scope across enterprise, product security, and SaaS infrastructure 
• Why visibility and asset management stay the hardest foundational problem 
• Shadow SaaS turning into shadow AI and the return to DNS, endpoints, and data lineage 
• Non-human identities, API keys, and lifecycle controls as AI agents multiply 
• Data protection choices for LLM usage using endpoints, proxies, and guardrails 
• Applying AI to the SOC first and then expanding into vulnerability management and GRC 
• Balancing faster AI adoption with risk trade-offs and business objectives 
• Least privilege becoming least agency with agents and unpredictable action paths 
• Automated penetration testing and continuous validation without humans in the loop 
• Threat themes ahead including credential exposure and shifting exploit patterns 
• Startup advising, conflicts of interest, and staying transparent 
• Practical advice for CISOs on ambiguity, risk decisions, and continuous learning 
• Talent growth through reskilling, automation, and focusing humans on judgment 

Transcript

Welcome And Guest Introduction

SPEAKER_00 0:02

Welcome to the Unicorn CISO Podcast. This is Pedro from 33N. Let's get started with today's episode. Okay, so welcome everyone to another episode of the Unicorn CISO podcast. Today we're uh having the opportunity to discuss with Mandy Andres, CISO of Elastic, and I'm sure it will be a great discussion. Hello, Mandy. Thanks for taking the time.

SPEAKER_01 0:25

Hello, thanks for having me.

SPEAKER_00 0:28

Okay.

From Accounting To CISO

SPEAKER_00 0:29

So we we wanted to start the discussion with uh an overview of your career to date. I mean, you've been three decades in in security, uh, and you're especially coming from a non-traditional background, let's put it that way. What has shaped the way you think about the Caesar world today with with some so long in the in the industry and also with this unusual background?

SPEAKER_01 0:51

Yeah, I started my career, my degrees in accounting, CPA started in auditing, moved into systems auditing, and always had a love of tech. And so found that to be a good combination of uh my interest in business and my interest in technology, and got into it right as the internet was really launching and taking off, and technology and online technology became more and more critical to businesses. And I was able to get in fairly early in that process and grow and expand my knowledge and experience as you know, the internet and security needs were expanding with businesses. And so what I find with the the business part of the background, especially say in the last decade, when security challenges became much more of an existential business risk and became a top enterprise risk, being able to go into conversations, not with just the technology perspective, but also with the business perspective and being able to speak that language and look at security challenges from kind of a business impact business operations perspective. And then along the way, I threw a law degree in there with the expansion regulatory frameworks and requirements. So really then coming at it from looking at security implementations from the perspective of regulations and control frameworks, and how do we communicate that, whether in contracts or to auditors? And so really bringing a broad perspective and able to really use the language of different stakeholders when talking about security. And it's been a really fun ride.

SPEAKER_00 2:47

Okay.

Elastic’s Security Scope And AI

SPEAKER_00 2:48

What's your today in what's your scope in in Elastic? And it's both a company that has a forward-looking uh front. And I'm sure with Elastic having such a relevance in as a as a tech and security vendor, it it also has some some implications for the outside, right? Does that does that cover this is the scope you cover?

SPEAKER_01 3:10

Yes, at Elastic, or SCCO, lead our information security program, and have full accountability for all information security, cybersecurity risks within the organization. So we cover Elastic as an enterprise. Product security from the perspective of secure software development and all of the processes and components that go with that. And then lastly, we sell SAS services, so all the infrastructure and components for securing that, securing where our customers share their data with us. Elastic overall, you know, Elastic to Search AI company. If we continue to look at the reliance that we have more and more and more on data and understanding what that data is and can tell us, and using AI to help leverage that Elastic has always been a core component of a back-end infrastructure for many, many organizations. And then as we move into deeper AI and security in that combination, security has always been really a search problem if you look at it. It's finding patterns, it's finding activities amongst large amounts of data, and that amount of data just continues to expand. And so Elastic kind of applying the expertise and specialist experience within search, starting to apply that into AI, uh into security with AI, allowing us to create more natural language interactions and chat interactions and deeper analytics that allow analysts to then focus on business impacts of moving system to system, whether it's to find context or understand first what their environment looks like, and being able to pull all of that information in automatically and really moving into that SOC of the future and Egypt SOC as we're talking about today.

Visibility Gaps In Cloud And SaaS

SPEAKER_00 5:24

You often say, I mean, you can secure what you can see, right? And I think visibility and given the sprawl of AI, SaaS, I mean, it is for sure a topic of the day. Where do you see that in your experience or speaking with peers, visibility is still lacking?

SPEAKER_01 5:44

Well, we we have talked for years that understanding what's in your environment, being able to see what's happening with what's your environment. If we look at top 10 control lists, asset management understanding is always number one. And we say that it's very hard to achieve. So one of the things that the migration to the cloud provided was a much easier way to get visibility into the assets, so everything being API-based created larger sprawl, more complexity within environments. So while easier to get data, much harder to understand how all of those components are fitting together. But it also expanded dramatically the amount of data available to us. So now applying AI into that, it allows you the ability to better understand or better analyze the data that you have to see where you have gaps, to see where you might have actions, say whether it's network activity, but not necessarily seeing the related assets in your inventory. So it allows you to find those blind spots much, much easier. But we're also from, you know, we had shadow SaaS, now of course we have shadow AI. And so it's also finding how do you get visibility into what is happening in your environment and in your organization from an AI perspective. And that's you know, getting back to some fundamentals, looking at DNS, looking at what's on, what's running on endpoints of users and workloads. It's going back to a lot of data lineage and understanding where data is going. It's moving potentially to many different places much faster. So still, you know, for me, a very foundational component of the security program and something that we continue to, as an industry, focus on and as practitioners need to ensure that we're we're doing the best that we can, but also continue to make our jobs more challenging, adding more complexity and more technology, uh, just as we're we're moving into the AI world these days.

Non-Human Identities And AI DLP

SPEAKER_00 8:01

It's it's kind of specifically in in securing AI or or next gen. Some people refer to it as next gen, DLP, the SPM. It's definitely a challenge when you have a new attack surface, which is the whole EI stack, but also you have AI being infused in SAS and in any other desktop applications that that employees might be running. And what's your view on having different vendors, having consolidated view? I mean, what's what has been solutions that uh that you found are interesting to cover holistically this problem?

SPEAKER_01 8:35

There's a lot of different solutions available, and effectiveness really depends on each organization, how they operate, what they see the risks being. I think for me, a couple of key areas that AI has continued to increase in importance, first non-human identities. Certainly a lot of conversations the last few years with threat actors, just leveraging exposed credentials, API keys to log into environments. AI is only creating a much larger impact of that. We're creating many more identities to support agents. And so solutions that are helping identify, manage life cycle of non-human identities is an area of solutions that is something that's is critical from my perspective. And now mentioned DLP and data protection. So getting into a new area is just how do we want to do that in the AI world, whether you're talking about data going into LLMs and you have an endpoint approach, you have a proxy approach, several different approaches and they they work well in different areas and in different organizational types. So certainly something that helps you understand where your data is going and to be able to provide guardrails and controls to help manage that you are able to implement whatever security controls, data protection controls your organization needs. And then certainly just applying AI into security processes. So a lot of kind of first area of application in security is a lot on the SOC, the security operations center. That's because it's high volume, highly analytical role that the humans are doing largely very manual tasks and aren't able to do the deep analysis, just the way human brains operate. And AI is able to really expand and improve on those processes. So it made sense. It's the first application for AI. So a lot of the solutions that are looking at how to create an agentec SOC, how to apply AI into SOC, and then starting to see it from a security security program perspective expand into whether it's vulnerability management, GRC, starting to see some initial plays there. So it'll be it'll touch every aspect of an organization. Security teams, security programs, security processes will look very different from my perspective over the next few years. And that's all driven by AI and how we how we find the best and the most opportunistic ways to apply it in our organizations.

Balancing AI Speed With Risk

SPEAKER_00 11:27

Well, one that touches an interesting point, which is the speed of AI adoption versus really re-re-architecting processes, really accounting for risk and making sure that the productivity gains are not compromising on the risk that you might incur with with both typical hallucination as well as just expanding the tech surface of your environment. How do you balance this? And and you've seen uh the the hype go up and now and now coming a bit more to the real balance of accounting also for risk and and and re-architecting of processes. How how did you see this from your perspective?

SPEAKER_01 12:10

Yeah, it's been interesting to track for the last several years of just the go back to when ChatGPT entered the zeitgeist and conversations on what data are we putting in there? Do we allow it? Do we block it to all right? We're allowing use of LLS largely, then moving into agents, and agents are coming, and now agents are here. So every year, now every six months, and it's just getting faster. Evolution and new technology, new application of AI technology. So I only see that velocity and that trend continuing. And then how do you take that and apply it within your organization and being able to understand and look at, all right, how can I best take advantage of that? And where that balance comes in is first really understanding your organization's business objectives. What is the company trying to achieve to hit what they define as business success? And then looking at how AI can or where AI can or cannot help, both within business processes and within security processes, and making a very risk-based analysis, but also a general trend that some organizations need to realistically take on more risk, at least in the short term. If you look at things like methos and models being able to provide deeper vulnerability analysis and find vulnerabilities in code and talking about needing to really compress patch cycles and things like that, you know, that creates risks on the other side. You have patch cycles too, and you allow for testing to minimize downtime and impact on business operations. And so it's continuing to really understand what those trade-offs need to be now in an AI world and balancing how how fast do you need to move within your organization with AI versus what potential impacts can you absorb or do you should you absorb? And those conversations are now happening, whether it's at the board level, senior leadership level, and only are expanding as we see the AI technology expanding. And so I you know, we've talked for several years about security now as an enterprise risk. It's a top risk in most organizations. And AI has only compounded that. And so get the combination of kind of AI use AI technology, and the need for that just for business to be successful, is also a key component of a security program to allow uh the evolution of an organization to both uh take advantage of AI, but also use it to enable their organization to move faster.

SPEAKER_00 15:15

I guess in code security, it is even more obvious or top of mind and the the trade-off between risk and and and productivity. Were there any hard trade-offs that you from from the InfoSec organization were were having in advising or or balancing the speed of software development with with security guardrails and controls?

Least Privilege For AI Agents

SPEAKER_01 15:42

The challenge has been not so much directly on supporting development, but supporting all of the connections, interrelationships, automation that where you want to take your AI tools and connect them to all sorts of business systems and business processes. And so that's where just not always understanding or anticipating what approaches AI and especially agents may take. So we're you know, we're looking at our threat models and things from a kind of human perspective, and we don't always look maybe as literal as sometimes an agent will look at a prompt and follow the objectives that are defined by it. And I think that's where we will see challenges and have seen challenges of just being able to make sure that because we can't always anticipate the approach an agent will take, that we really need to get back to least privilege, least agency with new terminology to talk about agents. And that's always been hard, and we haven't done it well in the past, and we need to figure out quickly how to do it well with the in the agent world because you know agents uh make interesting decisions. And one example of just what we've seen is have agents in the environment that have their credentials. If that access key doesn't work, then it just looks scans the entire vault to see if there's any credentials that they could use that make sense. And that wasn't necessarily anticipated behavior going in. So it just demonstrates how while you may not anticipate everything, you need to be uh prepared and make sure that you're putting any barberails to minimize your tag minimize lost radius.

Automated Pen Testing Gets Real

SPEAKER_00 17:45

Indeed, in in particular, there's there's also a big a big concern with how you're you're taking open source, right, and in software, right? And and and uh this is part of the the problem. Again, to your point of having the agent, agentic perspective or the perspective of an agent, there's a lot of tools coming around of let us again use an a pen testing approach or a genetic pen automated pen testing approach to to testing some of these things, right? I'm sure many organizations, the large vendors, the large uh financial companies first are having access to the latest models, but I'm sure there's a wave of others catching with almost as good kind of models and hopefully improving in the future. And I'm sure many companies will will startups, vendors will be built around that, right? I'm I'm better in in testing for your vulnerability. So have you have you seen this? Is it uh something you'd uh either as a business engine, right? Because you also have are quite active in and uh in startup ecosystem. So you've seen this emerging.

SPEAKER_01 18:56

Definitely seen it emerging, and it's leading to a lot of conversations kind of your your traditional buy versus build conversation that we've had for decades, and that's really come back to the forefront with the power and capabilities that AI and LLMs are bringing much more in build over buy at the moment, uh, and that from creating some challenges for startups. I don't foresee that being long term. I think while it's easy to build things right now, it is it has been and continues to be hard to maintain and scale. Uh, and I think as organizations learn that and find that resources, their limited resources may not be best served in trying to scale and support some tools that they've built in-house, going back to buy over over build. And that's where the expertise and a lot of the security startups that are hitting the market today are really finding their niche, finding their focus on how to help organizations either effectively or securely leverage AI within the organizations. And there's a lot of unique approaches, a lot of very interesting approaches. And I just continue to see that evolving over the next few years.

SPEAKER_00 20:33

But generally, this this topic of automated pen testing or genetic pen testing is one where you've seen uh that there's there's space, there's there's a real need.

SPEAKER_01 20:44

There's definitely space and a need. And what I've seen, the models that were released in late 2025 really upped the capabilities. For me, if I'm looking at genetic pen testing, what I in a teaming, what I define as effective solutions in that space is something that is able to, once it understands your environment or understands where where to start, it's then self-identifying, it's chaining, whether it's different vulnerabilities or access, and doing all of that in a highly automated fashion and where there's really no human in the loop. And starting to see some of those really evolving into pretty mature capabilities just over the last few months. What that allows, both it's certainly cost effective from an organizational perspective, but allowing you to have that continuous testing, that continuous understanding of potential vulnerabilities within your environment as the the threat landscape continues to change more and more quickly. Doing even if it's weekly tests against your environment these days, that's not sufficient. And being able to just have through these automated pen testing and routine lean solutions, just have that happening all the time and having that continuously updated as your environment changes and evolves and as the threat landscape changes and above all of that happening and being integrated into all of that work automatically is a huge benefit to security teams.

SPEAKER_00 22:29

Great to hear good. Sometimes speaking with with the offensive teams and also especially after the hype and then the the the moment after when when uh many people were were coming around and saying no, this is all of uh marketing and and stunt and and a lot of hype around, but in in in all truth, I mean it's still uh yet lagging. So it's it's good to see that for sure there's it's not a hundred percent sure that we'll we'll we'll take the human out of the loop, but but for sure there's there's big big benefits, right? Not only on costs. So so good to hear.

Threat Trends For The Next Year

SPEAKER_00 23:03

I'm I'm curious on other topics, other emerging topics. Again, as an organization, Elastic is super uh forthcoming and and have issuing also their your global threat report. And and so I'm I'm always curious. I mean, in the next one to one and a half years, what what are you seeing? What are things that you've been discussing more and more topics that are becoming more interesting?

SPEAKER_01 23:26

It is a challenge to talk 12, 18 months, sometimes even six months out. Uh so that that that would be number one, just velocity and evolution is happening so quickly that while I can't necessarily anticipate specifics. I can anticipate general themes. And first and foremost, just going back to the fundamentals and foundations of security. You know, in our last global threat report, credential exposure and just predator logging in as was the top dexist method. And I see that continuing, just the ability to apply AI and machine speed to define those credentials and just try them very broadly. So I see that continuing. The the vulnerability, there's been a trend of less use of exploit code over the past three years. I anticipate that that will increase, just as we've seen, increases in the number of vulnerabilities, probably found out by CBEs or zero days. I think that will exist over the next 12 to 18 months. And I think velocity in general, they mentioned before, will be a huge, both from a threat actor and also from how defenders need to adjust their approaches and their posture and their hospital space. And exactly how that plays out over the next 12 to 18 months. I have a feeling next month we'll be talking about something very different than what we're talking about this month, and that's what keeps the area exciting right now.

Startups Ethics And Conflicts

SPEAKER_00 25:10

Totally understand. You're you're very active as as a CISO telastic and very visible. I mean, participating in in leading edge conferences and and and supporting also big organizations, I'm sure. But then you also, as a business angel, I mean you also keep uh advising, supporting companies, and and having a balance here of what is mature and what is kind of uh innovation and and and coming in in the next few months or years. What what are synergies that you have and in in your activity with startups and and what is I mean misconceptions, right? That uh that you often see people having.

SPEAKER_01 25:52

I think synergy is it's a a lot of you know, really get into the space and trying to understand what what are the leading edge approaches in security company. You mentioned Elastic, we are on the forefront of some of that. And oftentimes the challenges that we're facing, there aren't existing solutions to help us solve that. So having access into the security startup ecosystem, just understand what ideas are out there, what different approaches are being developed, is how it started. And having those conversations uh for myself personally are just very engaging and very exciting. It keeps me interested in in the space overall. And I'd say areas that the folks don't necessarily anticipate all the time. I think for me, just reputation, conflict of interest is top of mind, and so being very mindful about how you approach that. So certainly with a last having some participation in the security space, networking with organizations, uh, conflicts, conflicts with other organizations that I may be advising or investing, just being very upfront and transparent about all of that. Also, just everything is is out in the open. And I think some folks don't always take that as as seriously as they they may need to, and just really separating the like the clear hats that I wear, being what we sell, and then being an advisor, looking at the future of the of the industry overall.

SPEAKER_00 27:31

I mean, elastic as a company has also been coming closer to to the security side, right, over the years and and and also AI helping, uh blurring the boundaries. So have have you ever been in a position, for example, in which you were advising a company, and then later on you saw, hey, listen, this is potentially too close to the business of Elastic, for example.

SPEAKER_01 27:54

Yes, absolutely. There's been a few that I started as I started devising and then evolution, either evolution on the startup side or on Elastic side, things merged a little too close, and I abandoned those those relationships because of that. And it's something I'll keep a very close eye on.

SPEAKER_00 28:15

Okay.

CISO Advice And Team Scaling

SPEAKER_00 28:16

One piece of advice to to CISOs that are joining. I mean, fast-growing tech companies, the trade-off we always discuss in this podcast is you're you're balancing being on the front for cyber with business speed. And and I'm sure listeners that might be in in tech growing companies, uh, hopefully fast growing companies, are having this balance. And you've been you've been in this position for a long time, right? So, what are pieces of advice experience that you would change you'd share with listeners?

SPEAKER_01 28:50

These days for folks becoming a CISO being comfortable with ambiguity. We've talked about that for a long time, but it's uh even more critical today because we don't know what's going to happen tomorrow. We need to make decisions based on the information that we know now. We need to make risks decisions. We need to be able to both articulate risk but also be comfortable accepting more risk if that is what the business and leadership want to do. So it's looking at your role as a risk manager and advisor, but not always from what, but being comfortable to take a stand on things that are potentially extremely critical or existential from an organization perspective. And those are hard conversations to have. So I think that's one area you see so it's coming in, just make sure you're comfortable with that and focusing within those specific areas and understanding how your business makes money and how that then flows from the intertechnology perspective and how security impacts that. So going back to that business relationship and business understanding that we talked about at the beginning of our discussion. And then second, I would focus on you know, security has always been a space where continuous learning is needed to stay current and even just to stay effective. And that's that's required as even more critical today in the speed with which you need to learn new things and you know, the round and times, and just finding ways to where you can you get your information where you're comfortable just kind of understanding what's happening, whether that's through your networks, whether that's through podcasts like this and um community groups and um articles and things like that. So just finding ways to stay current as best you can so you're stand ready to able to provide that advisory to your leadership and your organization as questions come up as they are doing every day in this in today's world.

SPEAKER_00 30:59

Very clear. One one topic that also might also arise here on related is is talent management, right? Then there's a lot of automation coming out, but the the threat landscape is also increasing, right? So, how how do you manage team growth? How do you manage team learning and and the development of the definitely an interesting area right now?

SPEAKER_01 31:23

So it's looking at one, you need to make from a talent perspective and a skill perspective the transition into how do so having them learn AI technologies, how to use them in their day-to-day, but also the impact that whether threat actors or having it in other processes, the volume or velocity of things that a security team needs to respond to is increasing often exponentially, and looking at how can we scale our program, and that's where adding people usually isn't the the answer to scale at that pace and at that level? It's uh it's having folks on the team being reskilled or learning, right? How do I look at what I'm doing today? Realize that the volume is increasing 10x or more, and has already increased 10x or more. How can I challenge our approach and look at it from a different perspective? How can I use whether it's automation or AI specifically to help in this area? Whether that's a lot of times right now, it's a lot of focus on using AI to help do some initial triage or bring in context and kind of all of those looking at anything that folks are doing manually today, how can that be in some way automated, being then able to apply human resources where that critical thinking, sign blocks, analysis and decision making and judgment need to be really reapplied. And that's the changes the dynamics of how a security team operates, how it's structured. So going to be a significant evolution in how into and how security teams operate over the next few years, and just to over the very beginning of that. And that's uh going to be some some fun times for us all.

SPEAKER_00 33:25

Mandy, thanks once again. Very useful advice and insights. Really appreciate you spending the time with us. Hope we we can speak soon. Thanks once again.

SPEAKER_01 33:35

Thank you so much.