How is Agentic AI redefining cybersecurity leadership? Watch the recording of Akitra’s July 2025 Interactive CISO Roundtable to discover how top security leaders are adapting in the era of autonomous intelligence.
In this thought-provoking session, distinguished CISOs and industry experts dive deep into how Agentic AI is transforming the traditional CISO role—from operational oversight to strategic orchestration. Learn how to lead effectively, strike the right balance between automation and human judgement, and stay ahead of emerging AI-driven risks.
Highlights include:
- The impact of Agentic AI on the security landscape
- How the CISO’s responsibilities and leadership are evolving
- Achieving the optimal blend of AI automation and human expertise
- Accountability and governance in autonomous security environments
- Tactics for addressing AI-powered threats and compliance requirements
Speakers:
Vivek S. Menon — CISO & Head of Data, Digital Turbine
Michael Machado — CISO
Prabhat Sharma — Founder & CEO, OpenObserve
Host: Naveen S. Bisht — CEO, Akitra
Explore More with Akitra:
📚 Discover cutting-edge compliance and security insights on our https://akitra.com/blog/
🌟 See how our customers lead with compliance: https://akitra.com/customers/
🚀 Ready to future-proof your security program? Book a Live Demo today at akitra.com/demo
🔔 Don’t forget to subscribe to our channel and enable notifications for more expert discussions and compliance best practices.
#CISO #AgenticAI #Cybersecurity #Akitra #AILeadership #Compliance #SecurityLeaders #AIEthics
In this thought-provoking session, distinguished CISOs and industry experts dive deep into how Agentic AI is transforming the traditional CISO role—from operational oversight to strategic orchestration. Learn how to lead effectively, strike the right balance between automation and human judgement, and stay ahead of emerging AI-driven risks.
Highlights include:
- The impact of Agentic AI on the security landscape
- How the CISO’s responsibilities and leadership are evolving
- Achieving the optimal blend of AI automation and human expertise
- Accountability and governance in autonomous security environments
- Tactics for addressing AI-powered threats and compliance requirements
Speakers:
Vivek S. Menon — CISO & Head of Data, Digital Turbine
Michael Machado — CISO
Prabhat Sharma — Founder & CEO, OpenObserve
Host: Naveen S. Bisht — CEO, Akitra
Explore More with Akitra:
📚 Discover cutting-edge compliance and security insights on our https://akitra.com/blog/
🌟 See how our customers lead with compliance: https://akitra.com/customers/
🚀 Ready to future-proof your security program? Book a Live Demo today at akitra.com/demo
🔔 Don’t forget to subscribe to our channel and enable notifications for more expert discussions and compliance best practices.
#CISO #AgenticAI #Cybersecurity #Akitra #AILeadership #Compliance #SecurityLeaders #AIEthics
Category
📚
LearningTranscript
00:00Whenever I see that people do ask me if I was there during the Capital One breach and I was
00:26not. That is the time I made the switch to Chase and within like six months the breach happened.
00:35But being in the security space for 10 plus years and overall in the data space for 25 years
00:40and excited to be back. This is my second time with the CSER on table. So, you know,
00:45some familiar faces and then looking to make new friends as well.
00:49Thank you Vivek. And our second panelist here is Prabhat Sharma. Prabhat is the
00:55founder and CEO of OpenObserve. So, Prabhat, I'd love you to tell your background a little bit for
01:01the audience and what all great things you're doing with OpenObserve. Sure Naveen. Thanks a lot. And
01:06thanks a lot Vivek for your introduction too. And everyone great to meet you here. This is my first
01:12time at CSER on table. I'm the founder and CEO of OpenObserve. We are an open source observability
01:20company. We've been building the company for around three years. Prior to that I used to work for AWS
01:30and yes, I was at AWS when the Capital One breach happened. And I remember those days very, very
01:36vividly. And the fixes that at AWS that we had to build to ensure that this never happens again.
01:45Essentially, IMDF, we do. That is what came as a result of that. If you folks remember that.
01:52We specialize in providing, while full stack observability solutions, but what we are finding
01:59is in the industry, and one of the things that I talked to Naveen earlier was that
02:06in the case of Asian DKI, very large volumes of data that people are trying to crunch,
02:10especially in the area of security, think about CDN or security logs. And we'll talk more about that.
02:18But that's something that we are actually noticing extreme volume of data. Think about hundreds of
02:24terabytes of data being generated generated per day. And then I'm trying to analyze and run any kind of
02:30queries on top of that. I'll stop here for now.
02:32Thank you, Prabhat. And our next panelist is Mike Machado. Mike has also been CISO in a number of companies,
02:40including most recently with Beyond Trust, Shippo, and earlier RingCentral. And so, Mike,
02:47would you like to elaborate a little more with your background as well as whatever you'd like to say about yourself?
02:52Sure. Hi, everyone. I'm Mike. As Naveen said, I'm a multi-time CISO startup advisor.
03:01Occasionally, you know, contributor to the community. I have a fun, exciting project coming up that
03:08I've been working on with the Cloud Security Alliance and Noma and Haze and Harmonic Security.
03:15We're all going to announce that next week, in fact. So that's a fun industry project.
03:22But, you know, kind of by day CISO is what I do.
03:26Thank you, Mike. Really appreciate it. So let's, why don't we get started with our first
03:30topic is, you know, how the CISO's roles have evolved because, you know, both of you as well
03:37as even in the audience, we will reach out to them after this, is, you know, how you think five years
03:42and now and with this Asian PKI, whether will we have Asian CISO virtual or real CISO in future?
03:48So Vivek, would you like to start from there? Yeah. So I will tell you how the role has evolved over the last
03:54six to nine months. I've not seen a technology so rapidly adopted within a company in my 25 years in
04:05the industry. You know, we talked about breaches and AWS and Cloud. Like there was some forewarning
04:12about Cloud coming and then it took like a few years for people to warm up to it. But there is no
04:17forewarning with AI. There is no warming up. There was this AI came with so much hype around
04:25performance improvement and productivity enhancement and so on, that everybody in most enterprises
04:31thought, hey, we just use AI and magically things happen. And ChatGPT certainly did not help,
04:37given how, you know, uncanny it was in figuring out what you wanted and giving you answers at what you
04:43wanted. So shadow AI is a real thing. It has been a here and now problem for the last three to six
04:50months already. And there is no clear solution on how we would go about actually containing it,
04:57much less so fixing it. So that's one aspect. To answer your question about like whether we would
05:03actually have a CISO five years down the line, I think personal or digital authors that are very
05:10personalized would be a common thing, according to me. It could certainly be doing most of the
05:17non, you know, cognitive heavy thinking that a non-human can do. And then we humans would be left
05:25with sort of thinking about the what if scenarios that only we can with a certain amount of imagination.
05:32But it's yeah, it's an interesting time. And I guess the role will evolve far more rapidly in the next
05:38five years than it did in the last 20 years. Thank you, Vivek. Mike, how about you? I mean,
05:45you probably get that question I asked, you know. How's the role evolved in the past five years? You
05:51know, it's continuing to evolve. But over the past five years, my observations is it's moved from
05:57tactical to strategic, you know, tactical being, you know, we need to patch these things or solve the
06:02security problem. Just to strategic saying now it's a critical facet of of our business health
06:11from technical to business, meaning the interaction at the business level is is, you know, far more
06:17important than it ever was and continuing to be more important from, you know, friction inducer to
06:25business enabler is another transition of the role. The scope increases, you know, old technology
06:32never dies, new technology emerges. So most recently, we have AI, all of that needs all of the same
06:39security motions that mobile did that web did that cloud did that, you know, on prem did. And so
06:47that requires a constant kind of multifaceted development of ourselves. The roles evolved.
06:54Thank you, Mike. So, and Prabhat, you know, I, you know, I'm sure you also, when you said your
07:04solutions, CISOs are part of the buying process on the other side. So if you want to share any of your
07:10interactions with the current company or in the past, you think, you know, how you've seen the roles of
07:16CISOs evolving from your, from your perspective. Now, I'll, I'll leave this to Michael and Vivek.
07:23I've not, never been a CISO and I'm not really an expert in that particular topic.
07:29Yeah, no, that's fine. No, no worries. So, but we have other CISOs in the, the, the, in the audience.
07:35You see Randy here and a number of other folks. So if you would like to kind of provide your
07:40perspective with the, all this, uh, infusion of now agentic AI, I mean, the things have moved so
07:46fast, right? ChatGPT, gender AI, now it's agentic AI. So I don't know what will happen in a year.
07:51Yeah, no, I agree. Um, so Randy Barr, um, been a CISO, still a CISO and, uh, advisor for many
07:58different companies, but yeah, just a few years, it's changed a lot, right? The, uh, the adoption of
08:05AI, especially the hyper adoption compared to how we adopted cloud, how we adopted ASP, how we
08:10adopted mobile is a lot faster than any other technology that I've ever seen in my career.
08:17And CISOs before used to be able to be one of those, no, let's stop. Let's analyze. Let's
08:23understand before we move forward. But that today is now a business disabler. They have to enable the
08:29business. And so they need to be able to adapt to new change, new innovation, new adoption of tools.
08:34Um, and they're expected to help the organization securely move forward. That's the biggest thing.
08:41And, and it's a lot faster than what it used to be. Now, one of the things that I hear is AI is
08:46going to replace security roles. I don't think so. It's going to be just like how people used to say
08:50IT roles are going to be replaced by cloud computing. It's not going to eliminate it. Instead,
08:54it's just going to shift that focus because security professionals are going to be the ones that need to
08:59understand and how to implement solutions like prompt filtering, any tool execution guardrails,
09:05so it doesn't execute tools that are not supposed to be approved in that automatic workflow that is
09:10being introduced by agentic AI. You know, these tools or, or MCP, the protocols are now introducing
09:16secrets management. How do you protect against that? You know, do you introduce, uh, time to live
09:21for some of this stuff? How do you do it? Um, and also how do you ensure that's, uh, that every action
09:27that's taken by the agentic AI is auditable and that you can actually put guardrails or gates before
09:34it gets executed to the production environment or into your environment that could do a lot of damage
09:40in case something bad was introduced to that workflow. So a lot more to talk about that. So,
09:45but yeah, it's definitely changed. Yeah. Thank you, Randy. Appreciate it. Um, anyone else? I see
09:50number of other security senior executive, any of us, um, uh, do you have any, um, insights or any
09:58questions for the panel? Uh, I will, I will add more as we, as we continue. Uh, yeah. Thank you.
10:05All right. So why don't we- Well, I'd like to add a point. So I'm David Brown. I was the senior
10:12director and executive for multiple organizations in the middle of the last 20 years. I'm now DPA
10:19of operations and we've developed AI to actually replace security analysts at that tier one triage
10:28level and it's worked very well. So AI can be coming in and can successfully take over where a human
10:39analyst was performing at that triage level. Now I'm not saying more advanced areas, but in the triage
10:45area, absolutely. We've trained that model, that model operates and functions as a tier one analyst
10:52would, and then it picks it up or it puts it down depending on what that playbook and what that, that
10:59training does. So yes, there are places where I think AI is quickly adopted and they're bringing
11:05into place and helping actually grow and mature the cybersecurity space in a good way compared to
11:15what we're seeing like with marketing in other areas where AI is just generating just trash campaigns
11:23and spam emails, things like that. But if done correctly and under supervision, it can be a enabler
11:31to reduce workforce, improve efficiency, and then elevate the workforce you have to higher levels.
11:38So that's my take on that situation.
11:41Right. A hundred percent. And so those analysts now could focus more on strategic, do more investigative,
11:48and do more specific areas that would enable the company to do more than just looking and triaging
11:56those alerts. A lot of those AI solutions definitely helps replace some of the work that analysts are
12:02doing, which is, you know, very, very, not something that I would want to be doing.
12:06Absolutely. That's it, right? When you focus and understand where AI can be an enabler and an optimization
12:16and a maturity growth, you will win. When you just put it out there, and like somebody said earlier,
12:23CISOs are forced to adapt early and fast because businesses are shutting down. Where I was in the
12:30Middle East, in the UAE, right, Microsoft was giving away co-pilot organizations. So they were encouraged to
12:37use it, right? And they were using it incorrectly in 98% of every case. And the security around that
12:44was horrible because now employees were putting IP into co-pilot being shared who knows where,
12:52because they thought that that's what they were supposed to do. And there were no rails or guards
12:58around any of that. And it just went haywire. Now, roll that back, deploy it strategically, train it,
13:06keep it under supervision. And so when we deployed this, my team's job was to take a lot of those
13:13analysts, that's the AI did, and review them periodically. Say, hey, out of every 12, let's
13:20look at two of them. Did it answer it correctly? Would you have made the same decision? And if the
13:25answers came up no at any time, we go back to the training model, we retrain that model. And it's
13:31incredibly effective in that path. All right. Thank you, David. So let's, one, one area I wanted to focus on
13:39was governance and readiness from a data perspective. I know Vivek, when we talked the other day,
13:45actually yesterday, I think, on a prep call. So just checking on the questions, you mentioned that
13:53a lot of work in that area, and I'd like to hear your insightful perspectives over there.
13:58Yeah. So I think, you know, when we were discussing about questions to prep for this conversation,
14:05the one that I wanted to specifically talk about was the data readiness piece. You know,
14:10David was talking about how they trained a model slash agent to replace a triage analyst or a SOC
14:18analyst. I'm only assuming that they are able to do that and have that sort of human in the feedback
14:25loop kind of part because they had pristine data. A lot of us, when we think of AI use cases,
14:32especially the non-technical folks in our company, they think of AI being able to conjure up magic
14:40without actually understanding that if you don't have the proper underlying data assets with the
14:45right amount of metadata governance, this, it's almost impossible because your models won't be
14:51precise. It won't be tuned to your specific needs. And while the chances of hallucination is very limited,
14:58because it is not necessarily generating content, but you getting an answer where the probability of
15:05or the accuracy being sub 75 or 70% is high. So I just wanted to highlight that, like, if there are
15:12individuals on this call who feel like, man, either me or my organization is missing the boat when it
15:19comes to AI adoption. You are not. I think if you get, if you have clean data, you are 70, 80% of the way
15:28there. Let's let the big organizations fight out what foundational LLM model will win. But if you focus
15:36on the application layer, coupled with the clean metadata layer or data arrangement, you would still have a lot
15:44of time to catch up and actually get it right once instead of trying to iterate through and stumble
15:50through an AI solution.
15:51Okay. Thank you. Anyone else wants to chime in in that area? Prabhat, I know you had some other
15:59perspectives on the use of agent-tiki AI.
16:01I've got a few.
16:03Yeah. Go ahead, Mike. You want to jump in?
16:07It's been said a few times and also challenged something that's been said a few times.
16:14We've talked about using AI to free up the time of security folks to do other work and more advanced
16:21work. That's the camp I'm in. We've also talked about using AI to replace roles. I'm not in that camp.
16:29I don't think that's realistic, especially not in the near term.
16:33We, in the security community, have already had so much work that wasn't getting done because we
16:42didn't have enough people to staff it, both in terms of skill shortage as well as just personnel
16:48levels of various departments across companies, across the world. So what I see AI bringing to the
16:57table is one, enabling us to do some things that we just didn't have coverage to do before.
17:03Two, enabling us to do some things better through combination of human AI partnership working
17:09together. Three, democratization of some skills that used to be more rare on security teams,
17:18you know, scripting, coding, things like that. Now you can use generative AI, whether, you know,
17:24in a form of like chat GPT or co-pilots, but also platforms like, you know, lovable to build things,
17:33to build your own tools, right? Now we can start to overcome gaps in the product space in ways we've
17:41never been able to do before, you know, write scripts to pull from APIs.
17:46That's because of why, I don't think anyone's going away. I just think some of the toil
17:54tools, to kind of use a DevOps term, is now something we can automate away.
17:58Thank you, Mike. Mike, I would actually partially agree to you
18:03that people are not going away. Because what I have seen is that
18:08these things have not really been working. I mean, as David mentioned, that things were not working
18:17before, but their analysts are actually able to do things that people were actually doing earlier.
18:23And, in your words, it's enabling people, you're always short-staffed. And even three years down
18:30the line, five years down the line, we'll always be short-staffed. That's going to happen.
18:35However, what I'm seeing in a lot of companies, and I was talking to the CEO of an MDR company,
18:45who is trying to automate things. And what he mentioned to me was that,
18:50until six months ago, they were trying to push or put all the agentic AI into a lot.
18:58I think someone is too eager to join.
19:00I think you have a friend who's agreeing with everything I've said.
19:04That's right.
19:05There is a future AI agent.
19:08Yeah.
19:08She wants me to take her to somewhere. She just came back.
19:13Yeah. But the story being that, things required a lot more intervention and a lot more supervision.
19:23But as technology is becoming better every day, now, six months later, and he started the company
19:32roughly around two months, two years ago, until six months ago, they were not doing a whole lot
19:38with agentic AI. But two months ago, they tested it again. And it was like, damn, this is working.
19:45And it's like, everybody shift and put agentic AI out over here, because this really, really works
19:51out over there. Something that I would have done personally, a lot of these things, probably a couple
19:56of days, or maybe a couple of hours, if not a couple of days. Now, we can do it in a couple of minutes.
20:02And intelligence will keep rising. And at some point in time, probably, if there are 10 junior
20:10roles, there will possibly be five junior roles. But at the top, where you really need people to
20:17make strategic decisions, I think it is going to be very, very helpful for them, as well as very
20:24empowering for them. So I agree. I'm concerned about your comment, though, about fewer entry
20:33level roles, or early career roles. I think you're right, by the way. I'm not disagreeing with you.
20:40But obviously, early career roles are how people become more experienced in their careers. So I
20:46think this is a bit of a nuance we're going to have to work through. And I agree with you, right?
20:52I mean, the tools that the technology is able to do progressively more and more, which is awesome.
20:58I just think it will be freeing up people's time for other things, as opposed to eliminating people.
21:04Someone was talking, though, and I went first. So please.
21:10It's okay, Michael. Yeah, no, I think, you know, there's a lot of hype around agents. And, you know,
21:15I have a contrarian view about this. I think, you know, for most of this are still in the dev
21:21staging phases. I think for real production deployments with massive scale agentic, like,
21:27you know, fully autonomous, we have like probably like several years away. It's I think it's just a
21:33lot of unknowns from memory, from cost, from reasoning, from hallucinations, from everything.
21:39Right. So I think we just have to take it in a caution, not get overexcited, like too quickly,
21:45too fast, you know, before, you know, it's charge switch is allowed.
21:51I think we'll have to take it like one step at a time. And I think coming back to AI was there
21:56even before LLMs. I think what LLM has added is the NLLP wrapper around it. So it makes things around
22:03scripting and data integration, the front end part of the equation, perhaps a little bit smoother.
22:08But in terms of pure automation, I mean, AI has been around that has solved it. So I think the real
22:14question would be is what is LLMs giving you on top of that, that pre-LLM AI is not able to do
22:21at a pace at which where the cost of the economics and the deployment models justify, justify this
22:27new mousetrap. Right. So I think that's an open question to everyone here and to the broader
22:32community as well. Right. So it's a lot of unknowns. I don't have the answers.
22:36Thank you. If I could jump in real quick again, I agree with what
22:44to the analyst roles and things like that might be limited in new worlds because AI can replace
22:52intelligence. But as I've said in the most repeated before, those senior level roles will never go away.
22:59It's the one thing that we've seen time and time again that we cannot let AI do is make kinetic
23:06decisions, right? That ability for AI to make a kinetic decision, whether that be the blocking of
23:14something, the terminator of something, the shooting of something, whatever that is, that must still remain
23:20with a human. Because the AI in almost all cases, no matter how well the models train or how advanced it
23:29is, it always looks at the odds in the numbers and goes, if I have to protect asset one and asset two
23:38at the same time, I'm going to divert skills between those to try to do it equally. Or it's going to make a
23:45decision to prevent asset one or two where an analyst can sit back, immediately confer with peers,
23:54make a better informed decision of which asset needs to be protected more, take resources away from the
24:01least important asset to ensure that the more important or depending on what the attack or what
24:07the outcome is, what the objectives are, see that three, four steps ahead, that the AIs just don't
24:14have that capability today, tomorrow, or maybe even five years from now to do that. So those kinetic
24:21decisions can never be in an AI hand. They must be in an analyst hand. Those analysts will be senior and they
24:28will have their job security. Where that triage analyst, yeah, we may hire 10 today, but tomorrow only hire
24:37five because I can replace five with that AI. I think that's a legitimate thing to look at by two
24:43cents. Okay. Anyone else has any questions or insights you want to add? So I think AI can get
24:51us there though, right? I mean, we can create that autonomy from beginning to end where it does do that
24:56execution in the production environment, but there are things that you can implement in place that, you know,
25:01like you said, David, we need individuals to put those controls in place, not necessarily look at it and
25:06review every single one. For example, you can make sure that before it goes to production, any rollout goes to a
25:13sidecar or proxies. Somebody has to build that structure. Somebody has to pull that policy in place. Somebody
25:19also has to put in any specific action that says, hey, if you're going to roll this out, limit the way it's going to
25:25roll out, or make sure you have rollback procedures in place right away, and make sure that's part of the process. So
25:30somebody has to think through that and put that in place. But that's the thing, those people that are
25:34working as those analysts, now you can help shape their career faster by having them work on the more
25:41difficult stuff. But yeah, there are certain areas that we can absolutely replace current human work
25:47because it's repetitive or that it's something that you can actually analyze very, very quickly
25:54through AI instead of somebody having to read and interpret.
25:58Thank you, Randy. So shifting, what I would like to do is shift gear into more, now this is more on
26:04the technology side, but from your perspective, whether Mike, you know, Vivek, Prabhat, how do you see in
26:11terms of the, like the customers buying really, you know, there's a high technology, a lot of solutions
26:19have been given, but from a budgetary perspective, how do you think, Vivek, you are evaluating some of
26:24these technologies in your own organization or, you know, and then others also would, it would be great
26:30to hear. I mean, you've got a capital of money.
26:33Yeah, just to be clear, Eric, are we talking about buying AI tools?
26:38Yeah, yeah, BI.
26:40Yeah, so I, at least in my case, we have to make our own budget, so to speak. We have to show that we
26:50are reducing spend elsewhere to justify the spend that you're making on a net new tool.
26:57And so I'll give some concrete examples. We have been building, so we are a Google shop,
27:04and so we have adopted Google agent space as much as we would like to be working with Zapier and
27:10and all the, you know, stuff that actually works. Being a Google shop and being a security guy, I have
27:16to limit everybody in the company to work on Google agent space. We are working through it. And what it
27:22has allowed to do is it has allowed us to do, allows to automate some basic use cases around security
27:30questionnaire answering that we have been able to automate about 80 percentage. And we continuously,
27:37you know, evolve that, that agent to do better, better and better. We have also been able to sort
27:44of create a chat bot for the company, where they can come and ask for security policy updates and so
27:50on. Whereas otherwise, they would have reached out to us, etc, etc. But that's us leveraging an existing
27:58capability that Google has provided, being that we are a Google shop, where I do see AI that we can leverage
28:06today is around compliance automation. And it's, again, something that we have seen
28:13some results of using the agents that we have built internally using agent space, but agent space
28:19as a platform is still maturing and is not there. But compliance automation is also where we have
28:26multiple touch points to some of the senior tech resources in our company. And there is always
28:30pressure on the compliance team to reduce that exposure, so that they can go back and to do what
28:37they do best, which is to actually build good tech. So in that case, you're looking at compliance automation
28:42tools, we're also looking at automated test tools to help out our QA team, and so on. So very practical
28:51applications of AI in our company, that we feel we would get a leg up. I think Mike was making the point that
28:58there can never be enough resources. That's always the case. I own data as well in the company.
29:04And I can never have enough, forget about the security, but that I think I, in my lifetime,
29:08I would never be fully staffed. But even on the data front, we are not, you know, we don't have enough
29:13data scientists. And so we have looked at AI tools to sort of accelerate that journey. So the investment is
29:19very specific and pointed. But also, it is not something that we can go to the board and ask for net new
29:26money. We are, we have to leverage, we have to let go of a tool that was not giving us a right benefit,
29:31and then, you know, shift that investment over to buying any new AI tool that might benefit us.
29:37Thank you Vivek. Great perspective, and great use cases. Mike, how about you? I know you've been,
29:44you know, working with even CSA, as well as many other companies,
29:47companies, where I'm sure you're looking at how they are buying patterns, right?
29:53Yeah. More on like, is it real? I mean, you know, obviously, there's a high solutions,
29:56lots of solutions, whether their AI tools are being bought or not.
30:00Oh, it's, it's real. I mean, it shows up in quotes all the time. But,
30:04but,
30:05But are they writing the checks?
30:07Yeah, I think, here's, here's my thought. If you're getting AI capabilities, that's one,
30:13one element. Another would be any project where you're looking at various ways of doing it. If
30:19the AI version has a higher ROI than the non-AI version, that kind of speaks for itself. Third
30:26would be optimizing existing run rates, stop investing in a tool or save money in a renewal
30:32to free up dollars to invest in somewhere else. I will say, I'm not personally convinced that many
30:41things that like, like I, I see charging for AI as in the experimental stage at this point,
30:47right? Everyone's investing in AI, they're trying to figure out how to make an agent agent, you know,
30:55in many ways, treating like agents, like we could sell you a bunch of agents, and then we could sell
30:59you a bunch of extra seats to use those agents, right? I think long term, we're going to start to
31:05see a consolidation where AI features, or, you know, AI is just going to be like a feature set,
31:12right? And, and, you know, it's like saying, I want to charge you extra for reporting. At some point,
31:17it will just be a built in expectation table stakes that tools operate with AI.
31:23And this is, I think we'll, we'll stop seeing efforts to have a separate line item. I think
31:31that's down the road, but not, but I'll, but I mean, to, to further Mike's point, I'm already seeing
31:40that the Google Gemini used to be a separate skew that we had to buy, but now it's embedded and it's
31:46part of a GWS. And I think agent space is going down the same route. It will eventually be part
31:53of the combined workspace skew and not like a separate skew that they sell today. So it's,
31:58it's going to, it's going to happen. Just a matter of maturity.
32:03Yeah. Yeah. Actually, that's a great point, Vivek. I just saw an email from Google workspace. Oh,
32:08you can use our video editing tool now free right now. Right. So I just, that's similar. And at some
32:14points people start using, then they want to charge, then they will, you know, they'll say,
32:17okay, we have to give it. We see that anyway, it's even in software, right? You start with one thing,
32:23they say, okay, I need another one. And you keep getting the prices and you keep giving them free,
32:26right? So, uh, great to hear from your side. Um, are you collecting some money by selling some
32:34observability agents? I am seeing something very different. I mean, um, orthogonal to what, uh,
32:42Wake and Mike mentioned, uh, a lot of people have started using a lot of, uh, these agents and the
32:50side effect of, of that is that you're making a lot more API calls, not only to these LLMs and,
32:59but these LLMs are making a whole lot more calls now through these LLMs. You're making a whole lot
33:03calls to your underlying systems to get data. And especially in the area, and, and we are seeing
33:10some impact of that is that many, many, many companies are locking down their APIs or reducing
33:17their rate limits. What we're also seeing is that, especially in the area of security,
33:25of course, and in observability, because the underlying data, for example, you have got
33:30tons of logs out over there and you want to do log analysis and your MCP server, you created an MCP
33:36server and your agent is hitting that MCP server. Now, if you have got a couple of hundred terabytes
33:43of data or a couple of petabytes of data in your, um, logging system, say for example, open observe,
33:49Splunk, Sentinel one, wherever, if that system you're calling that first, each and every call is
33:56actually quite expensive because it has to shift through so much of data. And it, you can't really
34:04make so, so, so many calls unless you have extremely efficient systems, extremely fast systems.
34:10So that is something that we are seeing where many of these systems are actually like,
34:16I think is any system that you, that we are building, they're not able to leverage the underlying
34:20system very well, simply because the underlying systems are overloaded, are getting overburdened.
34:26So that is a fact getting, that is a side impact of the agent, these systems right now.
34:31And unless we build the underlying infrastructure that can actually handle all of that,
34:37this is going to be a challenge.
34:41Yeah, I think this is, this is back to my earlier point.
34:43Sirena, can you hold on? I just want to cover some other folks here before we go.
34:48Sorry about that. Manish Rawat, I think you put a good, nice, it'll be great if you can,
34:52if you can be on the video too and just say two, two lines about yourself and maybe you can ask us,
34:56bring out the points you're bringing up in the chat.
35:03I thought Manish was there, maybe just put on the chat and then he's gone.
35:07All right Sirena, it's your turn.
35:09No, no, I was just, this is just a hit, hit a pivot, but the democratization of the agent
35:14runtime footprint, I think you know, we're still in the early stages, right? So the cost, the economics,
35:19the security model, everything.
35:21All that I think they're still in the early
35:23days. We have to hit that escape
35:25velocity before it can be
35:26perceived as wide stream deployment.
35:29I think that's what I'm trying to say.
35:31Okay. All right.
35:33So I think Raj
35:35here, you have a good question you wanted
35:36to ask the panelists. Would you please?
35:39I know you're on mute.
35:44Looks like
35:45Raj. Yes, please.
35:46Yes. I was just, you know,
35:48I'm on the hardware side just to
35:50give everybody a background. I'm
35:52putting in the fiber cabling
35:54and I just got back
35:56from SFO
35:58from a plane ride to Louisiana
36:00where Meta
36:02and a few other companies are
36:04building lots of data centers.
36:07So I guess my question
36:08is somewhat
36:10a young question.
36:12How is the cybersecurity
36:14practices
36:15moving to defend against
36:18sophisticated
36:19nation state
36:22and ransomware attacks?
36:24you know, what are they going to do?
36:25How are they going to react
36:27to these
36:28rising
36:28AI powered threats?
36:31There's a lot of data.
36:33I think Meta is up to
36:33100 data centers
36:34and
36:35you know, I'm curious
36:37how
36:37we see this evolving.
36:39I think that's a very complicated
36:48multifaceted question, right?
36:51There's everything from poisoning of data sets
36:55which I think is probably going to be as hard to uncover
36:58as
36:59backdoors in code
37:01which is, you know,
37:02like a really difficult challenge.
37:04Sometimes
37:05some people say
37:06an even
37:06impossible challenge.
37:08But then there's other things
37:09like
37:10using AI
37:11in the way that
37:13adversaries use AI
37:14as sort of this like
37:16you know,
37:17ladders and walls race, right?
37:19Like
37:19adversaries using
37:21LLMs
37:21to generate
37:22malware variants
37:24that
37:25bypass
37:26detections,
37:28right?
37:28but at the same time
37:30detection vendors
37:32using LLMs
37:34to do the same thing
37:35to try and find
37:36these variants first
37:37that bypass
37:38detections
37:39so that they can update
37:40their detections,
37:41right?
37:41So, I mean,
37:42these are just two examples.
37:44I don't think they're exhaustive
37:45by any means
37:46but in some cases
37:47I think
37:47my
37:49year
37:50yeah, I don't know what happened.
37:51I literally got brand new internet today
37:53that seems to work fine
37:54but maybe my Euro mesh
37:56is now having problems.
37:56Maybe it's like
37:57analogous to like
37:58new AI agents,
38:00you know?
38:00Yeah, you know.
38:03All right.
38:04So, Srini,
38:05you wanted to add in
38:08something.
38:08No, I think all I'm saying
38:09is that, you know,
38:10you have to skin the layer
38:12at multiple dimensions,
38:13right?
38:13So, there's model security,
38:15there's API security,
38:16there's prompt security
38:17and then there is
38:18the traditional web layers
38:20and all the facets
38:21that touch it
38:22are now
38:22because the threats
38:23can now
38:23transcend
38:24and cross
38:25jump from one
38:27to the other
38:27it just makes
38:28the problem
38:29of detection
38:29and enforcement
38:30much harder, right?
38:31So, all these
38:32are good opportunities
38:33there are lots of
38:33young startups
38:35and companies
38:35who are trying
38:36to address it
38:36and all could
38:37potentially become
38:38big companies
38:38but we are still
38:40early days in that.
38:42All right.
38:44Anyone else
38:46from the other folks
38:47who haven't participated
38:48if you have any insights
38:50or you want to ask a question
38:50so you just want to make sure
38:51you are also
38:52Ravi Lingarkar
38:55anyone else
38:57Mihir
38:57you have any
39:01insights
39:02anything you want to add
39:03or anything
39:04Anurag
39:09okay
39:11My question is simple
39:13like
39:13is there a popular
39:14agent application
39:18that
39:18is in CISO
39:20red arts
39:21other than
39:24building agents
39:25themselves
39:25Do you mean AI tools?
39:31Yeah
39:31something like
39:32like drop zone
39:33like
39:34like vendors
39:35that are already
39:35sort of
39:37in our
39:37mental
39:38models
39:39yes
39:40well that's one
39:41that's in mine
39:42okay
39:44drop zone
39:44yeah
39:45I'm actually
39:47thinking much
39:47more practical
39:48meaning
39:49much more
39:49practical
39:50to me
39:50which is
39:51perplexity
39:52you can do
39:54perplexity
39:55has
39:55you know
39:56I don't know
39:56if there are
39:57perplexity users
39:57in this forum
39:58but
39:58if you have seen
40:00their evolution
40:00over the last
40:01six to nine months
40:02I've been using
40:03perplexity
40:03for the last
40:0418 months
40:04approximately
40:05but
40:06it's
40:07like
40:07you know
40:08there are articles
40:11about them
40:11being Google killers
40:12and all that
40:13obviously overblown
40:14but
40:15they
40:15Google
40:16certainly is talking
40:17about perplexity
40:18in their board meetings
40:19and their executive
40:20meetings and so on
40:21but they have made
40:22some really good strides
40:23I've also
40:24you started using
40:25Comet
40:26their agentic browser
40:27for the last
40:28two weeks
40:29and I
40:30I'm amazed
40:31at what it can do
40:32from a day-to-day
40:32task point of view
40:33so we are
40:35been in
40:35earlier in the conversation
40:36I talked about
40:37like
40:37digital avatars
40:39and
40:39you having your own
40:41personal assistant
40:42et cetera
40:42et cetera
40:43I don't think
40:44we are too far away
40:45from that
40:45from that day
40:47maybe a year
40:4718 months
40:48where I'll tell
40:50Naveen
40:50my agent
40:51will talk to
40:51your agent
40:52and we'll fix
40:53the time
40:53for us to speak
40:54right
40:54it's
40:56it seemed
40:57far-fetched
40:58two years ago
40:59then 18
40:5918 months ago
41:00seemed like okay
41:01but now
41:02it's almost
41:03almost there
41:03and
41:04so I don't
41:05yeah
41:05drop zone
41:06is a good one
41:07but Anurag
41:08I
41:08just to keep
41:09things simple
41:10just looking
41:11at perplexity
41:12and the evolution
41:13that they have
41:13gone through
41:14over the last
41:15six to nine
41:15months
41:16and this whole
41:17agentic browser
41:18thing which
41:18they were the
41:20first to market
41:20with
41:21yeah
41:22lots and lots
41:23of possibilities
41:24right
41:25things that
41:26yeah
41:26things that we
41:27do today
41:27or we have been
41:28doing for 10
41:29plus years
41:29I think
41:30some of the
41:31younger
41:31professionals
41:33coming into
41:34workforce
41:34would not even
41:35do that
41:35a year
41:36from now
41:37yeah
41:38so perplexity
41:39will fall
41:39in the
41:40productivity
41:40space
41:42right
41:42I mean
41:42it will
41:43improve
41:43the productivity
41:44whereas drop
41:44zone
41:45is more
41:45like a
41:46IT application
41:47that can
41:48be leveraged
41:48for
41:49internal
41:50usages
41:50right
41:51yeah
41:52but it's
41:52very interesting
41:53to know
41:53that perplexity
41:54has made
41:55so much
41:55drive
41:55that
41:56it's
41:56entering
41:57the IT
41:57departments
41:58internally
41:59so
41:59that's
42:00huge
42:01yeah
42:02thank you
42:03you asked
42:04me
42:05my question
42:06is
42:06what we
42:08are doing
42:08right now
42:09is
42:09we are
42:10basically
42:10replicating
42:11all the
42:12all the
42:13what I
42:13will call
42:13the legacy
42:14tools
42:14with a
42:16you know
42:16AI washed
42:17tools
42:18even take
42:18the example
42:19of drop
42:20zone
42:20alert
42:21fatigue
42:22yes
42:22so we
42:23are
42:24there are
42:24many many
42:25tools now
42:26solving
42:27a single
42:28one aspect
42:29of the
42:30whole
42:30security
42:31workflow
42:32so what
42:33will happen
42:34is we'll
42:34end up
42:35creating
42:35yet another
42:37layer of
42:37tools
42:38that are
42:39just AI
42:40just plus
42:41AI
42:42whatever
42:42plus AI
42:43and we
42:45are not
42:45converging
42:45this is not
42:46a con
42:46we are
42:47diverging
42:48we keep
42:48diverging
42:49and then
42:49complaining
42:50why the
42:51security
42:51problem is
42:52not being
42:52solved
42:53this is
42:54pure
42:54divergence
42:54we keep
42:55adding more
42:55and more
42:56tools
42:56and average
42:58enterprise
42:58already has
42:59you know
42:5950 plus
43:00tools
43:01120 plus
43:03sorry
43:04120 plus
43:05tools
43:06yeah
43:06depending on
43:07those
43:07values
43:07this is a
43:10divergent
43:10problem
43:11if you know
43:11control system
43:12there is no
43:14feedback loop
43:14there is no
43:15dampening factor
43:17here to bring
43:18it to converge
43:19so
43:20we're just
43:23blowing
43:23everything up
43:24again
43:24and that's
43:26where I
43:26think it's
43:27very interesting
43:27where we're
43:28at today
43:28right
43:29because
43:29you know
43:30with
43:30Anthropic
43:31coming out
43:32with MCP
43:33there's a lot
43:33more people
43:34that are
43:34integrating
43:35their own
43:35tool sets
43:36with their
43:37own LLMs
43:37and trying
43:38to figure
43:38out how
43:38do they
43:39put all
43:40that together
43:41you don't
43:41have to
43:41wait for
43:42a vendor
43:42anymore
43:42these days
43:43you can
43:43start
43:44building
43:44your own
43:44solution
43:45in-house
43:46the problem
43:46with that
43:47is that
43:47it makes
43:48it easier
43:48for anybody
43:49in the
43:49company
43:50to do
43:50that
43:50and therefore
43:51the security
43:53opportunities
43:53just explode
43:54it's like
43:55the shadow
43:56AI
43:56and then
43:57when that
43:58person
43:58who built
43:59it
43:59leaves
44:00then there
44:01is panic
44:01nobody's
44:04there to
44:04maintain it
44:05that's why
44:06it is
44:06the carrier
44:09path is
44:09resistant
44:09proof
44:10right
44:10so
44:10security
44:11you always
44:12need it
44:13thank you
44:14this is
44:15some other
44:15like Bill
44:16and Krishna
44:17do you guys
44:18have any
44:18insights
44:19or you
44:19want to
44:19ask any
44:20questions
44:20you had
44:22an opportunity
44:23as well
44:24this is
44:26Bill here
44:27now I
44:27just I'm
44:28enjoying the
44:28conversation
44:29okay great
44:30awesome
44:31and Krishna
44:32no I don't
44:34have any
44:34questions
44:35all right
44:37how about
44:37Nikhil
44:38Jason
44:39Ram
44:39Jason good
44:40to have you
44:41here
44:41do you have
44:41any questions
44:42or any
44:44insights you
44:44want to
44:44provide
44:45okay
44:46all right
44:48okay and
44:49how about
44:49sorry someone
44:51was saying
44:51something
44:52very interesting
44:55perspective for
44:56many people
44:56and one
44:57comment made
44:58was that
44:59your agent
45:01will be talking
45:01to another
45:02agent to set
45:02up a meeting
45:03and there was
45:03a comment by
45:04David saying
45:04that kinetic
45:05decisions are
45:06going to be
45:07difficult and
45:08it may not
45:09be possible
45:10at some
45:11point in
45:11but I think
45:12if your agent
45:13is exposed
45:15to as much
45:15as your
45:16ability to
45:17think how
45:17much you
45:17want to
45:17train your
45:18agent your
45:18agent will
45:19be able to
45:19do kinetic
45:20decisions
45:21depending on
45:21what is the
45:22amount of
45:23freedom and
45:23leverage you're
45:24giving so I
45:25think the
45:25world is going
45:26to get there
45:27and anything
45:28that is not
45:30so critical or
45:31even fairly
45:32little critical
45:33such kinetic
45:34decisions will
45:34all be
45:35automated and
45:36that's going
45:36to happen
45:36very fast
45:37like someone
45:37else was
45:38mentioning in
45:39a matter of
45:3912 to 18
45:40months yes we
45:41will start
45:41seeing that
45:42happening a
45:42lot more in
45:43our domain
45:44at least
45:44thank you
45:47all right
45:49so we are
45:50almost coming
45:51to the last
45:51three minutes
45:52so what
45:53would be
45:54you know
45:54is there
45:57anyone who
45:57can foresee
45:58like what
45:58will happen
45:59in two
45:59years at
45:59least now
46:00since in
46:01the last
46:01two years
46:02since the
46:02chair GPD
46:03came I
46:03can't even
46:04predict you
46:04know what
46:04even though
46:05we keep
46:05progressing in
46:06our own
46:07platform
46:07also fast
46:08but we
46:08don't know
46:08what next
46:09two years
46:10so I
46:10think maybe
46:10from your
46:11side
46:12you know
46:13I think I
46:14already made
46:15my prediction
46:15the my agent
46:17will talk to
46:17your agent
46:18part I
46:19think it's
46:20it's real
46:21it will
46:22happen I
46:23also foresee
46:24sort of this
46:25hype almost
46:25beginning to
46:26die down
46:27we are seeing
46:28a lot of
46:29M&A activity
46:30around the
46:31you know
46:32AI space
46:33already but
46:34two years
46:36is like
46:37you know
46:38is a long
46:39time in the
46:39AI space and
46:40we would stop
46:41having this
46:41fight over
46:42AI talent
46:43by then
46:44that we're
46:45seeing
46:45currently
46:46today
46:46with like
46:47hundred
46:48million plus
46:48packages
46:49and so on
46:49Mike
46:52I think
46:54we'll see
46:55normalcy
46:56within two
46:56years
46:57right
46:57it's
46:58it's
46:58it's
47:00still going
47:00to be
47:01obviously
47:02widely used
47:02widely deployed
47:03but I think
47:04it'll be
47:04less
47:05less feared
47:06and more
47:07normal
47:07kind of
47:08more
47:08more
47:09in everyday
47:09things
47:10yeah
47:11how about
47:12where do you
47:14see in two
47:14years
47:15you deal
47:15with a lot
47:15of data
47:16at least
47:17I know
47:18of
47:19since you're
47:20my customer
47:21yeah so
47:23I think I
47:25would echo
47:25with what
47:26everyone has
47:27said that
47:27there is
47:27going to be
47:28a whole lot
47:29more agent
47:29again
47:30okay okay
47:32okay okay
47:32okay
47:33there are
47:33going to be
47:34a whole lot
47:34more agents
47:35the only
47:39thing that I
47:39fear is
47:40how much
47:41of load
47:42all of this
47:43is going to
47:43put on
47:44infrastructure
47:44which we are
47:47already seeing
47:47and unless
47:49we build the
47:50underlying
47:50infrastructure to
47:51be strong
47:51enough
47:52things
47:53we might
47:55start seeing
47:55some cracks
47:56out over
47:57there
47:57that
47:58at least
47:59in my
47:59industry
47:59where I'm
48:00operating
48:00with customers
48:01I've already
48:02started seeing
48:02some of
48:03them
48:03yeah
48:05so there
48:06are some
48:06of the
48:07folks I
48:07didn't
48:08Hiram do
48:08you have
48:09any questions
48:09or insights
48:10you want to
48:10provide and
48:11you just want
48:13to make sure
48:13that you
48:14want to
48:15anything you
48:16wanted to
48:16ask the
48:16panelists
48:17no I
48:18actually just
48:19really enjoyed
48:20the conversation
48:20it's nice
48:21hearing the
48:22back and
48:22forth of
48:22everybody and
48:23I just
48:23appreciate
48:23the great
48:24conversation
48:25okay
48:26thank you
48:26and how
48:27about
48:28hopefully
48:29I'm not
48:30butchering
48:31the name
48:32okay
48:36anyone else
48:38Cooper
48:38if I
48:41missed
48:41okay I
48:41think I've
48:41covered anyone
48:42yeah
48:42Randy you
48:43wanted to
48:43say something
48:44yeah in
48:45two years
48:45I think
48:46we're going
48:46to start
48:46seeing a
48:46lot more
48:47case filings
48:49against
48:49individuals
48:49that are
48:50not meeting
48:50the AI
48:51requirements
48:52especially in
48:52the EU
48:53AI Act
48:53we're seeing
48:54that right
48:54now
48:54we're going
48:55to start
48:55seeing people
48:56not being
48:56able to
48:57sell into
48:57certain
48:58countries
48:58we're going
48:58to see
48:59startups
49:00that need
49:00to carry
49:01a certain
49:02burden
49:02around
49:03any
49:03potential
49:04fines
49:04against
49:04them
49:05we have
49:06a startup
49:07in San
49:07Francisco
49:08that cannot
49:09sell into
49:09Italy
49:10and haven't
49:11been for the
49:11last three
49:12years
49:12and then
49:12recently got
49:13fined
49:13two million
49:14I think
49:14it's two
49:15million dollars
49:15if every
49:16other country
49:17starts following
49:17suit that
49:18startup company
49:19is not going
49:19to be able
49:19to survive
49:20so we're
49:21going to
49:21start seeing
49:21a little
49:21bit more
49:22structured
49:23on how
49:23we sell
49:23into other
49:24countries
49:24the
49:25responsibilities
49:26a lot
49:27of startups
49:27have
49:27including
49:27larger
49:28companies
49:28on how
49:29they adopt
49:30the AI
49:30and publish
49:31it
49:31so we're
49:32going to
49:32start seeing
49:32a lot
49:32more
49:32transparency
49:33and maybe
49:34figure out
49:35who is it
49:36that they're
49:36going to
49:36have to go
49:37to in a
49:37company
49:37when AI
49:39is not
49:40met
49:40based on
49:41the regulations
49:41that are
49:42out there
49:43today
49:43or going
49:44to be out
49:44there in
49:44the next
49:45year or
49:45two
49:45yeah
49:46thank you
49:47Randy
49:47really
49:47appreciate it
49:48anyone else
49:49Vivek
49:50or anyone
49:51has lost
49:51and so
49:52otherwise
49:52we will
49:52adjourn the
49:53meeting
49:53Vivek
49:53you want
49:54to
49:54I was
49:55just
49:55going to
49:55ask
49:55since we
49:56have
49:56a good
49:57group
49:58of people
49:59can I
50:00just
50:00poll
50:01to see
50:01what
50:02and where
50:03do people
50:03go to
50:04keep up
50:04with all
50:05the AI
50:05news
50:06I'll
50:07share
50:07mine
50:07if it
50:08matters
50:08I
50:09subscribe
50:11to the
50:12rundown
50:12that's
50:13Andrew
50:15Eng
50:15it's a
50:17newsletter
50:17that comes
50:17out
50:18definitely
50:18a weekly
50:19sometimes
50:19twice a
50:20week as
50:21well
50:21and then
50:22I follow
50:23Sanya
50:24Oja
50:24she's a
50:25partner
50:25at
50:25Bain
50:26Capital
50:26on
50:27LinkedIn
50:27and she
50:28pretty much
50:28does a
50:29daily
50:29rundown
50:30on all
50:31things
50:31here
50:31and then
50:33she wraps
50:33it up
50:34with a
50:34Sunday
50:34weekly
50:35as well
50:35and I
50:36think
50:37hers
50:37is far
50:38more
50:39insightful
50:39but
50:40I just
50:40wanted
50:40to ask
50:41what
50:41people
50:42do
50:42is there
50:43anything
50:43that you
50:44would
50:44recommend
50:44folks
50:46here
50:46should
50:46follow
50:46or read
50:48similar
50:53avenues
50:53sorry
50:56go ahead
50:56I
50:57subscribe
50:57to this
50:58this
50:58Turing
50:58post
50:59I think
50:59it's like
51:00a
51:00it's a
51:01it's a
51:02it's a
51:02it's a
51:02blog post
51:03okay
51:03it gives
51:04all the
51:05latest
51:05it's a
51:05little bit
51:06more technical
51:06but you'll
51:07enjoy it
51:07it's called
51:08Turing post
51:09yeah
51:09yeah
51:11I also
51:11get a
51:11lot of
51:12the
51:12photo
51:12do
51:13I
51:13think
51:13something
51:14didn't
51:15come off
51:15my mind
51:16I think
51:16Randy
51:17was also
51:17sharing
51:18something
51:18yeah
51:18oh yeah
51:20I'd take
51:20a more
51:21podcast
51:21approach
51:22but I
51:22use
51:22notebook
51:23LLM
51:24to go
51:24search
51:25the
51:25latest
51:25articles
51:26and then
51:27put it
51:27together
51:28you need
51:30an agent
51:31for that
51:31all right
51:36that's a
51:36good way
51:37to catch
51:37up in
51:37addition to
51:38all the
51:38readings
51:38that I
51:38do
51:39all right
51:40okay guys
51:41so thank
51:42you so
51:42much for
51:43everyone
51:43joining
51:44today for
51:45a very
51:45interactive
51:46discussion
51:47you know
51:48really appreciate
51:48your time
51:49and a lot
51:49of you
51:50are like
51:50Midwest
51:50or Mountain
51:51some of
51:53the folks
51:53really appreciate
51:54you all
51:54being here
51:54thank you
51:55again
51:55thank you
51:56thank you
51:56everybody
51:56bye
51:57thank you
51:59all
51:59thank you
52:00all very
52:00much
52:00thank you
Comments