Latest / State of the AI Union / Mark Azad: Don't run the risk of Snake Oil AI
Transcript
- 0:06All right, and welcome everybody to the State of the Union.
- 0:09I am your host, Laura. And joined with me today is Mark
- 0:13Azad, CRO of Across AI. Mark, would you like to
- 0:17introduce yourself and across please?
- 0:19Yes, thank you, Laura. Mark Azad, as you mentioned, I'm
- 0:21CRO here at Across AI. I've been acro and a seller for
- 0:25the better part of 20 years primarily in database and
- 0:27infrastructure technology. I work for a company called
- 0:30Across AI, but we've built reasoning graph technology with
- 0:33long term memory and one of the first use cases we are applying
- 0:36that to is within a go to market world, but we're also expanding
- 0:39into marketing and operations and we'll talk more about that.
- 0:43Yeah. So Mark, when we talked earlier,
- 0:47one of the things that you had talked a lot about was how the
- 0:50reasoning graph is a lot more, this is the new standard.
- 0:53It's the new standard for how AI should operate.
- 0:56Tell us a little bit more about that and how that differentiates
- 0:59from RAG and how everybody has been talking about it thus far.
- 1:03Yeah, certainly. So if you're familiar with RAG
- 1:05database retrieval systems, they, you know, one or two years
- 1:09ago they were very impressive, right?
- 1:11They could provide quick research, quick analysis, but
- 1:15you quickly came across enterprise deployment
- 1:18capabilities that I guess limited the value and what they
- 1:21could provide. What we've seen in the market is
- 1:23that you need a much more robust underlying technology that
- 1:26specifically accounts for things like context, inferences and
- 1:30memory. And so you can think of the
- 1:32reasoning graph, at least how we built it here at Across AI, as a
- 1:35system that not just takes the raw data and provides analysis
- 1:40on top of it, but it starts with the end in mind.
- 1:42And what I mean by that is you take a next best action that a
- 1:45human wants to apply. It could be a marketing use
- 1:48case, a sales use case, operations use case.
- 1:51And then we reverse engineer it down to the raw data and we say,
- 1:54what are the different steps that a human took in order to
- 1:57achieve that next best actions? And it requires a bunch of
- 2:00inferences. We call them collections that
- 2:02say, OK, take the raw data, give me the first pain points, give
- 2:05me the second pain points, and then mirror that.
- 2:07The real unlock and the value of the reasoning graph is that it's
- 2:11able to take a process, a defined process, and provide
- 2:15those next best actions at reliable scale that are highly
- 2:17accurate that basically mirror what a human would do.
- 2:20Yeah. And and you use the important
- 2:22word memory, right? And we talked about how
- 2:25currently and a lot of enterprise AI just does not have
- 2:28that memory. And so tell us a little bit more
- 2:30about what memory is and how that supports workflows today.
- 2:35Yeah, certainly. So it's really quite simple in
- 2:38the sense that, you know, if you think about what a human does in
- 2:41any process or any workflow, we take in information from
- 2:44meetings or calls or lunches, dinners, etcetera, we synthesize
- 2:48that information specifically applying what we knew about that
- 2:51company or that meeting or that that entity.
- 2:53And then based on that, we take an action.
- 2:55And that action could vary from, as I said, a marketing use case,
- 2:58a sales use operations use case. Up until now, agentic systems
- 3:04lacked true memory. They might be able to provide,
- 3:07as you pointed out earlier in a rag database memory limited to a
- 3:10use case session or taking information and putting it into
- 3:13cloud or Gemini. And it gives you valuable
- 3:16insights into one what happened with that specific siloed piece
- 3:19of information. When we talk about memory, we're
- 3:22talking about a reasoning graph that can take all of the
- 3:24historical information that occurred between a company and,
- 3:28you know, use a sales example, everything that it's occurred in
- 3:31previous meetings, months and years of information and
- 3:34understand that those details specifically within the context
- 3:38of a process that's predefined. So I'll give you an example.
- 3:41Let's just say, take a sales use case.
- 3:43You might have a best practices that says, hey, for every
- 3:46interested customer that reaches a certain stage, we want to
- 3:49produce a proof of value document or Apoc document.
- 3:52That's a technical criteria document that outlines what the
- 3:55customer is going to be evaluating you against, what are
- 3:58the technical speeds and feeds that it expects and what are the
- 4:00outcomes. When we talk about memory and in
- 4:03the actual application of long term memory, we're talking about
- 4:06a tool or a system that's able to take historical meetings,
- 4:09emails, call recorders, CRM data, all the things that a
- 4:12seller would use and and then based on the ingestion of that
- 4:16information, update apov document.
- 4:18Not just with what was said in one call, but multiple calls.
- 4:21And using a very specific example, applying that long term
- 4:24memory in the form of competitive landmines in apov is
- 4:28a highlight what it understands in one call against competitor
- 4:32A, the agent can apply memory update apov document with a
- 4:35specific landmine that we know trips up that competition and
- 4:38then includes that in the document that the human reviews
- 4:41and then sends out. That's just one example, but
- 4:43there are many, many different examples of how the memory
- 4:45manifests itself. And if I'm hearing you
- 4:48correctly, search and retrieval in the first version of a Gen.
- 4:52AI, it was, was looked really good.
- 4:55It sounded really good and it was very conversational, very
- 4:58real, very, very realistic. But with reasoning graphs and
- 5:02memory. What it really is doing is
- 5:04protecting our institutional knowledge and making sure that
- 5:07that doesn't decay. Just like when an employee
- 5:11leaves, right? All of that institutional
- 5:12knowledge judgment goes out the window with with that employee.
- 5:16This is sort of like helping us preserve those things.
- 5:20Yeah, that's correct. I mean the idea of being able to
- 5:23onboard a new user or educate employees with all of the
- 5:27historical information is exactly what that long term
- 5:29memory provides. And and you mentioned the RAG
- 5:32data systems a couple different times.
- 5:34The space is changing so rapidly, as you know better than
- 5:37ILORA a year or two ago, producing A RAG system for data
- 5:39retrieval would have been eye opening, right?
- 5:42The speed with which it could search vast amounts of data and
- 5:45provide insights was incredible. As we touched on earlier, the
- 5:49real unlock we're searching for are when agents can take
- 5:52actions. Agents that actually execute
- 5:54actions that we as humans would otherwise do.
- 5:57And that's where the long term memory and reasoning graph
- 5:59capabilities really come in. Not just for not losing
- 6:02historical context, as you pointed out, but can you produce
- 6:05actions that I as a human find highly reliable, maybe even
- 6:08better than I would have done on my own.
- 6:10Yeah. So let's talk about let's talk
- 6:12about buying AI. And I think, I think the last
- 6:17two years have been sort of an experimentation and a lot of us
- 6:22have been, you know, very open to experimentation.
- 6:24Let's let's say vibe buying is the term here.
- 6:27And that's going out the window a little bit because there's not
- 6:29a lot of pressure to actually show how what the real ROI is of
- 6:33an AI purchase. Tell us about how that
- 6:37transition from sort of, you know, Gen.
- 6:40AIV one versus V2 here is changing and how we as buyers
- 6:45should be looking at it? Yeah.
- 6:48So I'll give you a couple different thoughts on that.
- 6:50One is that the age-old question of buyer beware has never been
- 6:55more true, right. There's so much snake oil out
- 6:57there in the AI space that the fundamental principles of the
- 7:01way buyers, technical buyers should be evaluating vendors has
- 7:04not changed. And I'll give you some specific
- 7:06examples in our world before I do that.
- 7:09The the steps that one must take that we've always provided with
- 7:13vendors has never been more true.
- 7:15Meaning you're not just looking at technical features and
- 7:18functionality, but are you taking the time to clearly
- 7:21document tangible delineated ROI that can be realized over A123
- 7:26month period and being able to objectively measure that and
- 7:30prove it has never been more important.
- 7:32So I just wanted to, to mention that because of all the claims
- 7:35that are being made in the market, it's really never been
- 7:37more important. Now from our perspective, you
- 7:39know, given the, the technology that we provide, there's two
- 7:42distinct buckets. There's, there's major
- 7:44differences in your question between SMB, mid market,
- 7:47enterprise and the enterprise. Regardless of the differences,
- 7:51the way we think about it is that they need to always be
- 7:54thinking about the capabilities beyond today's use case.
- 7:57In other words, this may solve 1 pain point today, but things are
- 8:01changing so rapidly as we've talked about daily, weekly
- 8:03basis, things are coming out. So what I'm buying, is it
- 8:07extensible? Can I take this tool and apply
- 8:09it to use case number 2 #3 #4 or is this a 1 trick pony?
- 8:13And that's a very fundamental question that needs to be
- 8:15universally applicable regardless of whether you're a
- 8:18mid market or an enterprise. And Mark, let me interrupt here
- 8:21because of what she said there. Just it just, you know, stirred
- 8:24my thought a little bit. How do you know what it's
- 8:27sensibility looks like if you can't even predict what those
- 8:30new use cases AI could do for you?
- 8:32How What are some What are some tips that you'd recommend bias
- 8:36think about? Well, well, I would say that
- 8:37that you can predict what the use cases are in the sense that
- 8:40when you evaluate a technology or the vendor is going to say,
- 8:44look, we have this underlying technology, but let me show you
- 8:46the use case that we are applying it to.
- 8:49To the extent that you can document that in very clear
- 8:51detail, that starts to become the foundation of then where it
- 8:55could lead to. Again, it still needs to prove
- 8:57its value against use case number one across the lens that
- 9:00I mentioned a second ago, but it's not an unknown.
- 9:03This comes down to your ability to sift through a lot of the
- 9:05marketing in the sales and get down to the underlying facts
- 9:08that a tool can provide, whether it's a paid pilot or APOC.
- 9:12But once that is documented and it's proven, now you can very
- 9:16quickly see, OK, if that came to fruition, now I can extend it to
- 9:20the second use case and using concrete examples, you might
- 9:24say, OK, I have a cross sell up sell use case.
- 9:26I want to apply a gentech technology and can it
- 9:28meaningfully provide uplifts in my revenue with my existing
- 9:31stall base? If that's the case, then how can
- 9:34I apply it to say an executive briefing center where my
- 9:38executives want to have kind of an FBI briefing dossier that
- 9:41captures all of the information, all of the sentiments around my
- 9:45top 20% spenders. And I can give you more
- 9:47examples. But I would just say that while
- 9:50it is very, you know, you have to be careful as you said to, to
- 9:53predict too far out. You can very get a very clear
- 9:56line of sight into what is capable, measure it, and then
- 9:59expand. Thank you.
- 10:00Thank you. Thanks for letting me interrupt.
- 10:02You were going on to talk about Yeah.
- 10:04I was simply gonna say that as it relates to kind of the buyer
- 10:08personas and the vibe buying being dead as you said in an
- 10:13experimentation failing to scale.
- 10:15I would say that you know for us we bought, we bucketed into two
- 10:18groups. We are actively looking to
- 10:20engage with VPS of sales in CRO and the SMB mid market.
- 10:24But in the enterprise space we are talking to the VPS of AI,
- 10:27the brightest minds in AI that have been grappling with a lot
- 10:30of the deployment challenges and the enterprise deployment
- 10:33challenges for the better part of three years even before you
- 10:35know chat was released. And what I would say is that you
- 10:38know the specific decision criteria.
- 10:40If you look at bucket one SMB mid market is basic, basic
- 10:44questions of all the go to market challenges I have as a
- 10:46leader, can this solution solve the highest order business
- 10:49problems for me? That's number one.
- 10:51Number 2 is what is the time to value like what are the
- 10:55deployment capabilities? What are the onboarding steps
- 10:57required? And will this be worth my time
- 11:00if these documented ROI metrics are achieved and then how
- 11:04quickly can it be measured and how objectively can it be
- 11:06measured? So those are the things that we
- 11:08think about when we're talking to the the leadership in the SMB
- 11:11mid market. Now bear with me a second as I
- 11:13pivot to the enterprise because this is where I think is the
- 11:16most meaningful change that we're seeing in the market.
- 11:18The specific decision criteria that we talked about with VPS of
- 11:22AI and enterprise are very different.
- 11:24They're saying, of course, the initial use case matters, but
- 11:27what they're looking at is how broadly applicable is the
- 11:30underlying technology and can it easily be configured to multiple
- 11:33use cases? What enterprise deployment
- 11:36capabilities does it have? Meaning, has it accounted for?
- 11:39The technical challenges that I know as AVP of AI will
- 11:42inevitably pop up when I scale beyond 100 users?
- 11:45And I can give you examples if you're interested.
- 11:47But those are the core tenants that are popping up across the
- 11:50board. It doesn't matter if you're
- 11:51NVIDIA, Palo Alto, if you're LinkedIn, if you're boxed.
- 11:54They are all wrestling with the same enterprise deployment
- 11:58capabilities. That I can address if there's
- 11:59interest. Yeah.
- 12:00So, so on the functional side, like on the SMB side, outside of
- 12:06the enterprise, you go to the functional leader and just talk
- 12:08about their functional business outcomes and whether or not we
- 12:10can we can help them there. Then enterprise side you're
- 12:14talking to like VPS of AI, like infrastructure people that are
- 12:18are thinking about how to deploy AI across the organization and
- 12:21you were going to give some examples.
- 12:23Yeah, yeah, you're, you're correct in in the summary you
- 12:25provided and I'll give you very concrete examples of the things
- 12:29that keep those people up at night in the enterprise.
- 12:32It's one thing to have a a rag database retrieval system
- 12:35across, you know, X number of users with one data source.
- 12:39The moment you start opening it up to cross functional teams,
- 12:42multiple users attaching to multiple data sources come
- 12:45across the very real issue of access level controls.
- 12:48What Laura should see based on her database access level
- 12:51controls might be different than Mark, but yet we're working on
- 12:53the same project. Same thing can apply to
- 12:55leadership. So who should see what and when
- 12:58the agent is offering up something that is a non trivial
- 13:01task that all of these VPS of AI are are dealing with compounding
- 13:05knowledge that spans weeks and months.
- 13:07So does it incorporate its understanding of the past in a
- 13:11useful and reliable way? This is where the memory comes
- 13:13into play. Data size and complexity, does
- 13:16it have an explosion in file counts?
- 13:18Are we dealing with duplication all over the place?
- 13:20Is my ultimately that underlying data exploding in unuseful ways
- 13:24that might be prohibitive and then contradictory data like how
- 13:28do does it handle the flood of information that the agents are
- 13:31accessing and producing? And how does it handle override
- 13:34logic? These are the things that the
- 13:36VPS of AI at NVIDIA and Palo Alto are dealing with, and
- 13:39they're looking for underlying technology that is already
- 13:42addressed that or is building it in meaningful, thoughtful ways.
- 13:45Yeah. And you know, these are some
- 13:46like very sophisticated buyers that that, that I'm hearing
- 13:50about, right? Like they're, they're very
- 13:51educated and they like they know what's up.
- 13:54How? How do you think selling has
- 13:56evolved when we are interacting with such sophisticated buyers?
- 14:02So as a lifelong technologist, I actually don't think it has
- 14:05evolved that much meaning. Back to the original comment
- 14:09that I I mentioned to your question.
- 14:12Buyer beware has never been more true technical buyers. 20 years
- 14:15ago when I was working at Oracle or at couch base they always had
- 14:19the same approach. I believe what you're saying but
- 14:22I need to verify it in Apoc. You're correct.
- 14:25The VP's of AI, NVIDIA and Cisco and Box and LinkedIn.
- 14:28These are the sharpest brightest minds of AI.
- 14:30They are cutting edge. They know exactly what the
- 14:32market offers. It comes down to age-old best
- 14:36practices in a technical criteria, in a technical Bake
- 14:39Off, clear, very delineated technical features that need to
- 14:43be addressed and then the associated business outcomes.
- 14:46These people are no different and if anything, they are more
- 14:48acutely aware of the pains that they've been through the last
- 14:51couple years looking at different vendors and that
- 14:54process needs to be in lock step with the vendor without
- 14:57question. And when, when you're
- 15:00implementing or you're, or you're buying something like
- 15:02this for the enterprise where you were, you thinking about
- 15:05things like, you know, access control and, and things that are
- 15:07like sort of fundamental to how an organization runs.
- 15:10What are some ways that we build the ROI for the CFO?
- 15:16How do, how do we, how do we even think about quantifying
- 15:20this? Yeah.
- 15:21I'll give you an example in sales and there's other examples
- 15:24that we are being pulled into specifically in marketing and in
- 15:27post sale. But let's just use sales as an
- 15:29example. If you can go to any leader in
- 15:33any enterprise and say I can lift up your revenue by 5% or I
- 15:37can increase your deal velocity by 10% with the same headcount
- 15:42without hiring new people, you have their attention but it
- 15:44needs to be proven. So whether it's in the SMB mid
- 15:47market or enterprise, what we're seeing are POC's that center
- 15:51around exactly that leveraging agentic technology in our case
- 15:56with underlying, you know, reasoning graph and long term
- 15:58memory and being able to automate in large part what
- 16:01humans would otherwise take hours or days to do.
- 16:04It's done in a fraction of a minute or it's done in a handful
- 16:07of minutes, freeing those people up to go chase more deals or go
- 16:12basically tasked them with higher order value processes
- 16:16that agents quite can't do yet or rely upon human to human
- 16:19interaction. So back to your core question,
- 16:21it really comes down to in the case of sales, take a 90 day
- 16:25period pre deployment and walk me through the metrics that your
- 16:29team with X number of of sellers can accomplish.
- 16:32Let me compare that against the next 90 days where the
- 16:36technology is installed and you should be able to show very
- 16:39objective measurements across deal velocity, revenue uptick,
- 16:43etcetera. That's where we're seeing
- 16:45customers saying, wow, this is objective, I can take this to
- 16:48leadership. It's the tool has paid for
- 16:50itself in a matter of weeks. Yeah.
- 16:52And and would you say that that's also the same approach
- 16:56regardless of whether you're going to a functional unit
- 16:58versus for example, VPFAI, like how did they, how did they
- 17:01assess ROI? It's a great question.
- 17:04So the ROI that I mentioned a second ago with the VPS of
- 17:07business is very objective in the manner that I just said.
- 17:10However, when you talk to the VPS of AI, these people are so
- 17:14advanced that they are less looking at ROI per SE, but
- 17:18they're looking about the things that I mentioned a few minutes
- 17:21ago, which is what is the extensibility and
- 17:23configurability of the platform. Because remind you, these are
- 17:26stakeholders who have 10B US, they have to support and they
- 17:30know based on their interactions that each of those different
- 17:33BU's has very different workflows, very different
- 17:35requirements. So if you're selling me a tool
- 17:38that's purpose built for use case, one that has trouble going
- 17:40beyond that, they're thinking along those lines.
- 17:42But if it's configurable and it's easily configurable to
- 17:46different use cases that span what you initially showed me,
- 17:48that's the lens with which they're looking at it along with
- 17:51the things that I mentioned before, access level control,
- 17:54duplication, the inherent logic that the system provides, that's
- 17:57the lens. There is no ROI to cut to the
- 17:59chase, Lauren, on that. But looking about it in breadth
- 18:02of functionality specifically within enterprise challenges,
- 18:05OK. OK, very cool.
- 18:07Awesome. OK.
- 18:08So if if we are an enterprise Oregon, an organization that
- 18:13already has, you know, adopted AI and we're not in this like
- 18:18Gen. 2 version of it, you know, V2 version of it, what do you
- 18:21think are some ways that an organization can very easily
- 18:24shift or maybe it's a step by step approach to, you know,
- 18:28moving from right to reasoning graphs?
- 18:32Well, I would say that it depends on the definition of a
- 18:35company being quote, UN quote AI adopted.
- 18:37And what what we found candidly is that a lot of the quote UN
- 18:42quote AI that people are using are what I'll just say insights
- 18:45driven. So, you know, using and I won't
- 18:48name technologies, but using different technologies from big
- 18:50vendors and small vendors alike, they are able to provide real
- 18:54value in siloed ways, synthesizing calls or
- 18:58synthesizing CRM or e-mail and providing you information.
- 19:02I think that's a a very it's a good start.
- 19:05But when you say like, how do we take that step if we're already
- 19:07there to this, it really goes back to what we talked about is
- 19:11let's let's put the fluff aside and give me a workflow that
- 19:15provides value to your company. Like let's not take a a use case
- 19:18or a process that's off to the side.
- 19:20If unlocked, if automated, would this be the number one or #2 or
- 19:24#3 priority for you? And let us prove over a 6090 day
- 19:28period that the vast majority that can be not only automated,
- 19:32but specifically mirrored with the next best actions that your
- 19:36top 1% employees would take. And I can explain more about
- 19:39that. So it's not just the automation,
- 19:41but it's really mirroring the behavior that leadership wants
- 19:43to see across their workforce. Right, right.
- 19:46OK, awesome Mark, that was very helpful and insightful.
- 19:49Let's learn a little bit more about you.
- 19:51You called yourself a lifelong technologist and you've been CRO
- 19:55for a number of different companies, AI and non AI
- 19:58companies, but still technology. What's been most interesting for
- 20:01you, I think in the shift? Yeah, it's a great question.
- 20:05So in the majority of my career selling as you said, database
- 20:09technologies, infrastructure and being a CROA lot of the sales
- 20:13and kind of a lot of the approach I've taken has been
- 20:16around technologies that have already existed.
- 20:18So a database technology or analytics technology, etc.
- 20:22Much like we've seen with other technical waves, other technical
- 20:25revolutions, the biggest shift that I've seen goes back to what
- 20:29a lot of the the Deloitte, the accentures, the PWCS of the
- 20:33world specialize in and that's change management.
- 20:35You mentioned a second ago, hey, how would you, you know,
- 20:38recommend to a company that's already had AIVI one go to to
- 20:41V2? It is a fundamental change in
- 20:44the way people think. It's one thing for a company to
- 20:47get on a call with us and say, yes, we're, we're interested in
- 20:49in applying A gentech technology.
- 20:51But when you really get into the weeds with them, we're talking
- 20:53about changing the way they think and changing the way they
- 20:56allocate resources. And so one of the biggest things
- 20:58that I've observed is there's a very clear line of delineation
- 21:02between companies that say they want to and companies that are
- 21:04functionally, organizationally ready to make the changes,
- 21:08allocation of resources that is necessary to adopt.
- 21:11A lot of what we've talked about today, that's nothing new, but
- 21:14it's never been more true. Yeah.
- 21:16And and as far as change management goes, how, how are
- 21:19you seeing that change management within your internal
- 21:22sales teams? Like, have there been things
- 21:24that you're doing differently now that you're selling in such
- 21:27a fast, fast-paced environment with your teams?
- 21:30Yeah, that's a great question. So, you know, I'm a little long
- 21:32in the tooth as a seller, right. And, and so for the longest time
- 21:35I've operated out of spreadsheets, I've operated out
- 21:38of To Do List on paper. And I'll be honest, when we
- 21:41first released our own tool dog food, I didn't use it as much as
- 21:45I, I probably should have. And that's because of the 20
- 21:48years of, you know, having a, a process in a method that works
- 21:51for me. And so slowly over time,
- 21:54realizing that the more I give the agentic technology
- 21:58information and feed it, the better the outcomes are going to
- 22:00be for me and the faster I'm going to be.
- 22:02And so today I am highly, highly reliant upon every call that I
- 22:07take, every sales cycle I go through my tool because it has
- 22:10made me exponentially faster and exponentially more productive.
- 22:14But that doesn't happen overnight.
- 22:15And so the way I would answer that is, you know, the change
- 22:18management concept that we talked about, at least within
- 22:20the sales ecosystem takes time for sellers who have been doing
- 22:24it for a long time. If you're earlier in your career
- 22:26and you're AI native, perhaps that doesn't apply.
- 22:30They're ready to go. This is all they know.
- 22:31And so it's very exciting to kind of work with different
- 22:34layers of experience in in the enterprises in the S&B mid
- 22:37market because you come across a wide variety.
- 22:40And has that written something different about how you are now
- 22:43assessing candidates because of this?
- 22:46That is a great question. The truth is that we haven't yet
- 22:49begun to scale, but that's going to change here in the in the
- 22:51subsequent months. But I have ideas on the way that
- 22:54I'm going to assess them. And lifelong practices of using
- 22:57actual projects or actual homework have never been more
- 23:00applicable. Much like we talked about buyer
- 23:02beware having, you know, very defined POV use cases having
- 23:05very defined outcomes. The same is true here where you
- 23:09want to be able to give candidates very easy tech
- 23:13projects in the form of the interview process and say go
- 23:16ahead and apply AI to this workflow and let me see what you
- 23:19produce. And you're gonna get very
- 23:21insightful results as a result of using that.
- 23:25OK. Well, to all listeners who are
- 23:26looking for a job in sales, they'll come find you a little
- 23:31bit then. Cool.
- 23:32Mark, thank you so much for the conversation.
- 23:34Anything else that you'd like to leave the audience with before
- 23:36we wrap up here? No, I appreciate it.
- 23:38It's a very exciting, fast changing time and I'm I'm
- 23:41pleasure to be on your podcast. Fantastic.
- 23:43Thank you so much. Thank you listeners.
- 23:46If you liked it, please drop us a a note like us and otherwise.
- 23:50Until then, we'll see you next time.
- 23:51Bye bye.