3534: Agentic AI at Scale: What 120 Million Monthly Conversations Really Mean

[00:00:04] Over the last 12 months, almost every conversation around AI has circled around one question. How do you move from experimentation to interproduction and producing real value and outcomes? Outcomes that actually work inside complex organisations? Well, my guest today, he's someone who has lived that challenge from multiple angles.

[00:00:29] His name's Joe Kim, he's President and CEO of Druid AI, and he joined me from Austin while I was here in the UK. And Joe has led companies across security, observability and enterprise software. But today, he finds himself at the centre of the agentic AI conversation, as enterprises look to move beyond pilots and into production.

[00:00:53] So we will talk about what separates conversational AI from agentic AI, why orchestration and workflow design matter way more than hype, and how at Druid AI they're scaling to more than a billion enterprise conversations while still delivering measurable ROI. So we'll talk about the education, healthcare and public sector use cases to bring that to life,

[00:01:15] and also explore why trust, compliance and resilience should always remain non-negotiable as AI becomes embedded into your everyday processes. So if you're curious about where enterprise AI is genuinely delivering value today, this is a conversation you are going to want to hear. Here at the Tech Talks Network, we now have nine podcasts and approaching 4,000 interviews.

[00:01:41] And that is only possible with some of the great friendships that I've developed over 10 years of podcasting. And a company that I'm proud to call friends of the show is Denodo, because not only have they been on this podcast multiple times, they also help make sense of the AI data chaos that we're seeing now. Because the data world is louder than ever. AI hype, lake house complexity and pressure to deliver more with less. These are things that I talk about every day on this show. But Denodo is helping businesses make sense of it all.

[00:02:11] Because they provide a unified data foundation for trustworthy AI, lake house optimization and data products to finally bring service to life. So whether you are a CIO or a builder, Denodo helps you activate your data with speed and governance. And their global partner network also helps you accelerate every step of the way. So if you're ready to unlock real outcomes, simply visit denodo.com today.

[00:02:38] But now it's time for today's interview. Let me introduce you to today's guest. So a massive warm welcome to the show, Joe. Can you tell everyone listening a little about who you are and what you do? Sounds great. Thanks for having me, Neil. My name is Joe Kim, and I'm the president and CEO at Druid AI. Fantastic. Well, thank you for joining me on the podcast today.

[00:03:05] I've got to ask, how has your experience leading large SaaS and enterprise technology companies throughout your career? How has that helped shape your direction to getting Druid AI as it enters its next phase of growth? Because you've been on quite a journey here, haven't you? Tell me more about that. Sure. I mean, I'm pretty much a newbie at this point at Druid, right? I think I'm at this point a little over 100 days in is what I would say.

[00:03:32] It probably feels like for the team and everyone else that I've been here for years and years. But, you know, I can still count it in days instead of years. But before joining, obviously, I've been in a lot of, you know, other big companies. Right before Druid, I was the CEO at Sumo Logic, which is a very large security and observability company. And then before that, I've been in tiny little startups, you know, as well as huge companies like General Electric,

[00:04:00] HP, and Citrix are some of the larger organizations that I've been at. And obviously, I've learned, you know, quite a bit in terms of what it means for scale and everything else that I would bring over to Druid. With all of that said, the reason why I picked Druid and why I was so excited is there are two very specific things that is really, really hard to change whenever you go into helping to lead a company. The first piece is around, you know, how's the market dynamics look like? Right?

[00:04:29] As good as your product may be, as much as you're helping customers. There's very little that you can do if the market dynamics are weak or if it's a long tail. Well, you know, obviously, you can continue to build product, but it's hard to really move the business. And obviously, for Druid, in terms of where we play with the Gentic AI, the market dynamics are super dynamic and incredible and lots and lots of tailwinds. So that's one consideration that I picked.

[00:04:56] And then the second piece is in terms of the moat for the company or the differentiation around the intellectual property or, you know, how old the technology is compared to, you know, newer players. A lot of that is really hard to change as well, you know, when you go in and start leading a company. And so from that perspective, you know, I did a lot of research on my end in terms of Druid's differentiation in this space compared to others.

[00:05:23] And, you know, as I was dwindling down the list, ultimately, it wasn't even close. I mean, in terms of what Druid had, in terms of its capability, was way, way further along than most competitors in the market. And for people listening, hearing about Druid AI for the first time, you are known for helping enterprises work smarter with human-centric agent-based AI, an AI that actually works as well. You're also powered by the Druid conductor.

[00:05:50] But tell me a little bit more about that, just for people that are listening, that are hearing about you guys for the first time. Sure. I think in terms of the market today, it's pretty bifurcated in terms of the type of value that customers are looking for. I would say on one side with a lot of the AI technologies, you know, people are looking for capabilities around generative AI, which is around generating new content like videos or audio or documentation or something like that.

[00:06:20] And I think predominantly a lot of that stuff is played with by the big, big players like OpenAI, Anthropic, and, you know, and the like. You know, I'll throw Microsoft in there as well with Copilot. And then there's this other side of the universe that's becoming more prevalent now around agentic AI. It's how do I utilize all those technologies to help make decisions and take action?

[00:06:44] And in that side of the universe, the dynamics, although you might be using similar technologies, change quite a bit around how do I get the accuracy of my answers and the decisions higher and higher and higher instead of the 80s, maybe getting into the mid to the upper 90s.

[00:07:00] And then from a security and compliance perspective, because a lot of the non-generative work, the actions and the decisioning part for enterprises mean that you need super high levels of security and compliance so that you can make sure that you are protecting data, you know, for your customers, you know, et cetera. And so for Druid, we don't really play in the generative space. We use a lot of those technologies. We're really a platform that you utilize to be able to do agentic capabilities for enterprises.

[00:07:29] And before you join me on the podcast, it's a great time for you. I was also reading how you've recently closed a $31 million Series C round. So what does that investment enable you to accelerate that was possibly harder to deliver before? Yeah, I think, you know, previous to me joining, the team has done an incredible job building up the technology and adding customers bit by bit, I think, over time.

[00:07:54] We've gotten big enough at this point where if you go out there and look at, you know, even something like, you know, the Gardner Peer Insights, you know, it's not the Gardner Magic Quadrant per se. It is actual customers going out there and rating what the product does, et cetera. We've actually been able to claw our way into a place where we have more reviews than anyone else of our competitors. And more than that, our scores are incredibly high. They're higher than everyone else. We're darn near, you know, five stars across the board.

[00:08:24] And so I think with the limited budgeting that we had, we did a really good job building the products and really helping our customers up until this point. What I'm looking to do with the money that we raised, Neil, is really, you know, get the news out there as quickly as possible to the broader enterprises, more specifically in North America, because we have a good presence in EMEA. But in North America, people are just starting to figure us out. So I want to make sure that we hit the accelerator button, certainly there.

[00:08:53] And it's partially, you know, certainly everyone wants their businesses to grow and become bigger. But for me personally, I'm trying to get out there to really educate and put the product in front of customers more quickly than not, because so many people are wasting money right now trying to build this stuff on their own. And I think it was around about exactly a year ago that Gardner predicted that 2025 would be all about agentic AI. Since then, every tech conferences I've been to this year talks about it at scale.

[00:09:23] There has been a few stories of businesses not being able to secure AI from some of those projects. But to bring to life a positive story here, I was reading that Druid AI has delivered more than 1 billion conversations across thousands of agents last year. So when you look at a volume like that, where do you see the clearest proof that agentic systems are ready for enterprise scale? Because, again, a phenomenal figure. But what does that mean to you?

[00:09:51] I think, you know, for us, it's certainly a feather on our cap more than anything else, Neil, when you think about it. And even since we've broadcasted there were over a billion conversations per year, we're already maybe 40% to 50% above that even as we speak today. So we're doing, I think the last count was over 120 million conversations per month. And that would obviously put us, you know, 40 to 50% of where we are.

[00:10:19] And it will continue to grow is what you will find. Ultimately, though, for each of the specific verticals, the value is going to be something quite different. It's going to lead to more conversations. It's sort of the way that we keep track. But some of the leading indicators that we see is, you know, either going to be one of two things in terms of our value.

[00:10:40] If you think about agentic AI and sort of what we help with an existing business process, as an example today, anything that we use our technology to be able to automatically fulfill that specific transaction. So, like, I don't know, let's use collections as an example. Every business need to collect money from folks. You need to collect money from folks, you know, as well.

[00:11:03] And when you think about that transaction, if you're using our capabilities to automatically make decisions and make the collections process end-to-end as automagical as possible, then a lot of the figures that you're going to see on the ROI is going to be around efficiency. How much more money or time can I take away from the existing process?

[00:11:23] We also see on the other end of the spectrum, not automating the process, but how do we interact with our users, the end customers, so that that process is executed at the highest speed and degree as possible. Okay. And what we're seeing in terms of how customers find that is through what we call conversational AI, where you are making the consistent transaction execution possible.

[00:11:51] But the way that you might interact with that transaction and the way that I may interact with that transaction becomes super unique for each user so that it's not just like press one, press two, press three, et cetera, but it's more conversational. What we're finding is that actually helps with NPS and the ability to execute that transaction better for the business ultimately. And a lot of those things actually lead to top-line growth, top-line growth.

[00:12:16] So for a university, as an example, in the admissions process, the better that you can actually quickly go through that transaction means that you can have more and more students come in. In a hospital, it will be the same thing. The more that I can schedule patients appropriately for the right surgeries or the right follow-ups, you're going to end up in a situation where it actually increases your top-line.

[00:12:40] But ultimately, for us, we know on one side or the other, the efficiency of the top-line, we know that it's being effective as we see the number of conversations spike. And you mentioned universities there, and I read that you described outdated campus technology as one of the root causes for student churn because they've got a different expectation now. But what patterns are you seeing across education that led you to call this out as an urgent problem for institutions?

[00:13:09] Because it really is, isn't it? And students now do have much, much higher expectations with the tools that they're using at home. Yeah, I mean, I think when you look at the spectrum of different industries that are being hit by sort of the disruption caused by AI, it's not equal across the board. I think certain types of institutions or, you know, types of businesses are hit harder than others.

[00:13:35] And there's many, many of them, not just higher education, but higher education certainly is one of them. I'm sure, you know, like, you know, in my case and when I talk to, you know, my, you know, customers as well, there are students and kids out there using ChatGPT and all these capabilities each and every single day. Sometimes for good purposes, sometimes maybe not so good purposes, but it is hitting education incredibly, incredibly hard.

[00:14:03] And so, and what that does mean because of the dynamics of education will change tomorrow with all these technologies is that there's going to be, you know, possibly even like fewer and fewer students. If you think about the demographic, you know, kind of demographic and everything else of the number of students that's going to be attending university just by a sheer, you know, number of people that's going to be available.

[00:14:29] In my case, I would say total addressable market, but the TAM is going to be shrinking. So that means that you have to fight harder for each and every student. And it really is on two parts. I think one part is on the admissions and enrollment part. That's how the revenue ends up initially coming in. And, you know, I touched upon that a little bit. The more unique that you can make that experience and easier that you can make it for a student, they're going to be more likely to be joining the university.

[00:14:55] The other side is, and it's funny, it turned out this way, is the collections. For many, many reasons, when they're going from semester to semester, maybe it's financial aid or there was a mistake with a building process or something like that. But what we found is as soon as you don't interact with that student in their unique situation enough for you to be able to collect the dollars, because they might not want to drop out. That's not the case. They might be waiting on additional funds to come in.

[00:15:22] And if you don't know what is happening and you accidentally drop that student, it is highly likely that they're either going to switch schools or they're just going to drop out. Because trying to re-enroll all back to the, you know, the right classes and everything else, because you've got fewer choices by that point, it's incredibly difficult for the student. And that's what we're finding. And also when I was doing some research, I was reading that Druid AI has grown APR 2.7 times year old. Yay!

[00:15:52] And also expanded your partner ecosystem with some pretty big names. I'll read some of them out here. Microsoft, GemPact, Cognizant, Accenture. I mean, what does all this tell you, along with the partnerships, about where enterprises want conversational and agentic AI to go next? What are you learning here? I think what we're learning is, you know, initially there was like huge excitement around agentic AI and conversational AI platforms.

[00:16:19] And, you know, between the partners, because they're dealing with so many different enterprises that are out there, they're seeing like literally infinite amount of possibilities in terms of how you could utilize this technology. Again, both for top line and for efficiency purposes. What we've learned over time, and you've seen a lot of studies out there where this initial excitement is starting to kind of peter out a little bit.

[00:16:42] You know, there was a study from MIT saying that, you know, 95% of, you know, generative AI or AI related technologies aren't delivering ROI is what people are finding. But there is a renewed energy when we go talk to our partners. We're actually even adding more partnerships than what we have right now because of the renewed energy. It's because it's not just like people aren't able to find ROI.

[00:17:05] If they've spent all of that money to do a lot of these technologies, they've actually learned a ton of new lessons in terms of where are the applicable areas for you to get ROI very, very quickly in these. And so we've only seen partnerships ramp up and get much more specific in terms of the value that we deliver. The good news, I think, on the Druid side, and this is the Gartner Peer Insights, why our scores are so high and why we have so many people leaving feedback for us,

[00:17:32] is partially because we've been in the journey with our customers to make sure from day one that we're identifying use cases that is going to bring value to the customer. Like we've always put that in focus so much so in the last symbiosis in New York City that we had held, actually sort of the thematic thing that we had pitched was the 5% club. You know, not the 95%, but the 5% that's delivering value.

[00:17:58] A lot of the folks in the room, they're able to illustrate where their 5% is coming from. So we find that starting next year, you're going to see ROI actually spike with a lot of these technologies. Oh, interesting. And Druid AI also earned the challenger spot in the Gartner Magic Quadrant on the Stational AI platforms for 2025.

[00:18:20] I mean, how do you interpret that placement in terms of market maturity, expectations for buyers, and indeed the opportunity ahead? Has it attracted a lot of attention your way as well? It has. It's been a good marketing material for us, for sure. You know, when we look at it, we just see it again as a feather in our cap, you know, in terms of our size and, you know, how we're impacting the industry today. Okay. But it also does mean kind of one, you know, two things.

[00:18:49] I think one is, because typically how those MQs work is the kind of the up and down vertical is more around the execution. It means that we're executing really well. If we're at the challenger side, you know, basically at the top of the challenger section above some very, very large companies. I don't want to call anyone out, but across very large established companies. But it also does mean that if we want to get into the leadership quadrant, that we have to be able to share more of the thought leadership that we have.

[00:19:19] Okay. I mean, we got, you know, folks that are in the company, Andrea, for example, the co-founder and COO has literally a PhD in Agentec AI that she wrote 10 years ago, right? Like way before this stuff was hot. So we have really good thought leadership that we have to be a bit bold about sharing with others. And you'll certainly see us doing that moving forward. We've both mentioned today that many enterprises have invested in conversational AI without seeing measurable outcomes.

[00:19:46] And obviously you're in that 5% of people that are. So what have you learned about orchestration, integration and workflow design that is helping teams move from pilots and into production and delivering real business value? I think there's kind of two things that would point out to that, Neil. I think one of them is, you know, outside of the initial super excitement around AI technologies, because it seems so easy for you to, you know, be able to get value out.

[00:20:16] And what I think is, you know, is that for non-production, more like B2C kind of use cases around generative AI. I think it is true. I think the excitement is real and people should be using it every single day. I certainly use it every single day. And there's a lot of value that comes from that. The thing that people are learning now is if you want to do this for an enterprise, it is actually a way more complicated implementation if you are trying to do this on your own. Yeah. Yeah.

[00:20:45] And the reason why I feel that our customers have seen such successes compared to others is because it just happened that they're partnering with people that are really looking out for the customer's good and making sure that we are doing everything possible to make sure that they are successful. Internally, I have this concept around making sure that one of our biggest missions is to make sure that our customers are the heroes for their company.

[00:21:13] And that is reverberating across the entirety of the company in every single department to make sure that we're doing what we can to make sure that they are seen in the best light possible as they're starting their AI journey. So that's sort of kind of bucket number one is sort of the culture and the partnership. I think the second part that's really been helpful is, you know, to be frank, there's some real big bets when the company first started about, you know, now going on eight years very shortly here, Neil.

[00:21:41] That if you think back at it now, it's like, of course, you would build it that way. But eight years ago, it was not so straightforward that you should build it that way. And the bets that the technology team made all worked out really well in our favor. And it is a lot of the complications that a customer may need to go through is now built into the platform. So I'll give you just a snippet of how complicated things could get if they were to try to do it on their own.

[00:22:07] So, Neil, let's say, you know, there was some kind of conversational AI agent that we built and you are interacting with the agent, Neil. So if you went in there and you asked the question, it is not uncommon for the Druid platform in the background to go through 55 different subroutines and checks before it gives you the first answer. Okay. So it's anticipating who you are, what kind of questions you may ask in the question.

[00:22:35] It will then, you know, compare against like every other person that might have asked a similar question. You know, what are they looking for? And then it also starts to anticipate what is the next question that you're going to be asking all before the first answer gets done. And so, like, if you're trying to build out a platform from scratch that does that for you, it is going to be a pretty big uphill battle for you to start building that out. And so they're just from a technical point of view, the team's been building for eight years.

[00:23:05] That gives us a big, big unfair advantage right now in terms of how we can deliver value to customers. And as you said, it was eight years ago. Things weren't so straightforward. And you're also one that has spent years working across product security, observability and infrastructure. I'm curious, if you were to look back now, how do those lessons that you may have learned back then influence the way you think about trust, safety and resilience in agentic AI platforms?

[00:23:33] Are there any join up the dots kind of moments when you look back? Yeah. You know, for me, for me, when I look back, it's it's incredible to see like just the kind of from a big animal picture perspective, Neil, the things in security in terms of, you know, what you need to protect are, you know, it hasn't changed in a long time, you know, since, you know, digital technologies and capabilities have existed.

[00:23:57] The the fine tooth comb of like the each of the the technical components when you go deeper and deeper and deeper may change. But ultimately, you have to be able to fulfill some of the promises that are going to be at the big animal picture level. So be it like protecting identity, you know, you're you're sitting in, you know, in a different part of the region than I am here in North America. So like in your region, things like GDP are going to be very, very important.

[00:24:24] And it doesn't matter if you're using AI technology or not. You have to be able to be compliant to GDPR. And for each individual, if they want to be able to say, hey, I want you to remove data that you may have of me. Unless you have certain control points, you may not be able to do that using AI technology, but you're not going to be able to circumvent things like that. And so, you know, that's how I feel about it is a lot of those lessons are still the same lessons that you need to be able to follow.

[00:24:53] But make sure that from a technical component and technical capability that you have all the sophistication to be able to handle it for customers. Love that. It's a great moment to end on. But before I do let you go for anyone listening, wanted to dig a little bit deeper on anything we talked about today. Learn more about all things Druid AI and the work that you're doing or just keep up to speed with some of the announcements on the horizon. Where would you like to point everyone? I would just say go to the website at druidai.com.

[00:25:23] We can update the website with a lot of the capabilities that are getting released. Again, like the two that I would like to highlight where customers, if they go to the website right now that I would look at, that's super interesting and I'm personally excited about, is one, we just launched a marketplace on the website. So if you go to the marketplace tab on druidai.com, you're going to be able to see all of the pre-canned conversational AI solutions that you can get out of the box already.

[00:25:53] Okay. So if you don't want to do your own development, you want to see what are the capabilities that you can just get from us or some of our partners. You can certainly go there. The second piece that I'm super excited to see being released is something that we call the authoring agent. Okay. And the authoring agent basically is a conversational AI agent that helps you build other agents by just speaking to it. It's super cool stuff.

[00:26:19] So there's a lot of demonstrations and things like that really kind of on our hero page, but there is a lot of videos and things like that, that you can look at that will give you a bit of indication of where we're moving forward in terms of our technology. So still best place to go is druidai.com. Awesome. Wow. I'll be adding links to everything to make that easier for people.

[00:26:40] And I would urge people to check out that Druid conductor we've been talking about and how you manage intelligent agents that simplify processes, enhance interactions and produce measurable business impact with enterprise AI. It's incredible what you're doing. I'm going to be staying in touch and maybe get you on next year, see how things are evolving. But more than anything, thank you for shining a light on this today. Appreciate it. Thank you for your time today, Neil.

[00:27:06] So a big thank you to Joe for being so grounded in this conversation, especially at a time where so much of the AI discussion is driven by noise rather than outcomes and deliverables. And what stood out to me was their focus on orchestration, trust and making customers the heroes inside their own organisation.

[00:27:24] And whether it is higher education, healthcare or enterprise operations, it's clear that agentic AI will only work when it's designed around real workflows rather than abstract promises or chasing after the next shiny thing. So for everyone listening, I'll add Druid AI, the marketplace and the tools Joe mentioned in links to the show notes so you can find everything there. And if this episode helped you think differently about conversational and agentic AI in the enterprise, as always, love to hear your thoughts.

[00:27:54] Tech Talks Network dot com. But that is it for today. We're out of time, I'm afraid, but I'll return again bright and early tomorrow. Speak with you all then. Bye for now. Bye for now.