·
9 minute read
Summary
On this episode of Direct from the Expo, The Collab Collective's Craig Durr talks with Erik DeGiorgi, CEO of NetSpeek, from the floor of InfoComm 2025, North America’s largest Pro AV trade show. Erik discusses how NetSpeek’s AI platform, Lena, is solving key operational challenges in multi-vendor UC and Pro AV environments. The conversation explores NetSpeek’s launch strategy, Fleet Partners, and how Lena delivers secure, enterprise-grade AI without relying on public LLMs.
Their discussion covers:
- What is NetSpeek and Who is Lena: A purpose-built, generative agentic AI platform designed to manage UC, AV, and digital signage environments
- Solving Multi-Vendor Complexity: How Lena centralizes administration across disparate systems and devices
- Fleet Partners: NetSpeek’s strategic initiative with vendors like Aurora, Barco, BrightSign, NeatFrame, HewlettPackard, Sennheiser, and Skykit to create AI-specialized agents
- Operational Automation and Security: How Lena automates real-time room remediation and protects sensitive data with in-house AI models
Listen to the Audio:
Or tune in on your preferred streaming platform:
Transcript
Craig Durr: Everyone, this is Craig Durr, Chief Analyst and Founder of the Collab Collective, and I want to welcome you to another edition of Direct from the Expo–InfoComm 2025. This is one of my favorite shows. It's the largest Pro AV show in North America. And talking about that market, it is great to catch up with a good friend, Erik DiGiorgi, who's the CEO of NetSpeek.
And we're going to talk about what you've been up to here. We actually had a chance to sit down in March when NetSpeek first made itself known to the world. It was a great coming out party at this point in time, but you've been busy since then, right?
Erik DeGiorgi: We've been busy. Time flies. So we're really excited. A lot's happened since then. We are demoing here at InfoComm–the full platform. As soon as we get back, we're moving into beta. We have four betas lined up and greenlit. Ready to go. We've got a Fortune 500, financial services company, we've got a biotech company, and a couple others. And then, we're targeting GA in the Q3 here this year.
Craig Durr: That's fantastic. So from March to Q3, we've seen a lot of momentum. So we're talking about this wonderful product being available. But let's take a step back and make sure the audience understands what's going on. Talk to me about what NetSpeek is and who Lena is.
Erik DeGiorgi: So NetSpeek is bringing to market the first, generative agentic AI platform, built specifically to support the day to day operations and administration of UC, Pro AV, digital signage spaces–all those technologies.
What we've built is an AI that is trained on those platforms. It has a knowledge base, it’s integrated with those devices, can observe and control them, and can perform a lot of the day-to-day tasks, especially the onerous tasks that people are struggling to do, making sure meeting spaces are functioning, making sure things are up to date and the equipment is up to date.
So we've built a platform that can be inserted into these networks, and connect to all of the technologies that are on the networks and collaboration spaces, and take ownership and responsibility of the day to day operations.
You mentioned Lena. Lena is our flagship product. We really personify Lena, which stands for Language Enabled Network Administrator. I mean, it's kind of the origin story of NetSpeek. The reason it's called NetSpeek is because it's all natural language driven. So in addition to all the automations and things that we're showing here and we're going to be delivering as part of GA, it's really meant to give you voice-enabled natural language experience just like you would with other consumer products that you're accustomed to using.
Craig Durr: And because that's the modality that people are really leaning into, especially with AI, which is fantastic. Now, the interesting thing you set up, you're talking about serving products hardwares across UC, AV, digital signage. That's the reality of most environments–multi-vendor environments right now create complexity and challenges with managing, but you're able to solve that is what you're saying.
Erik DeGiorgi: Yeah, these environments are inherently multi-vendor. I mean, think about a conference room. You've got a display there, you've got a UC system, you might have a room control system in the lobby or foyer, you might have a digital signage set up. So all of these technologies need to coexist in the same network.
And there's actually part of, you know, going years back, the kind of genesis of NetSpeek was people within the industry saying, “I have too much technology. It's disparate, it's not connected. There's no centralized system.” So even before the AI component, we were looking at how to solve that problem. Right then, the AI component came into it, and it really became like, “Hey, we can actually build the network operator. We can build an assistant for these people that are struggling with those spaces.” So it really evolved quickly. And here we are today.
Craig Durr: I have to say, look, I know you come as a product executive back in the digital science space. You know about the complexities of multi-vendor environments. So that's really key. But I notice that you guys have announced something else, which is some really deep integration with a certain list of partners. I think you refer to it as a fleet partner.
Erik DeGiorgi: Yeah, so our main announcements here are our agent fleet called Fleet Partners. Agent fleets are basically specialists on vendor technologies that are in our probably Pro AV and UC ecosystem. So I'm going to make sure and try to get the list right. But we have eight major vendors that we're launching with here: Aurora, Barco, BrightSign, Neat, HP Poly, Sennheiser, Skykit. And then, there are many more that we're actively working with that we're going to be announcing.
Craig Durr: That's fantastic. So the way I understand it is this: I might have the equivalent of a deep knowledge specialist within those ecosystems. So I have HP within my environment… HP, I would have an on hand specialist.
Erik DeGiorgi: That's correct. So there's kind of four tenets to what makes a partnership–there's training, there's awareness, there's observability, and control. And we break that down real quick.
So training is… we work with the vendor partner to ingest all of their technical documentation, all of their resources, and we train a specialized AI agent that is a specialist, just as you're saying. So HP Poly, we have a specialist built for UC applications that understands and is trained on all the technologies from that vendor in the space.
Awareness–this is a really cool part. So our platform it's not using generic LLMS or generic AI. Everything's built into our cloud. I have a background building enterprise SAS, so you know, I've sold to the largest and most challenging customers to sell to. So we took that as a starting point from the beginning. Security is huge for us, and we’ll probably get into that.
But our platform does not communicate with the internet. When you ask Lena a question, Lena is not going to do an internet search. It's only based on the training that we have ingested and structured and formatted for Lena to use.
But how do you stay up to date if that's the case, right… if it can't go to the internet? So we've built this awareness engine that sits outside of the product and is able to look for product updates, end of life notices, bug fixes, patches–all that stuff. So we work with the vendors to set up that infrastructure as well. And so we have this separate system that's monitoring all of that in real time. So when they post an update that comes into our platform outside of the product, gets cleaned, gets ingested, structured and then gets brought over. So we maintain that simultaneously the security element as well as the real time relevance.
Craig Durr: So what you get here is this training of just released information updates that you then make sure is secure and bring it into that specialist.
Erik DeGiorgi: Exactly. And then the final two components are the observability and control. So those are simpler to explain. Lena is able to see everything on the network, what is it doing? Is it doing what it's supposed to be doing? Is it not? And so on and so forth.
Then, the control is pretty self-explanatory because we've done the integrations with the vendors, with the technology, so Lena can actually perform actions and do things.
And it's important to understand because that's a struggle for a lot of people in the industry is building integrations, maintaining integrations. And it's an important value point for NetSpeek because all that comes out of the box. We do all that behind the scenes. We really want to give a turnkey experience. Someone turns on the platform onboards their rooms–they don't have to worry about programming, they don't have to worry about keeping up to date with the APIs and all that stuff. We do all that so they can just manage their room.
Craig Durr: You know, that is fantastic. You mentioned previously, I want to make sure we double click on this: security. That is really important. Data sovereignty. These days, when anyone hears about AI, they're concerned about… “Is my data being shared with some kind of LLM on the public network? What does that look like?” How do you address that?
Erik DeGiorgi: So let's just start with the foundation of our background in B2B, SAS and just understand like we're building from that enterprise foundation. So let's just focus on the AI component. Just trust me that all the other stuff is in place.
So there's a couple key fears, I would even call them. There's like you just said, my data–I don't want my know-how, my information, of course my personal information to be used for generalized training and/or have access to the outside world or the internet or something.
First of all, all of our AI components are ours. So we build and train models. It's all within our infrastructure. We're not calling out to some commercial AI or LLM or something like that. Everything's internal. Everything's trained.
All the information that is generated through use, none of that information contains any kind of personal information, know-how, trade secret or anything… even with one organization, the way that they use the platform is not used to train another organization that would potentially even be their competitors. Imagine one managed service provider versus another managed service provided, which is one of our paths to market. You know, if they have know-how or use, then of course, that can't train a generalized model that then can be used. So we're very sensitive to that. There's new areas, there's new attack vectors. There's something oh prompt hacking. There's all these kinds of things that are inherent to AI.
We've put in a tremendous amount of effort. In fact, as cliches as it might sound, we really want to be the gold standard in AI security for our industry. So we spend a lot of time, a lot of resources, and a lot of money going and laying the groundwork in our infrastructure.
So, I mean, every single time you talk to Lena, this is how granular it gets. Every word, every sentence is analyzed. If you're using language that is counter to how the platform is intended to be used, it could be based on a single word, it could be based on a phrase or a sequence of phrases. All of that goes through a gateway. All of that is monitored in real time. So we've gone above and beyond in that sense. And we can have a whole conversation just about that because it's very detailed. But we have a lot of information on our website, and of course, we love to talk about it.
Craig Durr: Well, it's great to hear this detail because like I said, it's a top of mind concern. But what I'm hearing is that being built from the ground up with your own in-house AI infrastructure, AI, agents is a great idea. Other people are taking faster paths to market by leveraging public LLMs. I understand what you're doing. This is fantastic.
One of the really fun things, though, for me, was I actually saw the working demos that you had down the booth. I mean, we're talking about how it was able to log in, do a room check, remediate changes or issues and fix it. Like tell me about that. That's exciting to see.
Erik DeGiorgi: It's really exciting to see. And it's been a challenge to figure out how to show it because if I just showed it working, you probably wouldn't believe me. It is a little magic. It's a little science fiction. So it's really cool.
We’re demonstrating, among other things, the room check module. So it's a very common thing, a very ubiquitous task that people have to do, making sure spaces are working. We all know it. But so we're able to configure a room really easily. We're able to set up what we call a target state. So what is the room ready to have a meeting? And you know, all the equipment needs to be set up in a particular way to do that.
We break the room. That's what we're doing. We're breaking the room. We're changing the inputs. I mean, we're doing all these things that would make it inoperable. And Lena’s, again, going back to some of those core capabilities is observing that it's wrong–it's not in the target state. And then it's actually generating a workflow to actually go and send the correct commands to those devices.
And it's real gen AI in place that's saying, “Hey, this isn't the way it's supposed to be. I'm going to go and generate a command for that device to rectify that.” Because that room check agent has been trained on that task. And we have lots of these log reading, room checking, vendor-specific agents, and they're just trained on the day-to-day operational activities in these spaces.
Craig Durr: You just saved on one of the most mundane tasks. This is what I call sneakernet, right? The fact that I might have to send a person around to all my offices, let alone my different buildings or campuses to figure this out. That's fantastic.
Erik DeGiorgi: High cost work, a lot of operational overhead to do it that way, and you have people in those spaces that are frustrated because they're coming to them, they're not working. So it was low hanging fruit to go into tech.
Craig Durr: This is great. All right, so if our audience is to find out more, what should they do to learn more about NetSpeek, and how can they come in and get involved?
Erik DeGiorgi: Well if you're at the show, we're here at InfoComm. And if not, we just launched a new website. It’s NetSpeek.ai.
Craig Durr: It's fantastic. And your beta program is launching, right?
Erik DeGiorgi: Yeah. We're greenlit.
Craig Durr: It's talking about being GA here Q3. That's fantastic. It's a lot of great stuff. Erik, I really appreciate you sharing the update. Yeah. I think the next time I see you, I can only imagine what you're going to be sharing with us next time. But I look forward to that kind of exciting stuff. All right, everyone, this is Craig Durr, Chief Analyst and Founder of the Collab Collective. This is another great, wonderful update from InfoComm 2025. Stay tuned for some more updates. Thank you, Erik.
