109: Using AI Meeting Notes to Turn Conversations Into Business Outcomes with Artem Koren

Speaker A: We actually had to build our own transcription engine because there were no engine of quality available in the market at the time.

Speaker B: Generative, supporting the coding and stuff like that, you didn't have that advantage. What was the development of this product like? Challenging, quick, easy, dirty.

Speaker A: We had to do it the hard way. Imagine that. Imagine people used to code on keyboards with it. Crazy. Google, Amazon, uh, Microsoft, they were way behind in terms of transcription accuracy, especially for multilingual domains, but even for English.

Speaker B: So was this a hard sell when you went to market with this?

Speaker A: Yeah, it was definitely an uphill battle. No one really knew what the hell we were talking about. They're like, what does it do? Why is it in my meeting?

Speaker B: How does this help me for 2019-2022 for you guys? Just like you realized we're early.

Speaker A: We knew we were early. We had an idea that no one else was thinking about. Being early was a strategy. We wanted to be the innovators, we wanted to be the cutting edge developers. Our goal was never to build a product that does notes. Well, that was sort of the first bus stop on the, on the journey. Our goal was to build a product that's.

Speaker B: Artem Koren is a technology founder, product leader and co founder of Sembly AI Helping businesses turn everyday meetings into smarter decisions, stronger execution and real outcomes. Welcome to Using AI at Work. I'm um, your host Chris Staigle. Each week we'll be learning how today's business owners, entrepreneurs and ambitious professionals are getting more done with smart use of tomorrow's tech. Let's get started. Right now, every business leader is asking the same question. What are we going to do about AI? If this is you, chiefaiofficer.com has the answer. We give you a simple path forward where we provide executive and team training so your people know exactly how to safely use generative AI in their day to day. We also manage the deployment and implementation to make sure tools actually get adopted and deliver results. And we'll also guide company wide transformation so AI becomes part of your operating system, not just another shiny object. The companies that act now will increase productivity, cut costs and grow faster than their competitors. Those that wait will get left behind. So if you want to make AI work in your business, visit chiefaiofficer.com and see how we're helping companies of all sizes finally get results from AI. Ladies and gentlemen, welcome back to another episode of Using AI at Work. My name is Chris Daigle. I'm um, the host of the episode and today we're going uh, to be covering something that I'm particularly interested in, and that's meeting intelligence, leveraging AI to just help us capture a lot more context, um, from the conversations that are happening in our business. And we've got the best person to talk about that, and that's Artem Koren. He's the, um, chief product officer, but also co founder of Sembly AI. They were early in on this space, and I think we're gonna have a fascinating conversation on maybe how we should be using it different, maybe some applications, um, that we're not thinking of, how they're using generative AI, what they're seeing in the marketplace, and all those good things. So, Artem, before we step in to the conversation, anything you want to share with the audience about who you are, what you're up to, and, uh, what simply is doing.

Speaker A: So. Hi, glad to be here. Thank you for inviting me.

Speaker B: Thank you.

Speaker A: We are Sembly AI, as Chris mentioned. Uh, we are one of the founding companies in this space, and I think that makes us a little bit different because we've gone through all of the evolutions from the early AI machine learning type work, uh, all the way out through modern large language models, generative and agentic, and we've survived all of those pivots along the way. Uh, so I'm happy to talk about how things have transitioned, what things are looking like now, maybe some things about the future.

Speaker B: So, um, and thank you for that, Artem. So I noticed that, uh, you and your co founder, there's a lot of, like, professional consulting background and experience, so you've been in a lot of boardrooms, you've been on a lot of conference calls and zooms and all that kind of stuff. So you're. You're better suited to talk about what a user would need for this tool than somebody that just was a Y Combinator kid that had an idea. Right. Um, I mean, would you say that was accurate?

Speaker A: Uh, I would say yes. Personally, I would classify myself as kind of a techie person that got lost in the world of business for about a decade and a half. So that's kind of how I bring together all my backgrounds to do what I need to do. Uh, but, yeah, we're fundamentally different. And our perspective, there's a different my name, my co founder than, let's say, the typical Y Combinator, um, cohorts.

Speaker B: Nice. Okay, so, um, you mentioned that you guys had. You were early in on this. How early were you guys? Like, when did you launch Sembly 2019.

Speaker A: Was when we first started, okay, and um, the environment was very, very different. There were very few people in general thinking about AI, uh, especially having products that uh, are forward with AI and use AI as the base foundation. Uh, the meeting intelligence space as such didn't really exist. There were a couple of products in the market, uh, none really comprehensive or robust. And uh, when we got together with my uh, co founder Gilmore, we thought about how valuable it would be to have a technology that understands what actually happens inside the meeting and then can do something useful work wise after that. And that's how the idea for the business was born.

Speaker B: So what was the pain point that you were, you were addressing? Was it something that you guys were personally experiencing with the standard, if I'm not using AI on my meetings, I know that I'm losing context. I'm not really paying attention because I'm tracking down notes like was there a specific situation, uh, or pain point that had you guys go, there's a business here?

Speaker A: I would say that we went a little bit of an unorthodox route because the pain point was that uh, meetings and team conversations were disconnected from technology in general and there was no way for technology to do anything useful connected to what was actually being discussed. It's not a clear pain point because no one leaves a meeting and says, oh, I wish I had an AI agent to create these things for me. Like no one wakes up in the morning and says that. Similarly to how no one woke up in the morning and said, you know, I need a, I need a phone that I can touch with my finger and things would happen. And no one woke up in the morning and said, I need an app that lets me take photos of my breakfast and share it with all my friends for Instagram. So, so it's not, it's not a classic founding of a product in that sense, but we felt that the gap was uh, very real and that the addressing that gap would be transformational in terms of value. And we pursued that, that direction from the get go.

Speaker B: You know, man, I'll, I'll concur. Like in personal experience I'd never, I had never considered the, the contribution that uh, an AI note taker or whatever sitting in my meetings would have had. Right. And like all during COVID that would have been extremely helpful. But I wasn't even using any type of AI, not even GPT 1, 2 or 3. Right. So I, I totally get that. It was kind of like with Henry Ford. He says, you know, if I would ask my customers what they wanted they would have said faster horses. Right. They weren't necessarily thinking about what was possible. So, uh, what was the. Was there. I mean, now anybody could. I'm not gonna say anybody. It's a lot easier to spin up tools with vibe coding and things like that. With Generative supporting the coding and stuff like that. You didn't have that advantage. Um, what was the development of this product like? Challenging, Quick, easy, dirty.

Speaker A: We had to do it the hard way. Human labor on keyboards. Uh, imagine that. Imagine people used to code on keyboards with things. Crazy. Yes, we're in that. We're entering that world. Um, and that world is real today. Yeah. Uh, it was some things that surprised us was the low maturity in the audio to digital world at the time. Uh, we were surprised that some of the major providers in the space, including Google, Amazon, Microsoft, they were way behind in terms of transcription accuracy, especially for multilingual domains, but even for English. And then there was also something called diarization, which is when multiple people speak in the recording and you want to figure out who said what. That domain was Very, very immature. And so when we started, that was one of the first things we realized we would need to solve, which is having accurate enough audio to text transcription, um, and having accurate enough speaker splitting, um, and then move on to multilingual and then m. Multilingual within a single environment as well. Like inside of a one meeting where you can speak English partly, speak Spanish, partly. So those are some things we had to solve. We actually had to build our own transcription engine because there were no engine of quality available in the market at the time. And the one we built for English language was competitive with just the top, top performers in the market, companies like Rev and companies like SpeechMatch. That was the first mountain we surprisingly had to climb. And then the next one was the connectivity aspect. So you need to be in the room with the team. Like you can't build a new app. There was a couple of, uh, products that tried to do that. They tried to build a separate teleconferencing platform that just did this and that. You know, they failed fairly quickly because people weren't going to give up their zooms and their Microsoft Teams and Google Meets. They want to be in that environment. So we had to invent a whole new technology to actually, um, human participants inside of the meeting environment across different environments. Um, and so we built that technology from scratch. Later companies grew up like Recall AI, that kind of did what we did, uh, a year before them. Um, and now many companies that try to do products that attend meetings use something like recall. But we had to build our own proprietary stack to do all that. So those are some of the interesting unexpected challenges that were not AI related. Surprisingly for a company like ours that was very AI forward.

Speaker B: So I don't quite recall. I was a client of Rev.com right back in the day. Um, but I don't recall. Did they provide any meeting intelligence or was it strictly the transcript?

Speaker A: Strictly transcript. So Rev was always thought, yeah, yeah.

Speaker B: So was this a hard sell when you started, when you went to market with this?

Speaker A: Like yeah, it was, it was. So no one really knew what uh, the hell we were talking about. When we would talk to people, they're like, what, what does it do? Why is it in my meeting? How do I, how does this help me? Like a lot of very officious. No one's ever seen anything like it. No one's ever seen an AI agent in your meeting along with humans. And yeah, it was definitely an uphill battle. Um, there was a lot of knocking on doors and then a lot of explaining and explaining and explaining. Um, it was a tough market and then it really flipped in 2022 when um, Chad GPT started becoming like a well known thing. So end of 22, early 23, uh, and that really changed the environment also. What changed the environment was there was a lot more investment into the media intelligence domain from companies like Otter as well as like others and that they help to socialize. That this segment of product exists and how it's gong is another example. And so people slowly started to be become aware of this new type of technology and that really changed the game. And I think about 22, 23 is when we first experienced our first major growth spurts.

Speaker B: So I launched my company, Chief AI Officer, right like early March 23rd and I thought we were already late. I thought Everybody's using chat, GPT 3.5, blah blah blah. Obviously wasn't the case. And being that early sucked. It was painful until it wasn't right. But it was a good thing that we had started as early as we had so that when the market was ready for that message, we were already on the scene. Was that the experience from 19 or uh, 2019 to 2022 for you guys? Just like you realized we're early.

Speaker A: We, I mean we knew we were early. Yeah, we knew we were early. We had an idea that no one else was thinking about, but that was part of the strategy. Being early was a strategy. We wanted to be the innovators, we wanted to be the cutting edge developers. And we have an exceptional team, technology team behind us that we nurtured from ground up. We've never outsourced, we've never outstaffed, We've always nurtured our own team. We started with a small team in Ukraine. I moved to Ukraine for a couple of years to handhold that team to fruition. And today our team is distributed across all of Europe. Um, still, you know, a bunch of people are in Ukraine as well, but we're pretty global now as well. We have people in Japan, et cetera. Um, so yeah, so it was the quality of technology and, and the quality of product and the product experience was always an important thing. I like to joke with, you know, when I talk to people that we're, we do the exact opposite of what typically a, uh, startup company probably should do. And I would recommend do. I don't recommend the path we took, which is you build the sales and marketing pitch first and then, and then maybe you build some UIs and then, you know, then you, at some point you try to develop something that resembles what you've drawn. And we did the exact opposite. Like in the first two, three months of just starting the company, we were playing around with ideas like having a microphone device with a, uh, Raspberry PI little device. Ah, that would be recording device, that would be streaming things. So our earliest prototype actually used the Raspberry PI and the, and the streaming kind of buffering setup to show that you can record in an office room and have the transcript come out on the other end in your cloud. So it was crazy what we did. Don't do that. Everybody listening, don't do that. Do the other thing I said, which is do the show and tell. Show the pretty picture, show the pretty interfaces that have no technology behind them. Get, uh, some responses, get some investments based just on pretty pictures and then do the hard work. Um, but we kind of do the opposite. We build the hard stuff first. Um, I wish I could say we've learned our lesson. We so haven't. We 100% have not. We continue to do this. Uh, but it has its own benefits, uh, which are that we are usually first to market with technology, generally speaking, and technology that actually works, like, not vaporware, not like, oh, I thought famously, there's another major company in our space that started with actual people, like listening to the audio and typing meeting notes. I couldn't believe that's true. I'm still not sure really if it's true or not, but when I heard that, I laughed uh, investors would tell me years ago when we would talk to them that there's another company that does this. And I thought it was absurd, but apparently that's what it is. Meanwhile, we had a technology. No humans in the box. There's no human inside. Under the chess set there. It's all tech. And um, you know, we're, we've, we've done a pretty good job of being the innovators in the space so far.

Speaker B: So the messaging that you used to kind of explain what Meeting Intelligence was then versus now, has it changed for sure?

Speaker A: It evolves, I would say it evolves almost every year. Um, I think when we started initially everything was really about meeting minutes, meeting notes. Like that was the low hanging fruit. And that was something that was easy to understand, something no one really liked to do. And it's very obvious how that saves time and effort and is useful. So that was the messaging from the get go. Uh, but I think that's a very low bar to reach from a Minion intelligence product. And so as we move forward, we pushed uh, more on our original founding principle that our goal was never to build a product that does notes. Well, like that wasn't the goal. That was sort of the first bus stop on the, on the journey. Uh, our goal was to build a product that's extremely useful to individuals, teams and organizations by virtue of being powered by the context of their day to day work. And meeting notes are like the lowest hanging use case for that. And so I would say the meeting search notes, um, even things like identifying tasks, risks and issues, all of that kind of was the first layer that we built and a lot of messaging was around that. The next level of messaging that we introduced was about connectivity to your third party applications. So now that you have meeting Intelligence, you can imbue all the applications in your ecosystem, your CRM, your knowledge base, your test management system, with the intelligence from your meetings automatically. So now you, your investment into all those applications grows, uh, in value because suddenly your CRMs automatically are enriched with actual conversations, actual notes from conversations, et cetera, et cetera, et cetera. So there's a lot of value baked into that. So that was the second generation of our messaging, you can say. And then in the, uh, kind of the more recent generation, like the third generation, we built something called Semblian, which is in some sense like a ChatGPT for your meeting library. So imagine if ChatGPT had all your calls for all your calls.

Speaker B: Wow.

Speaker A: And so, yeah, and specially tuned to do a good job of Researching for materials from those calls and producing answers as well as document materials based on those calls. So today you can go into assembly and can say, take a look at my call conversations with this customer for the past four weeks and generate a sales proposal based on those conversations. And it will create, you know, we'll, it will go out there, we'll do research on those conversations, it will pull together all the information that it needs and then it will generate a document that's hyper specialized to that customer, to your business. Uh, you know, very, very bespoke.

Speaker B: So I've got some questions because that's like we do a lot of that, right? Um, first off, it sounds like your second, the second iteration of what you guys were talking about, you were really early on like leveraging multiple sources of context, not just the meeting. So that awesome there because now that anybody can use connectors or whatever in GPT, it's making a difference. A big part of what we train when we go on site is it's one of the first things we do because most people are using just plain vanilla versions of the tool and they're not really, you know, like leveraging CRM and calendar and all that stuff as context for their LLM sessions. So you guys were definitely early on that. The second thing, I like this concept. A lot of being able to um, isolate just my meetings as a context source of, you know, uh, of queries for the, the intelligence to do and to create this bespoke proposal, uh, or whatever that's, that's referencing very specific and nuanced points that came from those conversations. But let me ask you, is the, the sales team, let's say in this case, are they able to create that proposal or do they extract the information and then go and create it wherever they, you know, what tool they were using? Does that make sense?

Speaker A: Yeah, for sure.

Speaker B: Okay.

Speaker A: They're able to create that proposal inside of Sembly. And there's a lot of value in doing it in Sembly because unlike so GPT, you can connect it to any data source. But because it doesn't really, it's not specialized in that data source, it's going to use whatever heuristics it will use to think about what it should do. Now you have, let's say you might have hundreds or thousands of meetings for the, you know, the year and if you just throw that at ChatGPT and say good luck, it's going to try to do something, but you're going to get what you're going to give and so you might. It's going to be very hit or miss. Our product is tuned to work really, really well with your meeting content library. So when you go into assembly and you ask it to generate that artifact, it's going to do a really nice job. More than that, it's going to ground things that it comes up with in that artifact, in that artifact to where it found it. So it will say, like, you know, it will mention like this, you know, like it will put a little tag and say, this came from this meeting. This came from that meeting. It's very, very, uh, careful. And we have special layers of technology, uh, that ensure that there is no hallucination happening, that it doesn't try to gaps with whatever it feels is good. If it has gaps, it will tell you, um, that's another part, because we understand the use case and the kind of customers that we have, and we know that it's not this. So that's the big difference. I think people think, oh, can't I just do it in ChatGPT? You can do a thing in ChatGPT.

Speaker B: Yeah, right.

Speaker A: Just like I can, you know, like, if I have Excel, does it mean, like, I don't need, uh, QuickBooks or, you know, TurboTax? Like you can do a thing in Excel, but is that going to be right? Like, are you going to respect all your organizational principles? What about privacy? What about personal information, et cetera? There's so many things to consider. If you're just by yourself working, it's just you chatgpt all day, go for it. But if you're trying to do this in a business environment, you want to make sure you're using a tool that's really thought deeply about what it means to create a good outcome, and that's what is able to do.

Speaker B: So I probably should have asked this at the beginning, but I just have an assumption. I understand. I believe that I understand what meeting intelligence is. What. How do you define meeting intelligence? It's not just, as you mentioned, it's not just the transcript and the, the summary. There's more to it than that.

Speaker A: M. I would define meeting intelligence as the ability to use what has been discussed in meetings of your teams to power business results. Okay, so if you kind of, you can draw kind of two points, right? Two dots. One dot is all of the different meetings your teams are having every single day. That's one dot. And then there's the other dot, which is some strategic target that you have as a business, whether that be a revenue thing or a client Acquisition or a, uh, cost efficiency you're trying to achieve. Whatever those strategic initiatives are and what Meeting Intelligence is, it gives you the ability to connect this dot to that dot. Now if you can do it, you've got Meeting Intelligence capability in your organization. If you can't do it effectively, then you don't. And of course capabilities have maturities. Maybe you can do it a little bit, maybe you can do it a lot. And so Sembly is here to draw a big fat red marker line from the first dot to the second. Give you a lot of power connecting those two things. Um, you can get little bits and pieces of the way out there with different technologies. But we're very strong in this domain.

Speaker B: Okay, so there's, you know, the, the, the concept of a, an AI powered meeting, note taker. Anybody that's joined a Zoom or a teams meeting, they see all these different flavors pop up, right? People have gotten used to having those join and in some cases there's more of those and there are people on the meeting. Um, but just because you're capturing the information doesn't necessarily mean that a, you're capturing intelligent information or that you know what to do with it after the call. Because like I uh, know plenty of clients there, they said, hey, this is a great idea. I've got thousands of hours of conversations, but I don't know that they've ever touched it. Right. It's just, it's context collecting dust I guess. And, and this context, to some degree it will, it's perishable. Like call I had six months ago isn't nearly as valuable perhaps as a call I had last week with that same client. Um, what would be, I guess, best practice of if, if I'm using Sembly or any tool really to capture this context, what should I be doing with it and on what timeline, event horizon should I be doing something with it?

Speaker A: I'm thinking how to best frame my response. As a product, uh, maker, I'd rather you not have to think about what should I be doing with my meeting content.

Speaker B: Good point. Yep, good point.

Speaker A: I would rather you think about what is the most impactful thing I can do in my business tomorrow. And then what I hope to do is give you a, uh, technology, uh, give you an interface, give you a tool that can help you get there as close to that target as I possibly can with as little investment from you, resource wise as I possibly can. That's my goal. And whether that means bringing in meeting context or something else like that's, that's a different thing. And so for example, so there are certain things uh, or certain aspects in our product that we have like we have something called old insights that it will go look at the meeting and it will tell you like for you personally for your role, what the next step uh, suggested things could be. So if I'm a chief product officer they'll suggest certain things. If I'm a salesperson, suggest completely different things. This tech already brings a lot of different context beyond the meeting. It understands who you are, it understands who you're in the meeting with, what company they're from, what company you're from. It looks at the, the, the trajectory of the meeting in terms of what is the meeting intending to accomplish. It synthesizes a lot of information and that, and then it goes through an adversarial process internally to come up with like here are some vectors of next steps that we think would be very valuable for you. So that already exists, that already exists in the product. But I think the future is more proactive and the future is also more client ready. Meaning that Sembly has a good idea of what's happening in your business day to day and there's no reason why after you had a call you wouldn't get a little notification from Semly saying hey, uh, you just had a call with this customer. I just generated this beautiful pitch deck using the customer's branding that highlights the things they were concerned about that your product can deliver in the best possible way. And the last slide is a call to action for a 15 day pilot or whatever. It's a complete ready deck that you can just look at and say okay, awesome and shoot it off like that, like that level of uh, automation and productivity. I think that's the direction we want to head into.

Speaker B: I like that. So I don't know if the technology or the, the functionality currently exists but am I able to set like I don't know, share the strategy with simbly or share some like uh, key things for it to be looking out for? Across all the calls happening in my company with you know, people that have simply accounts am I able to give it. I don't know what you'd call that like I don't know, um, uh, signal to look for.

Speaker A: Yeah. Today not universally like you can do it on each individual assembly interaction when you're doing multi meeting chat. Um, but you would have to like push it in the prompt every time to uh, we're, we're building something very, very innovative that's going to be released later this summer. That's a lot more along those lines.

Speaker B: Got it. I like that. Um, so I'm starting to get a picture that of all the tools that are out there, just because they're sitting on your meetings and recording things, they may not necessarily um, be ah, getting full functionality of what's possible with meeting intelligence. Is that how you would. I guess it would be like light competitors to, you know, like some of these companies that, because the people, um, I guess some of the names that people think about like a Fireflies let's say. Right. Like that is doing a good job. I guess there's a different category. There's meeting capture and then there's business intelligence. I guess almost is like a different category or division of this technology, this capability. Is that how you see it?

Speaker A: A lot of products in our space, they from like you know, 100ft away, they look all very similar. And as you get closer they have very different specializations. And so for example, like um, if we take something like Granola. Right. So they're very focused on like it's a local recorder without an attendee in your meeting.

Speaker B: Right.

Speaker A: Which is great for certain cases. Like for my personal notes. I like that. Um, but I don't use granola, I use sembly. But I would like that. But if you're in an organization of any size and if you ever have to ask the question, wait, are we being recorded right now? That's a problem. And that's a problem that granola can solve by having a local recorder easily. So that's one. So they all, so they try to do different things whether the focus is on um, actually meeting notes, um, or the focus is on capture or the focus is on sales or the focuses on intelligence and artifacts.

Speaker B: Yeah.

Speaker A: So we're touching on something important and sorry, I don't know if you're hearing the noise outside my window.

Speaker B: New York City. Yep.

Speaker A: I'm like smack dab middle. I'm theater district. I'm in the center of all the fire engines. So apologies in advance.

Speaker B: Oh good.

Speaker A: I think we're hitting on something important which is, I'm not sure that AI like meeting notes or meeting capture or even meeting intelligence is long term a product. And I say that because if you look around Microsoft Teams can give you a transcript. Google Meet can now give you a transcript. If you go to HubSpot or Notion, they can attend your meetings now through recall or whatever they want to use. Um, there's different apps that are much more special like gong that you know, can Power the sales team through a transcript Zoom does transcripts and has its own AI assistant. So this idea of capture the context and use, that's kind of the core idea. But then the actual value is coming into very different streams and our, our view is that meaning intelligence as a component or an aspect is very important. That's how you're able to connect to the most raw and the most up to date, uh, cadence of your business. Email won't do it for you, Slack won't do it for you. Um, it's the most clear, the most impactful set of content you can capture and use. But as far as what the product is in terms of the value it provides to the user, I don't think you can just stop at uh, that you have to really think more deeply as to what your product is trying to achieve for your customers. And I think we've evolved. You know, five years ago I might not be able to say this today. I think that's pretty clear. At least where, where we're standing.

Speaker B: That makes sense with the new release that you're talking about this summer kind of is following that trajectory. It's. Yeah. So um, and I think this is, you brought up a good point. They, from a distance they all kind of look the same. And for kind of like you know, we talk about AI developing your palette and that sort of thing. It's almost like with wine. Oh, that's red wine. To one person it's just red wine, but to the other person it's ah, a premier crew. You know, blah blah, blah, blah blah blah. So how should someone, I don't know that there's a ton of value unless you're just wanting raw, raw data, raw context for something like uh, a Dorsey style hierarchy to intelligence. If you're m familiar with what Jack Dorsey's doing at Block, maybe if you just want raw call transcripts there's plenty of tools out there. But if you actually want some ah, to leverage intelligence for the consideration on like what to do next, it's a much smaller category of all the what would be perceived to be copycat similar products. Right. Okay.

Speaker A: Yeah.

Speaker B: Um, so how should someone listening to this make that distinction? What should they look for? Because I don't, I uh, would much rather have leverage the intelligence to give me insights like you're talking about. Right. Or even suggest proactively like hey, I spun up this deck, do you want to send this out to the client kind of thing? Great ideas, awesome. If I'm interested in that as a business operator and uh, not so much on just having, you know, something recording or capturing. What characteristics should I look for? Like what sets simbly apart from the broad category of like meeting notetaker?

Speaker A: As far as factors you should consider when looking at this space broadly, first of all you should think about whether your organization is the kind that wants to do its own uh, development around your meeting content or if you want this to be a turnkey solution. Uh, if you are comfortable with getting into development around it, which means you'll have to manage the attendees, the speakers, the privacy, all that, that's great. Um, but then you can use uh, any solution that gives you good access to the processed meeting content. Which means if you just need a transcript and meeting notes and whatever language, um, you can just use a solution that gives you that, sembly gives you that, that's fine, we do a lot more. But there's other products that do that as well. So if you're trying to use the product as a pass through, um, that you know, I think that broadens your range and I think at that point you probably are just a cost sensitive buyer. You're just looking at what gives me my volume of data at the cheapest. Uh, if you don't, if you're not thinking about spinning up a development team around um, your meeting content, then what you should think about is what are the elements of my environment that can really benefit from being infused with what's going on in my meetings. So you know, sales is a very easy ah, layup example. But also recruiting is like if, you know, wouldn't be cool like if I had some value, like if my interviews can actually be used by AI down, down the road to, to figure out who my candidates are. All that kind of stuff, um, in project delivery, like wouldn't be cool if you know, my, all my visibility and reporting around what's going on in a project was automated and AI handled because AI is hanging out. So one way to, so if you're moving beyond like I want to develop this myself, if you're moving beyond that one way, that's a useful way to think about this kind of technology is think about it in the human form. Like if I could put a person in every meeting who works for department X, what could that person do for that department? And then just substitute person for AI agent or AI tool.

Speaker B: Yeah, yeah.

Speaker A: And I find that that's, that's a very useful um, substitution and it helps you to ask the right kind of questions.

Speaker B: Yeah, that's good. Um, you know, earlier before we started the recording, you mentioned that you guys are like a lot of companies and, and as they should be, uh, heavily into agentic application in this meeting. Business intelligence capture and that sort of thing. Can you kind of explain maybe what you see next phase looking like with the agent, the addition of agentic capabilities within the, the business intelligence, uh, extraction from your calls. Um, and maybe how companies like how they should be preparing for this? Because I know the, the, the chatter is the effectiveness of the agents is going to really be dependent on the, the, the formatting and the quantity and quality of the data that's that the agent has access to. Right. Um, so I don't know if that's a question, but um, kind of tell me where you guys are going and what we should expect to see.

Speaker A: What I like to always start with is this idea that agent or agentic is first of all more a concept than a tech or a product. And here's what I mean. So it's kind of like when you look at something, you can ask, is this thing alive? I'm not suggesting agents are alive, I'm just giving an analogy. But you can say is this thing alive? In the same way you can look at a product and say is this thing agentic? And there are certain telltale signs or qualifiers that make something agentic. But an 8 but it doesn't define what an agent is. Just like if I say oh, life, you know, that could mean m, a sea turtle or it could mean an elephant or a human. Right. They're very, very different. Um, so similar agentic. So agentic just means there are certain qualifiers. What are those qualifiers? Uh, first of all, it's AI powered in the sense that it's non deterministic, meaning that it's able to make dynamic um, responses. So that's as opposed to a bot. Right. So a bot is defined like if this, then say that. Right. If they say hello, you say hello. Right. That's a bot. Ajantic has freedom of response and usually that's a part of. Okay, so that's the first thing. The second thing is that uh, it pursues a goal. So it's not functional. It's not like is this number even? There's a goal to reach. Um, and the third is that it can take multiple steps towards that goal and use context and tools, um, and decisions as it does. So uh, that differentiates it from something like the classic ChatGPT. Today ChatGPT is getting more and more agentic actually because it does research, it does this, it Decides what it needs to do. It can invoke like Python and whatever, but classic charge of you ask it a question, you get back an answer. It was a multi step. Agentic means that it's going to think about what are the steps and it's at each step it's going to figure out what tools it needs and it's going to use those tools, it's going to get some results and then it's going to make a decision for, from those results to take the next steps. So if it's AI powered, if it's goal oriented, if it's multi stepped, it smells very agentic at that point. Okay, very likely it's agentic or an agent. I like to give this prequel to my responses around agents because whenever we say agents, everyone imagines some different thing. Somebody imagines a turtle, somebody imagines a, uh, rhinoceros, um, and that's fine, that's normal. Because of where we are. We're basically in the early 90s of the Internet equivalent of the AI era. And um, you know, the, it, the Internet, the, you know, the web evolved very quickly and terminology evolved very quickly. And so it is today. So it's important to get everyone level set on the same basis. So, okay, if we're agreed on what agentic is now, I can answer the question a little bit better. And I would say that everything, um, all applications are experiencing a paradigm shift, some faster than others. In a sense, that classical software was interfaces created for humans. So for example PowerPoint, PowerPoint was a tool that allowed a person to effectively design and manipulate a PowerPoint slide. So you can make boxes, resize them, change your font. It was the mechanism of creating a slide. And what agentic technology allows you to do is to abstract, uh, yourself away from those kinds of interfaces. And now those kinds of interfaces are for agents to be able to use really well. So an agent knows how to manipulate a PowerPoint slider really well. It's really good at that. And now your job is to interface with the agent in an effective way and communicate what you're trying to accomplish as best as you can, which includes connecting the agent to all the source material it should need to accomplish this job, to give it the context contours, meaning like this is for my business or this is for that business, or this is for this business situation. Give it the kinds of contours. Again, I'm going to invoke the human metaphor. Like think about it as a consultant who is good at certain things that you're trying to Engage to produce something for you and treat the agent that way. Um, and, and this is so all. I don't know. Okay, I don't want to say all, but the vast majority of technology is shifting into this new paradigm. So think about CRMs, think about Project management systems and what you're really trying to accomplish with those products. Uh, rather than like managing individual customer records, what you're trying to do is you're trying to make the process of sale to that customer as efficient as it can be by being as clear as possible about the value you provide and offering pricing in a way that positions your product to be very, very, uh, high demand. Right. Very coveted. That's the goal. Right. Like, everything else is detailed. You know, looking back, like, what do we talk about across all those sales conversations? That's all details. The goal is to make the sales process as, uh, frictionless and as enjoyable as possible for your customer. And that's the agentic layer now. So an agent, knowing that that's the goal, would then leverage tools like CRM or materials design or whatever to accomplish that end goal for your business. So that's where we're headed. It's away from these kind of interfaces where I have to do a lot of clicking and data management in organizations more to interfaces where, uh, I have to communicate with the agent effectively and I have to empower the agent to have access to the right information to make good decisions.

Speaker B: That was very helpful. I'm excited to see what you guys are releasing. You said end of summer, there'll be some, uh, significant updates.

Speaker A: That's right.

Speaker B: Right.

Speaker A: Yeah. It's kind of huge. Not going to lie. It's.

Speaker B: Can you talk about it?

Speaker A: I can a little bit. Um, I can kind of directionally say that one of the things. So chatgpt, Claude, you know, all amazing products. I use it every day. Everybody, I think, you know, in my space uses them very, very often now. And they can do really, they can create beautiful materials. They can create like slide decks and whatever have you. The problem is that it's, you know, you no longer have to work in PowerPoint, but you still have to work in Claude to get it to do the right thing.

Speaker B: Yeah.

Speaker A: And unfortunately, it's disconnected from the reality of your business and the reality of your business situation. And also it's disconnected from the visual aspects of it, like the branding aspects. So it will, it, it definitely gets you a good part of the way towards having materials that are really useful, but it really doesn't cover the last mile. Like, it doesn't generate things that are client ready. It doesn't generate things that you can just take and kind of with just a couple of tweaks and send over it gets you halfway like it's helpful but you still have a lot of work to do and certainly it's always a one off. So every time you do it it's a new little animal. You have to uh. And there's no organization level controls, it's not repeatable, there's no quality control there. And so how cool would it be if I'm going to give things away a little bit by saying this, but how cool would it be if you had the full context of your day to day meetings uh, with all the privacy and security built in so the right things go to the right place and then you had the context of your CRM, the context of your knowledge base and the context of your design and branding environment. And what if there was an agent that had access to all of that information and who you can go to and say we just had a call with this potential customer and I need a really strong pitch that focuses on this aspect of our product and this pitch should be five slides and then in a couple of minutes what you have in front of you is a brilliant pitch, uh, that's on brand, that's as impactful as anything you can imagine. Yeah, that you can literally. And it, and it talks exactly to the customer, to that customer, their industry, their business situation, their specific pain points, just them. Um, it talks to that customer, not to a customer in that vertical, but to John Q customer and his or her thing. So that's, you know, that's direction and think about, you know, that's like a pitch deck. But what if you wanted to create a case study that focuses on a specific customer's need? Um, that's not something that's really possible today. Like case studies are long and expensive and they can project the general value your business provides. But what if you can hyper customize it to what's going on with that target customer? Um, I've said too much but those are the themes I'm going to highlight.

Speaker B: No, I like it a lot. As a matter of fact, I'm looking forward to switching to simbly, uh, soon when this stuff comes out. Even before that. I like what I've heard a lot. Um, we've used some stuff that I'm realizing that we're capturing the business artifacts but we're doing exactly like you talked about. It's going from capture into ChatGPT, which you're absolutely right. It's. It can do a thing, but it's not built to do that thing. And as a result, I'm realizing that the intelligence that we're getting from these conversations and that sort of stuff, it's helpful, but human is still big time in the loop. Um, and we're, we're not getting a finished product. We're getting a direction from the LLM. When we introduce this and ask it questions, it's not proactive. It doesn't have broader context on everything else in the business. It's only looking at that discrete application or that conversation. And better than nothing, for sure. But not, not what I want. I kind of like when you think about Google. Like people say, oh, Google doesn't give you what you want. Google gives you a bunch of links, but you're looking for the answer, right? And we've kind of convinced ourselves that the fact that there's a meeting recorder there is the answer in that Google context. But it's not. It's. It's the raw information. I still have to massage it a lot, which is fine for, for now, but boy, it would be great if AI was proactive about that. Boy, it would be great if AI said, hey, here's a couple of bits of collateral that you can send to the client. Want me to send it kind of thing. Like, much, much different, much better, much, much stronger. Um, uh, experience. So, Artem, this has been fantastic. So outside of going to Sembly M. AI, which Sembly AI, is there any other place that you think individuals who want to find out more about what you're up to and I mean, obviously enroll in the services or test them out where they should go.

Speaker A: For sure. Check out our LinkedIn page, uh, for assembly, we have a webinar coming up. Uh, I believe I saw that June. Yeah, Yeah, I think that one is focused more on the sales, uh, community. But we have many, many happening, um, every month. So check, uh, out our, yeah, check out our LinkedIn page. A lot of the updates are there. Uh, we have a lot of funny content on Reddit if you're a Reddit fan. Okay, you can find me on LinkedIn as well. Uh, but yeah, LinkedIn I think, is the best place to follow what's going on at Sembly. All the latest stuff comes out there first before we, we publish it in other places. But check out the website. The product comes with a free trial. It's not commit. You can always check it out, see, uh, how it works for you and I'll also mention that, um, if you're listening to this podcast and you use the Coupon podcast, 2026, all caps, you're gonna get, you're gonna get an additional discount for checking out, um, Chris's podcast today.

Speaker B: Thank you. So, and we'll have all those links in the show notes for the listeners. And I, um, would encourage you if you don't. If you're capturing meeting notes and you're not necessarily leveraging business intelligence, this is something to explore. So, Ardab, thank you for taking the time out. Um, I know that you guys are busy with production that's, uh, launching imminently and all this and staying really like being competitive in an environment that's got a lot of people on your heels. So thank you for that. Um, and then for the listeners, as always, I appreciate that you invested, uh, your time here and kind of, uh, getting more perspective on what's happening in the. The application of generative AI into businesses. If you know others that would benefit from this who are on the journey, they're eager to learn more. Um, we'd love to have them kind, uh, of join our listenership. Uh, and so please feel free to forward, um, this along or tell them about it. And other than that, we will see you on the next episode of Using um. AI at Work. Thank you so much for listening. Thanks for tuning in to Using AI at Work. Don't forget to subscribe for more conversations about how to use AI at work. And a special thank you to our sponsor, Chief AI Officer for Empowering Businesses with AI Education and Training. Visit their website for a free AI Readiness Assessment and AI Strategy Guide to help you get started using AI at work. That's www.chiefaiofficer.com. follow us on Twitter at the handle usingaiatwork and visit www.usingaiatwork.com for free resources to help you harness AI in your role.