Meeting Intelligence: Making Every Interaction Count

[00:00:01] You're listening to the HR Mixtape, a podcast for leaders who want to understand people, strengthen culture, and navigate change with clarity. Today's conversation starts now.

[00:00:16] Joining me today is Artem Koren, co-founder and chief product officer at Assembly AI. Artem focuses on using AI to turn meetings into actionable insights and make collaboration more effective. Artem, thanks for jumping on the podcast with me. Great to be here.

[00:00:37] So you're building AI that listens in meetings, which I think is such a valuable tool. But what really made you obsessed with kind of improving how humans collaborate with that approach? I just had a very simple idea when we started back in 2019, which is that if technology is able to understand what happens in the meeting, it can be very helpful after the meeting. And from there, the idea was born and we built Assembly AI to realize that idea.

[00:01:07] I love that. You know, you talk about augmented work. What do you think that means for most teams? So it's kind of the ability to custom shape an assistant for your day-to-day activity. I think that's a good way of thinking about it.

[00:01:24] You know, we used to talk about it as a superpower. We used to talk about it as an Iron Man suit. But really, it's technology's today's technology's ability to contour to the things you're doing from day to day and then either improve the quality of the result that you're trying to achieve or improve how quickly it takes you to achieve that result. Do you think that with these types of tools that we can reduce some of the bad costs involved with, you know, poor meetings?

[00:01:52] I think about all the meetings that people have to sit in, especially virtually, and you walk away and you're like, that could have been an email. You know, how are the tools that you're creating kind of change that narrative, some you think? Well, you can now focus more on what you're actually discussing and what you're actually trying to achieve. And then the technology carries the context of the conversation well beyond the meeting.

[00:02:19] So it's really, and listen, it's, it's, the meetings are the scaffolding in a way. It's never, the meeting is never the point, right? It's whatever you're trying to achieve, that's the point. So whether, you know, you're, you're having an interview and you're trying to hire someone or you're having a project meeting, the meeting is just a facilitating touchpoint.

[00:02:40] And so it's great to have a meeting, but actually it's, if you don't need to have all those meetings, that's better if you can get to the result faster. And that's what our product helps you do. Tell me more about the product specifically. How does it help and what are AI tools like this helping people do better?

[00:03:00] So Assembly AI can generate artifacts, documents, insights based on the entirety of the content of your meetings. It comes with you wherever you happen to be, whether that's a Google Meet, Microsoft Teams, Zoom, WebEx, you can also do like an offline conversation and use the mobile app to record that.

[00:03:23] But the magic is that the content builds over time, right? It builds this library of your, of your day-to-day activity as you go along, which makes it more intelligent. And you can do things like, you know, at the end of the week, you can go into Assembly, you can say, you know, take a look at the last month of meetings with this customer, generate a proposal, or take a look at all the candidates I interviewed this week and create a table side-by-side comparison according to these factors.

[00:03:49] And that's going to go to work. So it's different from Chagipity in the sense that you ask Chagipity a question that gives you an answer. But when you're interacting with Assembly, you're interacting, it's an agentic product fundamentally. And so when you ask it a question, it does research across your content library. It might go out to the internet to do some research. Then it kind of compiles all that information.

[00:04:10] And then if it needs to produce a document, let's say a customer proposal or a statement of work or a project plan, it actually will kind of design how this document flows and then kick off on a separate AI to write all the different sections of the document using the research that it found. So it can save you a lot of time in creating documents. It can serve as an infinite memory for you from across all your meetings. And it can also preserve context between you and your colleagues.

[00:04:36] So let's say you had a bunch of meetings with a candidate, but there's another round of interviews. You can share all those prior interviews with the next round interviewee. And they can actually ask Assembly, like, what are some questions that this candidate hasn't answered yet? And then they can go, you know, interview them. I love the idea of keeping context because there's nothing worse than going through a really good, for example, brainstorming session.

[00:05:00] And you have all these really great ideas and you're trying to scribble down notes and somebody will say something that's like epic. And you're like, wait, say that again. And you've completely lost it because the person is never going to say it the same way the second time. So I love that concept of thinking about using tools like this more collaboratively. What do you say to the leaders who are still, you know, scared about implementing these kind of listening? And I'm using air quotes for that because this is a very specific reason we're listening.

[00:05:27] But implementing listening tools into our meetings with things like proprietary data and trust that people are concerned about. I mean, those are important things to be concerned about. Those are the right things to be concerned about. But they don't need to be blockers to implementing this kind of technology in your world. I understand that, you know, we started way before AI was cool. And so when we started, there was a lot of pushback to what we're doing today. It's, you know, it's the opposite. There's a pull into what we're doing.

[00:05:56] But there's still organizations that are reticent. So I would say, firstly, you know, we use, let's say, Gmail every day. And somehow we have trust that it's like your emails are your own. Well, it's just a file on a disk, right? Like, so anybody can read your email. Phone calls. There's all kinds of ways to record phone calls on the side of the provider. So I think it's kind of like this assumption that these things are safe and like everything.

[00:06:21] Things are being recorded left and right in logs and system logs and activities. So this idea that, you know, everything that happens, I'm not being recorded. I mean, that's a nice way to think about it, but it's mostly not true. So it's not the fact that it is recorded, but the fact that it's done thoughtfully and it's done with security provisions in mind in a certified way. So, like, for example, you know, we're SOC 2 Type 2 certified. We're Microsoft 365 certified. We support GDPR.

[00:06:50] We support data privacy framework. All sorts of things, which is, these aren't just kind of throwaway words. These are ways to say that we've thought a lot about this. We take this very seriously and we've engaged with a lot of organizations to say, hey, look at our stuff and, you know, let us know that this is all true. So you want to look for companies that you certainly don't want to just throw a random AI product in there to listen to what you're doing. So I agree with that. Like, don't do that.

[00:07:19] Just stop, right? Don't do that. But that doesn't mean that, like, all AI products are concerning, right? So look for products that are well certified. Look for products that put privacy first. Look for products that don't use your data to train AI. You know, we check all those boxes. And then once you have that, the next thing I would say is, you know, the bit flips. You kind of have to. Because if you don't, you're losing to all your competition today. You're missing out on tremendous amount of value. It was a lot of value before.

[00:07:48] Today, it's just non-justifiable as a business to not be using something like this internally. You're just not going to be able to sustain good competitive posture long term. As you've worked with clients, what are some of the questions that you would recommend that clients ask vendors like you as they're vetting these opportunities? I've talked about this a little bit, and I'm glad that you mentioned all of those different regulations. I agree.

[00:08:17] They're definitely not throwaway terms. They're terms that you should understand, especially as it relates to your business. But I'm continuously encouraging, you know, those that are in the buying space to ask tougher questions of vendors like you. What are some of the ones that are like table stakes? And what are some that maybe can help the buyer understand and dig a little bit deeper to serve as what is a good AI tool and what's one that you should pass on? Yeah.

[00:08:42] So I'd say table stakes are things like encrypted communication, encrypted storage. Table stakes are things like having specialized agreements with your LLM, so AI endpoints, that assure that the data isn't retained longer than it used to be. The data isn't used for follow-on training. That basically your data is kept confidential all throughout. And there is some level of control with that.

[00:09:11] So, for example, you know, there are more or less kind of strict providers in that. So, you know, we happen to use Azure, for example. That's very good at doing these kinds of things. Also, table stakes are, you know, safety controls in the way that you use AI. Certain model endpoints provide security out of the box. I'd say, you know, GDPR type privacy controls are table stakes at this point. I don't know if SOC 2 is truly table stakes, but almost.

[00:09:39] And then, you know, then there's much more advanced stuff like, you know, the ability to support sovereign data storage in your own geography. You know, that's a little bit more advanced. I think you hear that a lot more often from, like, EU-based businesses more so than from US. But it does exist. So, yeah, I think, you know, table stakes around the security and production of your data first and foremost. And how that data is used and follow on.

[00:10:07] And the fact that you own the data that hits the provider and no one else does. Like, those are table stakes. Yeah. Can you elaborate on sovereign data storage for those that might not know what that means? Yeah.

[00:10:21] So, often, like, a larger company is going to have a requirement that all the content that you store related to their account has to reside physically in a country's geography or in a particular geography. So, let's say I'm a company in France. I might have a requirement that if you store my content, it has to be stored somewhere in France. Like, the data center has to be in France.

[00:10:51] So, that's what I mean by sovereign data storage. So, like, we, for example, support, like, you can pick your own geography and kind of your own storage endpoint. There's also kind of two global endpoints that we support out of the box, which is US and EU. And this is becoming more and more of a common ask, I would say, across the world.

[00:11:11] I think it's such a good question because if you are working with a large organization who has redundant data centers, right, you might have one in your country. You might have one in a different country. So, I love that you brought that up as something to ask because if that's important to your organization, you definitely want to know what your data is going to be.

[00:11:30] Have you run into conversations as you've been talking to leaders about how this might impact things like psychological safety in an organization or maybe surface something in a conversation that becomes an HR issue, for example? Yeah. So, there are, you know, there's these, there are project meetings, there are, you know, recruiting meetings, there are all these kinds of things that this is kind of like, it merits less scrutiny.

[00:12:00] But there's certainly sensitive conversations that happen. There's, like, sensitive HR conversations that happen. There might be sensitive strategy conversations that happen, whatever happens. So, the best way to think about it is having somebody with you in a discussion is very similar to having another person with you in a discussion. It's very analogous. And when you start to think about it that way, a lot of questions kind of answer themselves.

[00:12:26] And so, it's like, let's, you know, let's say you're having a sensitive conversation. Well, would you invite another colleague into the room when you're having that conversation? If the answer is yes, like, you know, maybe this guy, like, let's talk about, you can think about that colleague as an analyst or an associate who just, you know, he's like he or she, they're helpful. They can kind of do and help you things after the call or whatever, right? But, like, would you invite them into this conversation if it's a sensitive conversation?

[00:12:55] I think once you think about it in this humanized way, it's much, much easier to make that decision. And then I would say don't invite somebody for conversations that are sensitive or kind of have a, that, you know, that human factor that you wouldn't just invite a random person into that meeting. Yeah. And you always have the ability to, if you start a meeting, you know, with tools like this and then the conversation goes somewhere sensitive, just turn it off.

[00:13:24] Like, that's also like sometimes some of that, like, just basic stuff we forget that we can just, just turn it off and be like, okay, well, we're not going to record that part. It's, it's not valid for what we're doing. One question we get very commonly is, well, how do I know I'm being recorded? Like, I think that's scary to people. It's like, am I, am I not? I actually had a bunch of conversations around this. Different products do it in different ways. And some products do it in a way that you don't really know if you are. And I think that's a creepy idea.

[00:13:54] That's almost like someone hiding under the table when you're having a meeting with your team. That's not cool. Like, I wouldn't be comfortable with that. Yeah. So I want to make sure there's transparency when things like this happen. And so, for example, you know, there are products out there like Granola that kind of quote unquote record from desktop. That's basically like having a, like a scary hidden person in the room that you don't know they're listening. I don't like that at all. Assembly doesn't do that. Assembly always attends.

[00:14:23] So when Assembly is with you in your meeting, you see it on the participant list. You can't miss it. It throws a message in the chat. Hey, I'm here. I'm going to record this. It's got a video thing that says, hey, you're being recorded. We're very explicit about it. And then what happens is if you're like, oh, we're getting into like sensitive territory, you could just kick Assembly out of your meeting like you would any other participant. That's the magic. That's the idea. That's so core.

[00:14:53] And it's really like it's my personal little pet peeve around people who have these browser recorders that can hit like a record button and you wouldn't be the wiser. I think that's crazy. Like I'm not I'm not OK with things like that. And I'm, you know, a co-founder of a company that's that all we do is do these kinds of things. So so that's what I would say. It's like, you know, people who are not comfortable. Right. You shouldn't be unless there's clear transparency. And what we tell all our customers is you always know if Assembly is there with you.

[00:15:23] If you don't want to be recorded, kick it out of the thing. If you if it recorded something, you didn't want to go in and delete it. But you're always aware that it's there. So sorry, I had to throw that in. I think that's important. No, I think that's hugely important.

[00:15:37] And it goes to one of the questions I had a little bit further, but it's around when you implement tools like this, what your change management and communication plan is to talk about the value of these tools without kind of introducing the big brother mentality that I think some people can have. So from a metrics perspective, what have you seen organizations start to talk about from an ROI when they implement tools like this?

[00:16:03] We did a number of case studies across the board on our customers. And we haven't done a follow up, which we're due to do because we have a bunch of new capabilities that I think can actually multiply the number a few times. But the original number is kind of mind blowing.

[00:16:20] So our study showed that for a $100,000 investment over over one year for assembly AI, your ROI is around two and a half million dollars. That's a crazy that's a crazy multiplier.

[00:16:42] And the reason it's so high is, well, there's the basic stuff like, you know, time saved on notes, like time saved on like sharing collaborate. That's little stuff. That's like a small fraction of it. But then you get into things like meetings you didn't have to have. Then you get into things like your CRM is now automatically enriched with conversations with your customers, the real ones.

[00:17:09] You get things like your artifacts are getting created automatically. So that statement of work that used to take you two, three days to draft, you can now have in two, three minutes and on and on and on. So once you look at the totality of that value, you arrive at some major number. So it's a real force multiplier for organizations.

[00:17:33] I love the artifact creation piece of it because I think of all the meetings we have and then you have to take that and simulate it into something that you can use and think through it. And having that first draft of that artifact created for you, that's a huge, huge win. You know, as you look forward in this industry, what do you think human AI collaboration is going to look like in the next two to three years? And that's like forever in this space. I know I'm saying two to three years, like it's nothing.

[00:18:03] But in this space, that's like 15 years, you know, like the speed at which we're moving. Yeah. Yeah, you're so right. I, you know, I can't stop being amazed. Like every couple of weeks, I'm like, what just happened? I think it's going to be a lot more human. Human AI collaboration is going to be a lot more human. AI is evolving to be a lot more personable, a lot more dynamic of a communicator. What do I mean by that?

[00:18:33] So rapport is not like secret sauce. There are courses out there like to make you super charming and to build good rapport. Some of us kind of naturally have it. Some of us learn it. But so, for example, like one of the things about rapport is like you can match the person's speaking style. And then you're talking to someone that speaks kind of like you. And that's very comfortable. Like they match your cadence.

[00:19:01] They match your kind of like tonalities. They match your things. AI will get better and better at interacting with us in a way that's going to feel very nice. It's going to feel nice. You're going to talk and you're like, this thing gets me. I don't know what it is. But it's just, wow, I love talking. It's like my bro or, you know, it's like my, you know, it's like my best friend. Like it's going to feel really good.

[00:19:29] Because think about in your working life, like your colleagues, you want them to kind of feel like, you know, you're part of the team and you're vibing. AI is going to be doing a lot more vibing going forward. So that's one thing. The second thing is that AI is going to be much more imbued into what's actually going on. As if it were sitting with you in an office, kind of like overhearing what's going on at the water cooler like that.

[00:19:55] It's going to start getting a lot more intelligent about what's happening across the board with you, with your team, with your organization. And so it's going to be a lot more contextually aware. It's also going to get a lot more skillful. So today, like, you know, when AI says, oh, like I could do a PowerPoint, I can do this. It can kind of do it. Like sometimes it does it. Usually like you have a lot more work to do.

[00:20:21] I think over the next few years, it's going to get really good at those things and it's going to produce like client ready or near client ready outcomes. And finally, and I think this is a big one. It's going to. And so I'm going to tell you two things. One thing you probably have heard, which is they're going to it's going to get proactive. So it's going to anticipate what you need and not wait for you to ask it. But then this this new thing is something we've been at assembly thinking about a lot is.

[00:20:49] And I truly believe this. And I think this is kind of like a 100 percent going to happen. AI is going to get collaborative with AI. And so the agents in your company that do certain things for you really well and are like contextually imbued and stylistically oriented to you. We'll have agents and other companies that have a trust network like, you know, you can talk to this agent. It's OK. There's going to be guardrails.

[00:21:19] There's going to be guidelines. But those agents will communicate with each other in support of achieving the best result for you. Man, I love it. It's so exciting. I am one who's very excited about the future. I know that there are others that are very scared about what this could do long term, but I have good, good, good hope for everything. So this was a fascinating conversation. I think it gave us some really great tactical pieces for those listening about how to work with with AI vendors.

[00:21:48] And just to get excited about the possibilities. So thanks for jumping on the podcast and spending a few minutes with me. Thank you, Shari, for having me on. Can I tell the audience about Assembly? Just www.assembly.ai, S-E-M-B-L-Y. Check it out. There's a free trial. If you use the code PODCAST2026, all caps, you get some discount. And so check it out. And I hope you enjoy it.

[00:22:20] Thanks for tuning in to the HR Mixtape. Like, share, review, and subscribe to support the show. And help more people discover these conversations. Until next time, keep the conversation going.