The Invisible Colleague: What Really Happens When AI Listens to Your Meetings

Should AI be in the room? Not "can it take notes" — but what actually happens when an AI system is listening to your leadership meetings, hiring interviews, and performance check-ins? When is that powerful? When is it a risk? And what guardrails do you need before you invite it in?

My guest is Artem Koren, co-founder and Chief Product Officer at Sembly AI, a meeting intelligence platform he founded in 2019. Artem has spent years thinking about how AI can enrich the way humans work — rather than replace the human judgment at the centre of it.

We dig into the "invisible colleague" problem — what happens to trust and psychological safety when people don't know whether AI is listening. We discuss why EU data protection regulations are making works councils push back against AI note-takers in performance reviews and hiring. And we unpack the hiring question: why most managers have never been trained to interview well, how AI can reduce bias and improve decision quality, and why AI should inform decisions but never make them.

🎙️ TOPICS COVERED • The real fears leaders have about AI in meetings — and whether they're justified • The invisible colleague problem: what it means when AI attends without full transparency • GDPR, EU data protection, and AI note-takers in sensitive workplace conversations • Psychological safety by design: building trust when AI is present • AI in hiring: reducing bias, enforcing structure, and what the research shows • Who's accountable when AI sits in your interview and you still make a bad hire? • Why general LLMs (ChatGPT) are not the same as purpose-built meeting intelligence tools • What good AI-assisted leadership looks like in 3–5 years

🔑 KEY INSIGHTS → AI in a meeting is like a hidden participant on Zoom — the discomfort is valid, and the answer is transparency, not avoidance. → Trust in AI tools mirrors trust in people: you need context, track record, and clarity on who sees what. → Under GDPR, employees and candidates can request the data collected on them — which changes what "behind closed doors" means for hiring. → AI consistently outperforms untrained human interviewers at predicting performance — because structure beats instinct. But the more AI is visibly involved, the lower candidate acceptance rates become. The optimal design: AI in the background, human conversation in the foreground. → AI should never decide who gets hired, promoted, or managed out. It should surface what humans miss and organise evidence for better human decisions. → Uploading CVs into ChatGPT and asking "who should I hire?" is not meeting intelligence — it's uncontrolled data exposure with no guardrails.

💬 GUEST Artem Koren | Co-Founder & CPO, Sembly AI sembly.ai

📌 RESOURCES MENTIONED • Sembly AI — sembly.ai • SHRM: only 23% of managers have had formal interview training • EU General Data Protection Regulation (GDPR)

This episode is for HR leaders, hiring managers, executives, and anyone deciding where AI belongs in high-stakes people conversations.

Keywords: AI meeting intelligence, Sembly AI, AI in hiring, interview bias, psychological safety AI, GDPR AI meetings, AI note-taker, structured interviews, AI hiring tools, meeting AI, invisible colleague, AI leadership, works council AI, EU data protection AI, AI accountability

Follow Konstanty Sliwowski on LinkedIn https://www.linkedin.com/in/sliwowskik/

For more insights check out www.schoolofhiring.com and newsletter.schoolofhiring.com

Transcript

SPEAKER_00 0:00

Today we're asking a question more and more leaders are quietly wrestling with. Should AI be in the room? Not can it take notes, but what actually happens when an AI system is listening to teams' conversations in your leadership meetings, in hiring interviews, in performance check-ins? When is that actually powerful? When is it risky? And what are the guardrails that we need in place before we actually invite AI into the conversation? My guest today is Artem Kort, co-founder and chief product officer at Assembly AI, a meeting intelligence company he started back in 2019. Now Artem's background spans computer science, product, and executive roles, and he's spent the last several years thinking about how AI can enrich the way humans work rather than replace it. In this conversation, we're going to stay away from feature lists and actually focus on what matters to you as a leader trust, psychological safety, hiring quality, and the reality of working with an invisible colleague in the room. Artem, welcome to the conversation.

SPEAKER_01 1:19

Thank you. Happy to be here.

SPEAKER_00 1:21

Now I wanted to start off by probably the elephant in the room. When leaders say they're uncomfortable with AI listening to their meetings, what do you think they're actually afraid of? And do you think those fears are justified?

SPEAKER_01 1:36

Sure. So I think first of all, it's new and different. And anything that's new and different is going to cause a little bit of concern, anxiety. What are the risks that I can think of, but maybe there are other risks that I cannot think of. And so it's very natural that this shift of AI getting exposure to your conversations creates a little bit of friction. Maybe real, maybe not. So that's the first kind of bit. It's not necessarily the fact that it's AI, it's just the fact that it's something new and different. The second thing is that I think a common concern is taking things out of context. So, you know, what if AI was listening but like it misunderstood what I was really trying to say, and then it's going to get misconstrued, and then you know, I'm on the hook for something I didn't really mean. And that's that could be another concern. And then, you know, finally we're all human, and sometimes we say things that maybe everyone in the organization doesn't need to hear. Or you know, maybe we're being a little bit too harsh on one day because that's just how we're feeling emotionally. And the concern is that if AI is always there, it's gonna latch on to those uh you know outbound things, out-of-bound things that one might say, and that's going to reflect badly on you later in the end. So I think all those are concerns. I think all those are are valid, and as with any such technology, we have to think about what is the weight of those concerns versus the value that the technology delivers, because in a funny way, for some reason nobody is too worried about like email and like Slack, and people feel very easy chatting on that, emailing back and forth. But in essence, that all those conversations are have the same level of security rapper that AI attended conversations do. And so there's really not that big a delta between you exchanging emails or you exchanging chats on Slack and you inviting AI to participate in a conversation you're having with your team.

SPEAKER_00 3:45

Now, in our pre-conversation, you actually said something that that really stuck with me. That inviting AI into a meeting is like inviting another person into the room, except you can't see whether they're actually there. Now, that's the invisible colleague problem. Can you unpack that a little bit?

SPEAKER_01 4:06

Sure. So when we are meeting with our teammates or we're meeting with a customer or with management, it's very obvious that we're who we're talking to and who is in the conversation. Because there's another human being or multiple on the other end of the line, whatever that line happens to be. When we're dealing with technology or with AI power tech, that may not necessarily be the case. Because it may not be so obvious that AI is actually hanging out with you on this conversation. And so that creates some concern. Like, is AI in the room? So it's it's this idea of a potentially invisible man or person that is in the room with you even though you're not aware. And that's you know, that's a creepy feeling for sure.

SPEAKER_00 4:59

Absolutely. I also can't help thinking about the difference in data protection regulations. There's a discussion here in Germany, and I know I know you you're based in the US, but you do also travel and have customers out in Europe. There's a big discussion about AI note-takers and allowing AI to take notes in performance reviews, in interviews. What needs to happen in order for AI to be allowed into that room? Do you think those concerns are well-founded? Do you think that you know works councils in German companies are truly managing a real concern that they should be that they should be dealing with, or is this simply being afraid of something new?

SPEAKER_01 5:43

I think looking at it from the perspective of the prior question, which is if there's something that's going to listen and understand what we're saying, we need to be aware of that, just like we would be aware of a person being on the call with us, is a good lens to take on this. And so from that perspective, it's definitely a concern equal to if there was a hidden participant on your meeting that you couldn't see. That would be a big concern. And so we should look at it the same way. It's like, do you allow a secret hidden participant in a meeting you're having on Google Meeting? Imagine if you were on Zoom and there was a feature on Zoom that someone or multiple people can join your meeting, but you would have zero notice of that in your Zoom call. Like you would just not see the person, you would not see them in the participant list, you just wouldn't know, but they're there, kind of like being invisible. That's a really scary idea. There's nothing stopping Zoom from implementing that technically, but if they were to do that, I think that would introduce a lot of uh concern around, you know, who's actually to whom actually is the information proliferating. And of course, in conversations that are sensitive, like HR and others, that could be a big, big deal and a big privacy risk. So, yeah, definitely an important concern, and it should be observed in the same lens as a hidden participant in a meeting.

SPEAKER_00 7:09

I would agree with you on that one, but it does also lead me into another question, which is what changes in a room once people know that their words are actually being captured and made searchable and potentially, you know, summarized by an AI. And how should leaders actually design for a safe environment, a psychological safety, if you will, if they do decide to have AI in the room?

SPEAKER_01 7:37

It really helps to humanize the technology when it comes to AI, because it's a lot closer to how you would deal with a human that has certain, let's say, skill sets uh uh and competencies versus how a tech works. Here's what I mean. So, one way of looking at this is like what what changes how people behave when they know that AI is in the room is well, think consider Bob the analyst. And Bob the analyst can be in the room with you, for example, on the call. Now, if you've never met Bob, you don't really know Bob, you're not really sure if they understand the sensitivity of the conversation, if they understand the lay of the land in your organization, if they're mature and professional employee or related to the organization, you're gonna have a lot of concerns with Bob joining your call, Bob the Analyst. And rightfully so, because there's a lot of aspect of trust that comes with letting someone in on a call with you. But if you knew that Bob the Analyst, you know, you've worked with Bob the Analyst for a long time, your entire team has, and Bob does a really great job of paying attention to what's being said, producing extremely quality tasks, notes, reports at the end of the week on things you've discussed, and Bob knows very well who these things should and shouldn't be shared with, will never violate your trust in that sense, will keep very everything very transparent as far as who has access to this information, then you're super happy for Bob to be there because Bob is going to make your work easier, it's gonna make you more productive, it's gonna let you achieve your results better. So it's really all about how you feel about Bob. And now substitute Bob with AI. It's the same thing. Now, if there's an AI there and you don't know what it's doing, who it's sending things to, how it thinks about your privacy, how it understands or does in your organizational layout, if it's a professional product intended for the comp mature companies like the one you're in, or it's just kind of like a personal thing that just kind of comes in and just does whatever. I would have a, you know, those are important questions to answer before you let the AI participate in your call. So it's the same as Bob. Uh, you want to make sure that they a the AI in the call, you have trust, and that the that AI has all the layers of grounding and wherewithal to do the right thing and to be transparent and professional with how it handles the kind of data that it's exposed to on the calls with you. If you trust it to do that, if you've had good experience with it, you understand that this product is designed with those aspects in mind, those important privacy aspects in mind, then by all means, you know, for me, for example, I get anxious when assembly is not on my call because I know I'm missing out on important things that I would get after the call following next day or next week or next month. Because for me, assembly is a bob that I really, really love. Like I love my Bob. But you need to get there with the product you're using, and it's you know, not every Bob is right for the conversation at hand.

SPEAKER_00 10:57

Absolutely, but that also does bring us to the question that sure you might trust your Bob. You know Bob. But let's say you're interviewing a candidate, and the candidate doesn't know Bob, and Bob's just basically sitting there and listening. That doesn't necessarily mean the candidate should trust Bob. Naturally, the natural instinct would be not to trust Bob. So how do you set up a conversation so that your counterpart that doesn't know Bob, or in this case the AI that's listening, be the assembly or any other AI that's taking notes and providing us the information, the feedback from the conversation afterwards? How do you set that conversation up so that your counterpart trusts your judgment on having your Bob in the conversation?

SPEAKER_01 11:44

I feel like we're gonna be talking a lot about Bob on this call today. Oh, we are. Yeah, we're gonna we're gonna love Bob at the end of this. So, first of all, we've already had so at Sembly that has happened because Assembly does come to pretty much all of our calls, unless there's something explicitly sensitive, including interviews. And that I have had candidates who are like, whoa, like I don't know about all this, like, am I being recorded? So the first thing I would say is it's much, much better for the candidate to be aware that they're being recorded than not.

SPEAKER_00 12:15

Because 100%.

SPEAKER_01 12:16

Without that, you know, we're getting into really troublesome territory of potential privacy invasion. So so that's that's the first thing. So first of all, they need to see that Bob is in the room with them, with the interviewer. I mean, the next question, the next natural question would be if I saw a Bob on the call with me, I'd be like, hey interviewer, by the way, who's Bob? Like, why are they here? And then the interviewer should explain what it is, how it's used. It should also explain, it would help, I think, to say any information that we um kind of record and collect, we'll make available to you. So like you'll have some kind of limited access to all the things that were recorded as part of this interview. So you know what we captured. So you can see it, you can see that we're not misconstruing, we're not you know, cherry-picking, like here's your full recording, here's the video, here's the transcript, maybe here are the notes that we generate from the transcript. At that point, it's you know, the the generated insights is already up to the organization how much they want to share. But I think introducing what it is, like what this AI is doing, how this information will be used downstream as part of the hiring process, and then providing some level of access to this information to the candidate should alleviate concerns. Now, there's still going to be candidates that go, no matter what, I don't want to be recorded. And at this point, it becomes an organizational philosophy. Is that okay? Do we still talk to a candidate like that? Or is it our policy that if a candidate is not comfortable with being recorded by our hiring technology, then that's not a candidate we can work with, unfortunately. And I've literally had this happen. I had someone come in in an interview say, like, I'm not comfortable being recorded. I'm like, well, we are an AI meeting intelligence company, so if you're not comfortable being recorded, it's gonna be very difficult for you in this environment. I don't think we can proceed. So that becomes a yeah.

SPEAKER_00 14:16

Interesting, interesting two points on that. First of all, I do know of situations where people are not comfortable being recorded, and this can be very quickly avoided by informing them when the meeting is being scheduled, that this meeting will have an AI note-taking solution present in the meeting. And if they're not comfortable with this, they should let you know before the meeting. That's one thing. The other thing is, interestingly, there's a lot of managers and a lot of companies that are pushing back against the idea of having notes taken by an AI during the meeting, precisely because according to EU data protection policies, they would need to make that information available to candidates afterwards. Which I think says something very interesting about how companies are not comfortable with how they themselves conduct interviews, how they collect data and evidence during those interviewing conversations, based on which they then decide to make hiring decisions. And I I think that opens up the door to a whole other segment here because having notes taken by an AI and then sharing it with all meeting participants does level the playing field to a huge extent between employers and candidates, for example. And I think that that has tremendous benefits to both sides. But it also changes the expectations that we should have of ourselves as, for example, interviewers because what used to be just behind closed doors, and we'd, you know, keep it nice and quiet, and you know, we could say, Oh, they just didn't feel right, they don't feel like a culture fit, right? And we just hide behind comments like that, which don't have any substance to them, is no longer available. And now there is evidence to back up that those claims are unfounded, that it was stereotyping or bias or whatever that was that was that was being uh ultimately introduced in that conversation.

SPEAKER_01 16:26

Yeah, absolutely. This is very true. It democratized the process, it digitizes the process. So it does a few things. It allows, for example, the organization to assess the quality of its interview process itself. And maybe the in the interviewing managers need some additional training, or maybe they just need to flag a couple of things about how they conduct the interviews. That could come out of this as well. And I think that's very important. But it also allows very objective criteria to be applied across all the interviews that you've conducted because now they're digitized, and AI can apply that criteria across the full set and support your hiring decision.

SPEAKER_00 17:06

Some there's some interesting data that came out on this. AI is consistently showing itself as better at predicting performance in interviews. If AI is leading an interview, it consistently predicts performance better. The reason for this is quite simple. It's structured, it's methodical. Most interviewers have not been trained in how to interview. So they lead conversations that are unstructured, that uh do not have a cadence that is repeatable. As a result of this, they are less predictive of on-the-job performance. However, what the data is also showing us is that when AI is involved in the more AI is involved in the interview process, the lower the probability of candidate acceptance of a job offer in the end. So I think you you actually hit the nail on the head there. The solution is to use AI to help maintain a structure and cadence in the conversation, but to leave the conversation to be ultimately a human-to-human conversation with AI in the background, but collecting the information and ensuring that we are using a structure of those conversations, that those conversations, those interviews are repeatable, for example.

SPEAKER_01 18:26

Yeah, absolutely. And you can even assess um, you know, did did the interviewer hit the important points that needed to be hit during an interview? Maybe some were skipped, maybe some were overemphasized. All these things now become available and they're literally at your fingertips. It's no longer like a half a year data analytics exercise, it's one prompt. AI will run across the full history of interviews and crunch you up with an answer across all of that data in just a couple of minutes. It's very, very powerful stuff.

SPEAKER_00 18:59

It is. And that actually brings me to an interesting question. Because as I said, you know, in hiring, most managers have never been properly trained on how to interview or how to recognize bias. SHRM says that only 23% of managers have had any training in that respect. And most of that training, honestly, is uh avoidance of asking the so-called illegal questions. It's not about making better hiring decisions. And this really does mean that bias already shapes what we hear in interviews, it shapes the conversation. In that reality, do you see AI being present in the room as we were discussing earlier, being to able to actually improve the decisions that are being made as a result of that conversation?

SPEAKER_01 19:48

100%. I mean, I use assembly for just this. So when I interview candidates for a position, at the end of the interview rounds, at the end of the week, I will load up those conversations in assembly, and I will ask Assembly to, I will give it the criteria I'm using to assess the candidates. I'll give it the job description, and I'll ask assembly to go through those conversations, assess those candidates on those criteria, and give me some feedback. And it's it's surprisingly good. I I just it's really good. Um I think because conversations are rich, they have a lot of content in them, and it gives AI a lot of things to play with. And then it can also come out with things that you know maybe are a blind spot for for a human evaluator. So for example, like on one of the candidates, it told me, well, this candidate is primarily in like very large budget company experience, and and for a smaller company they might not be the best fit. Like things like that. So even if a candidate is super competent, has the skill set, but they're just used to a different environment, I can suggest uh like we don't know if that will translate from a very large company to a smaller company. Maybe not something you're thinking about.

SPEAKER_00 21:00

Context is everything in hiring. The fact that someone has the right skill set doesn't mean that they can perform within your context. But if AI is sitting in your interview and you still make a bad hire, who's responsible? Depends who you ask.

SPEAKER_01 21:21

Look, I think at the end of the day, it has to be your at least for the next few years, let's say. It should be it should be a human decision because there are still things that both contextual and cultural and soft skilled and all this kind of stuff that AI cannot fully pick up on yet. And sometimes that makes all the difference. And you know best what it is that this person is going to be expected to do, and you know best what person is required to be successful in that role and to bring success to the organization. So I would never say leave the decision up to AI, I would say the human has to make the decision, but I'll We'll say this and maybe kind of throw a little curveball in here, which is that if the AI is making a decision, I think you're hedged for a lot of risks already. It may not pick the very best candidate, but it's a if the product you're using is of quality, at the very least it's going to pick someone who is a good enough candidate for this position without any critical gaps or risks. You're kind of hedged against a really terrible hire if you're using AI. But as far as picking somebody who is uh a peak performer, I would say you'd still need a human to make that determine.

SPEAKER_00 22:42

Okay, now this is where we start to disagree. I really don't believe that AI should ever be in a position of decision. And we actually made this a cornerstone of how we built Cloretta. AI should never decide. AI can definitely help you collect the information. It can definitely come up with questions to ask to help you collect the information. But the ultimate decision about who you hire should be very much a human-to-human decision. Sure, based on evidence, based on the facts that you collect. But AI, it doesn't have the the lived context to make that decision. And sure, it might make the right decision. It's still a human-to-human decision.

SPEAKER_01 23:27

I don't think we disagree that much. So I I agree that ultimately the human should make the decision. But you can never be sure whether that human takes AI's recommendation and just forwards it, or if they actually applied some of their own rigor to the analysis. And they just they just might say, Oh, I just I just agree with AI that this is the top candidate and move forward like that without giving it the due the due diligence required. All I'm saying is that if you do do that, that's definitely better than chance. Let's say it like that.

SPEAKER_00 24:02

AI is wonderful at giving you a confident answer, even if it's completely false.

SPEAKER_01 24:08

Fair.

SPEAKER_00 24:09

And thus I would even argue that AI should not even be making a recommendation. It should be presenting you with the data, it should be organizing the data for you to be able to analyze it and make that decision. But I don't think I don't think we're in a position to trust AI to the to the extent that it can make decisions about who we hire, who builds our culture, who builds our teams, who stays on our teams, how we promote people. I don't think that's AI's job here.

SPEAKER_01 24:43

But Constantia, I know some people, some of them in polit political positions of power, who can very confidently say things that are absolutely not true. So AI is unique in that particular skill. And in fact, I think it's becoming more and more of a in vogue from what I can see in the news. Ultimately, I agree with you. I don't, I would never recommend that AI is the ultimate decision maker. That's not what I'm saying. All I'm saying is that it can support a decision, and maybe the the kind of the controversial bit here is that if someone wants to spend zero time thinking about who to hire, and you have those situations, they're gonna happen. This this person is, you know, there's they're overwhelmed, there's stress, maybe there's personal life stuff. This is not like hiring their hiring recommendation is the last thing that they want to care about at that moment. For whatever reason, we're all people, things happen in our lives. And of course, maybe that's not the best course of action just to say, hey, I'm just gonna let AI do it, but let's be real, that's going to happen. And what I'm saying is that if that's going to happen, that is not catastrophic. I think that's you know, maybe the second or third best thing that could happen in that situation.

SPEAKER_00 25:59

I mean, I have seen a catastrophic instance of this where a manager basically tried to show me how he makes a hiring decision, and he takes a number of CVs, throws them into a chat on Chat GPT, and asks ChatGPT who should I hire. I have seen this, I have screamed seeing this because you know that's data protection out the window because he just uploaded raw, raw information with very little context. And of course, garbage in, garbage out. If you are not providing a safe environment for this analysis to happen, if you're not asking for the data, but simply asking for an answer, that's not gonna end well, is it?

SPEAKER_01 26:50

So I I think this is an important point to say that there's you know, everything gets painted with this AI brush, and you're right, like Chad GPT is AI, assembly is an LLM, right? There's LLM, but there's different products that are built around this foundational LLM technology, and there's ways to coax these LLMs to give you the right kinds of results. There's many, many ways of improperly using LLMs. And so when you you build it when you build a product around an LLM, so if it's a product like assembly, there's all kinds of things taken into consideration, including things like adversarial checking of your results, including safety checks, including the fact that you have deep context when you're providing results, including, for example, assembly, if assembly doesn't have enough information to give you an answer, it will tell you that. And that's also important, right? So ChatGPT won't. And so I don't want to suggest that anything that happens to have AI in the name is equally applicable to this particular use case, right? If you just upload a bunch of CVs into Chad GPT and you think that's gonna give you a good hiring decision, that's not what I'm talking about. I'm talking about products that are much more specialized for this particular example. Like for you know, for example, assembly that will actually point out like actually, I don't think I can make a recommendation because I don't have enough information. But here are some things I would suggest looking at. That would be assembly kind of answer if there wasn't enough context involved.

SPEAKER_00 28:29

I'm so glad you gave that answer because this is exactly what I was going after. You know, looking at OpenAI or Anthropic, those are very general, very large LLM models that have certain guardrails, but they're not specialized guardrails that are put into place. And we had to build the same thing within Claretta where it cannot create interview questions, it cannot create job definitions, job descriptions, job ads, or map you know, conversations to a scorecard without the correct context. And if it doesn't have enough information for that context, it will keep asking questions, it will ultimately be an AI-powered job onboarding form that is designed to get the information it needs so that you are making the right hire for your business, and it knows full well what your business is, what the organization looks like, and it takes that into context while creating this. And those guardrails are incredibly important in order to actually surface that information to collect those facts, to collect that evidence that we collect with Sembly or we collect with Claretta in order to help people make the right decision. Again, it's not about AI making a decision here, it's about AI providing assistance to us who have more complete information to make the right decision. We'll agree here. We will agree here. Now, look, if you look back three to five years ahead and assume AI will basically quietly be present in most knowledge work conversations, be this in-person conversations or online conversations, what will distinguish leaders who use it well from leaders who end up basically eroding trust and performance without meaning to?

SPEAKER_01 30:26

I think that if we're looking out three to five years, we should expect um AI in the workplace to be a lot more contextually imbued into what's going on in the organization. That's one. So it's going to be as well, if not more informed, about all the different dimensions of organizational activity that are happening. Um strategic orientation, competitive landscape, cultural elements, events, major events in the company, events outside the company, and so on, organizational change, the full deep context of an organization, three to five years, the work AIs that are going to be involved are going to have a very up-to-date what we call knowledge graph about that organization's current state and time. And the reason for that is because each instance of AI everywhere in the organization is going to act as a kind of a sensor and update that digital profile, that knowledge graph, almost real time. And so AI is going to have access to that. So it's going to be very deeply imbued into the context of the organization. The second thing it's going to do is it's going to be a lot more proactive. So it's going to be less like this QA, like, hey, Chad GPT, give me this, okay, here you go. It's going to be much more of a partnered experience. And so you would expect that a manager would have this uh essentially a teammate or a partner shadowing them, and this partner would have a team of agentic execution specialists as AI is aware of what your goals are, what you're trying to accomplish, what you need to accomplish those goals. It's going to be in the background directing those agents to create materials, knowledge, contacts, etc., to support you in the work that you're doing. And so the managers that are going to be effective with AI are going to be ones that keep their AI well informed. So really establish a good relationship. This sounds funny, but it's true. Establish a good deep relationship with their AI. It's almost like, you know, if you have an associate, let's say you're a VP and you have an associate working with you. You want that associate to really be like plugged into what's going on because ultimately they're the muscle power behind most of the things that need to happen. So you're going to need to treat AI as that. It's you're going to have to have a good deep relationship, make sure that AI is up to speed on what you're trying to do, what you're not trying to do, and also effectively control, and maybe this is the third and most important thing, is effectively control the boundaries of what AI can and can do autonomously. Because as we move forward, that's going to be the bigger question, which is, you know, what are the boundaries of control and access for AI rather than the other way, right? Because today it's like, you know, you can kind of turn on this, turn on that. But as we go into the future, AI is going to want to plug into everything all at once to help it do what it needs to do. And it's going to be up to the person to intelligently manage what the guardrails, what the AI is and isn't allowed to perform on the manager's behalf. Again, Bob comes in the picture, but you you have to give very clear outline to Bob who he can email. Like he can't just send random emails to the CEO. That wouldn't be okay. But maybe it's okay to send some emails to your execution team. Maybe that's fine. So Bob needs to have a good contour of what they're allowed and not allowed to do. Otherwise, Bob might cost some bruckus.

SPEAKER_00 34:11

That's really good. It just immediately made me think of ClaudeBot and why you should never let it loose on your main PC at the moment. Those guardrails, what it can and cannot do, those limitations are going to be increasingly important. Yeah, absolutely right. How we set that up, how we set those relationships are going to be ultimately defining for our businesses, for how we run our teams, for how we collaborate with people. Artem, thank you so much for this conversation. This was uh eye-opening and really insightful. Real pleasure to speak to you.

SPEAKER_01 34:46

Thank you, Constantine. Um, I like to uh promote a little bit of assembly while I'm on here with you, if that's alright.

SPEAKER_00 34:52

Yeah, absolutely go for it. So it's a great product.

SPEAKER_01 34:55

Thank you. Our website is assembly.ai, that's s-em-bly.ai. Check it out. There's we have three brand new plans that we just released this weekend. They're fantastic. We have MCP access now, so you can use Cursor AI, your agents, to work directly with assembly data that's available now on live. And we have AI chat across multiple meetings and artifact generation. So all these things that we discussed with Constantine on the call as far as you know analyzing your interviews and getting some feedback from AI and your candidates, you can do that today in Assembly with any of the regular plans. I'll also drop this. You can use coupon code podcast2026 and get 5% off across the board. So that's available to you as well. And um, if you need to connect with me, I'm Artem Koren on LinkedIn.

SPEAKER_00 35:43

Artem, I'm gonna add one more thing to that because I have used Assembly, I have seen what it can do, and the notes it has taken, as well as the synopsis of our last meeting, has been fantastic. So if you're considering an AI note-taking solution for your business, definitely worth trying out Assembly because it is next level to what I've what I've seen out there.

SPEAKER_01 36:07

Thank you.