Episode 62

Artem Koren on Sembly AI's Vision for Augmented Worker Intelligence & AI's Role in Recruitment

TIM: Artem Welcome to the Objective Hiring Show. Thank you so much for joining us.

ARTEM: Thanks, great to be here, sure.

TIM: It's really great to have you here. I feel quite excited and pumped because we've been speaking so much in the last few weeks about AI use cases, particularly in hiring, and I think with your domain knowledge, having been at the forefront of creating a product in this space—not necessarily hiring, but in the AI space—I'm just so excited to pick your brains and understand what's happening right now and where you see the next few years going. I wonder, to start with, if we could hear just a little bit more about yourself and about your product Sembly and what that's all about.

ARTEM: Artem Koren I'm the co-founder and chief product officer of Sembly AI. We started in 2019, so my background is in technology management consulting, and we started Sembly with my co-founder, Gil McLeff, with the idea that applying AI in meetings and then in business conversations could yield tremendous value to get better results to get work done faster and more effectively at the time. and when we were just kicking off, there was hardly any technology that tried to make some sense of what was going on discussed in the meeting. There was a lot of technology to facilitate how the meetings are held, like Zoom, etc., but really nothing that would try to make use of all of the discussion. And so that was the idea we started with: what Sembly is today is an AI teammate. One way to think about it is It's similar in many ways to the popular meeting assistant products. So products like Fireflies, Fathom, Fellow, TLDB, and products like that, in the sense that one of the things that Sembly does is it joins you on your hangouts with you there, and by doing that, it absorbs the information that was discussed, it supports over 40 languages. It supports all the major conferencing platforms, and it very neatly plugs in with your calendar, so there's really no downloads or installs. It works in a turnkey way, and so it hugs your day-to-day workflow, but then, by virtue of having visibility into your meetings, it can give you really valuable information afterwards. So the basics are that it makes all your meetings searchable. It generates contextualized meeting notes, meaning that the meeting notes will be very different depending on whether this was a sales call or a recruiting interview or a daily stand-up that those notes will be completely different. And it identifies tasks. It also can identify decisions, risks, and issues, and then you can take all that information and then automatically pipe it into your workflow apps, like your CRMs and your knowledge systems, etc., and all that is natively supported, and at this point, that's like the basic set of functionality that most quote-unquote meeting assistants have. but what's really special about Sembly is that's actually not where most of the value of the product lies, and I always find that kind of funny because in many cases, like that, that is the point for a lot of products, but actually for us, the point is the other piece, which is Sembly in 2.0 and what Sembly in 2 0 is able to do a few important things so it can give you personalized insights after the meeting, so instead of generic notes for just what happened at the meeting across the team, it infers it so it knows a lot about you and your organization; it knows about all the participants in the meeting. So it really has a good picture of you and your work, a really deep understanding of you and your work, and then it can use that understanding to provide you with personalized insights or suggestions after the meeting, so like for a recruiter, it can suggest based on something the candidate said to follow up on something or to provide something or to figure out some additional things. Understanding that you're a recruiter, understanding that this was a candidate, right, so it figures all that out on its own in a product management situation, it will do the appropriate thing; in a sales situation, it'll do the appropriate thing, so it's really customized to you, and then the most The interesting thing is that it will suggest your next steps and then the deliverable, so it will say, Here are the things you should do next, and here's the artifact or the document you should produce as a result. For example, let's say it's a sales call, and the customer in the sales call mentions a competing product; it might say, Okay, it would be actually really useful for you to create a side-by-side comparison with this product that this customer mentioned. and then it has this button called work on this, and when you click the button, guess what? It will actually make that document for you, so it can use all the information across your meetings and your team meetings. It also uses the web, and it can construct a full document—not just a chat GPT response, but like a 15- to 20-page document, full-fledged, that is very specific to our team. to the context at hand so you can act and know about your product from coming to your meetings with you, and it can pull information about the competing product by going to the web; it will also find things like reviews, and this is all done because there's a complete agentic framework underneath. So there are agents working in concert to produce these results for you, and so that's the most valuable part of the system: the Sembly and part, which kind of relies on all the information across your meeting, is this deep understanding of you to super speed document creation. You can also talk to all your meetings from Sembly. So you can ask for simple things, like what all my connections with this customer were over the past few weeks, and then we'll list that for you, but you can also ask it to trend things. You can also ask it to analyze things. You can say if you have weekly statuses with your team, it's like, Can you identify a few? trends in our weekly statuses across the team. If you're a manager and you have your team leads reporting into you, you have an option for them to share the meetings with you so then you can just say, What are my team members? What's the summary of my team's progress for the week? or something like that. It will produce that as well, so it's tremendously powerful technology on the basis of simple meetings.

TIM: Such a great explanation, and as soon as you're talking about what, to be honest, almost sounds like magic at this point, I was reminded of the last role I worked in years ago, where we'd have this recurring joke where we'd have a long, in-depth meeting about some fairly complicated bit of analytics we had to do. We joke about the do work button or the make stuff button. I'm just wishing we had that, but it sounds like we're almost there, which is really exciting.

ARTEM: No, we're not almost there; we're literally there; we're literally there.

TIM: And with this, this is great, so I really want to dig into this further at the moment. The level—how would you describe the level of quality of the artifacts that can be produced? Is it on par with what a human would do, or does it still need a bit of work? What's the current state?

ARTEM: So it's on par with what a human would do, and it needs a bit of human touch afterwards because you'll always want to have certain things included that it didn't think to, so you can think of it as a colleague, and maybe one way to think about it is that classic meeting assistants, and maybe it's like the more kind of classic teammates are more like the analyst colleague, where it's like you invite the analyst to your meeting, and that way they'll like jot down notes, and they'll distribute those notes, and they will take action points like analyst-level activity. and I think Sembly and brings us into associate-level activity, so I'm using a little bit of like consulting banking terms from us, but basically an associate is like after three or four years, an analyst graduates to an associate, and so Sembly in this associate-level activity, so what is an associate? So as a VP or as a senior manager, you would ask your associate to come up with a document and come up with a risk mitigation strategy. come up with a customer sales proposal, come up with a project plan, and they will do a lot of work. They're very brainy, like very capable-level people, and they'll produce like a 20-page document. Does that mean that document will have no edits or changes, like feedback from you? No, of course it's going to get redlined all across the place because that's how people learn and become directors, right? All the time, so this is the same, so Sembling gives you an associate-level quality result. It's very good, but it's not necessarily something that you would take and just shoot right across the board, like right across the wall to your customer or to your team. You could, like, nothing bad will happen, but hopefully, I guess not. I shouldn't say that bad things could happen. I won't say that, but as a diligent manager, you'd probably want to review it and maybe make some adjustments or additions. The cool thing is that you can actually make some of those in Sembly in itself, so you can say Oh, actually I also want to include this section in this artifact. so you can actually tell Hey, can you also add a section about this? It will reproduce the artifact with the adjustment, or you can say, Can you expand this section to include additional information about X, Y, and Z? and it will do that so you can actually AI live edit your document in a way But you can also download it as a Word document, or you can download it as a PDF and then edit it yourself further on, but the idea is to take what would be days, sometimes weeks, of work into minutes, and then the rest of the day you just spend adjusting and polishing and making it your own.

TIM: That's amazing. What do you then see as the evolution of this being over the next, I don't know, six to 12 months, or some time horizon that you can see without giving away all the secret sauce? Where do you think it's going?

ARTEM: So I can tell you where we're going, and the vector that we're pursuing is something we're calling augmented worker intelligence, so this is a cool idea, and I think this might be the conversation where I publicly shared this idea. I've had a lot of private chats with a lot of people about this, but I like it a lot. So let me try it on you. Okay, so I see that there's kind of two camps forming in the applied AI space. One camp is something that we can call digital assistants, and that's like JAD GPT Plus, so it's like you ask it a question, it will give you an answer, and it's super useful, right? So I would call that like the AI assistant space where you can go in and ask, and it will tell you it's like the grandchild of Alexa, right? So Alexa was like a very ancient personal AI assistant, and now we have the grandchildren of Alexa, who are like a lot more advanced, okay? That's on one end of the spectrum, and then in the other camp, the other end of the spectrum is digital workers. So digital workers, they're like, We are—we have a result in mind, like we're going to generate sales leads, or we're going to generate traction with marketing, or we're going to do support. and it takes that and solves the problem end to end, soup to nuts, and the ask is for you to like it regardless of what your specific business situation is; your specific organization contour or topology is you just bring this, like, digital worker in, and it will do the things that it knows how to do the way it does them. and the hope is that there's enough points of contact between how you generate value and how it generates value to create value for you, but it's almost like bringing a consultant into your organization that's very rigid in how they do their work, and they're like, Good for you that this is how you run your business, but this is how I do sales. And just like this, I'm going to do it, and if it works for you, great, and if it doesn't, I'm sorry. Those are digital workers, and those are like becoming more mature products. Over the next 12 to 18 months, we are doing something a little bit, and okay, one other thing I'll add is that between those two camps, both of those camps are converging into a reality where it's both highly productive but also highly contoured to how your organization works and how your organization is structured. and so they're from both ends because right now the digital workers are rigid, the assistants are super pliable but not very useful, and so both are trying to converge into this place where they're both super useful, like highly valuable results consistently, but also very pliable to exactly how your organization wants and what it needs. and they're not there yet, and I think over the next 35 years we're going to start to see results of that convergence, so we're making strides in that convergent space, and we're starting from the assistant end, and today I would say humbly that we're probably the furthest along as a company that's moving towards this convergence point from the assistance. And so starting from that kind of end of like you, it will not assume what your goal is; you will need to ask it, and we're calling that convergence piece augmented worker intelligence. So on one end there's a system, and on the other end there's digital workers, and then the convergence is augmented worker intelligence. I'm going to work; it's all just as highly valuable, consistent results contour to exact business dynamics that you have, and so we're moving in that direction, so what does that mean specifically? Our product will do a much better job over the next year or two in understanding Your role and the context of your role, like what you're doing and what the environment is in which you're doing it, like your teammates, like the dynamics of your business, like your business cycles, your business workflows, like that kind of stuff, so there's you, and then there's the work or the organization around you. It's going to do a much better job of understanding that. It's going to do a much better job at understanding what your goals are at different timelines, like day to day, week to week, month to month. So what are you trying to accomplish as a worker? And then finally, and I guess this is the cool one, is that it's going to do a much better job in helping you to accomplish those goals, sometimes prompted and sometimes unprompted, so it will be like more and more self-driven in anticipating the kinds of things that you need done and doing those things for you. And I anticipate that, like when we get to a certain level of maturity, there's going to be things that it does, like without you needing to ask it, so like you're trying to, let's say, I don't know, like vet a list of candidates. And this is going to become your, now this kind of is growing out to be your, like, AI recruiting supporting agent. and it knows the positions you're hiring for, and so it will maybe pre-screen some candidates for you, like it will say, Here, I understand the role; I understand this is the goal; I understand this is the timeline. I understand these are the dynamics, and here is a prescreened list of candidates from the ones that you have submitted already. and you didn't even ask it; it just knows because it knows what you're doing because it's hanging out with you all the time.

TIM: Is part of the challenge then in helping get to this point where it has that additional context? Is it just that it needs access to more data? Is it as simple as that? It needs to have it; it needs to know more about you, or is this still just sitting on top of all the meeting data?

ARTEM: So more data is good with a caveat. I had this interesting conversation with one of the most popular, if not the most popular, project and task management platforms out there. Like, very, like, the senior leadership of that product—awesome product, awesome team—can talk about it, who it was, but I think there's just a very popular project and task management product, and we were talking about this very thing, and they asked, Can this thing absorb emails? You guys absorb meetings, and can you absorb my email? And we've thought about this; we think about these kinds of things a lot; we're weird that way. And my answer to that is eventually probably yes, but think about this: if I were to look at your inbox versus look at what your meeting notes are for the last several meetings that you've had, which is a better, cleaner, sharper reflection of the work you're doing, right? Like for most of us, I think the inbox is a messy mess, and hopefully you can suss out a few emails that are important, and as you fight through everything that's not, I think that sussing-out process is critical, and I think just absorbing your inbox is going to create more bad than good if it's not done in a very clever way in the way that actually gives you things that are really relevant to what you're doing and what's going on in your life. because I think if somebody reads my inbox, they'll probably think that I'm obsessed with hiring offshore developers across the world, and I'm obsessed with auto-generating leads, and I'm obsessed with, like, various random conferences. That's probably what someone would think because that's what my inbox looks like, but it's really nothing like what I'm working on.

TIM: I mean, if we go down this little rabbit hole for a second just out of interest because I'm sure you guys have thought about this, would it be that hard to filter out some obvious crap like you've got all or nothing? It's been read; maybe anything that I've replied to is more important or more relevant. I don't know what other data points you get back from Google or Outlook, but I feel like you would be that out to do a first pass and get rid of probably 90 percent of the crap. But even then you feel like the kind of noise-to-signal ratio is going to be too high.

ARTEM: I think it's definitely possible to get it to a point where it's really good, but I think that's a complex problem to solve, so what I mean is it's not as simple as just us API plugging into your email account and suddenly we're that much smarter. In fact, I think that would make the overall result dumber if we did that. and then to actually make it smarter, you need a whole big brain on top of that operation to make sure that the quality you're absorbing from emails is good enough. The similar thing about your Slack chats is that I think if we were going to make a very broad generalization, I would say, Yeah, it's super valuable to get access to additional data. That you have available, so like our product already can connect to dozens of third-party app endpoints, CRM text management systems, document management systems, like wikis, and set it and things like that, and so right now that happens in a one-directional way, like we can send items into those apps, like transcripts, notes, and tasks. but in the not too distant future we are, you know, very open to potentially being able to read information from those apps and then incorporate that information into the kinds of results that we regenerate, and yeah, having access to more data is good, but it's critical that that data is well curated. So it's not just like data for data's sake that's very important, and two, I would say that the currency, like how current that data is, is also important because something that happened or was said, like two weeks ago, may no longer be relevant because it was overridden by something that was said yesterday. And so just having data is also not the answer, right? You need to be able to, over time, make sense of this data and then actually historically navigate that data as well.

TIM: One thing struck me as you've been talking about this, and I've been thinking a bit about the general space: a lot of businesses now are back in the office, back in the quote-unquote real world, face to face, where a lot of data is not captured necessarily, for example, in meetings as I imagined. A lot of people would do a meeting face to face and not record it; that's probably the norm. I'm assuming I've worked in an office for a while, but I doubt that most companies would be inviting in a recording of a meeting where everyone is face to face, and there must be other data points like that that are just never captured because it's not digital, but the way I see things going is there is going to be some point at which there is a stupendous advantage to purely digital companies where literally everything they do is online. They're remote, or at least everything's recorded, every meeting internally and externally, all did like just a 100 percent digital footprint business if we're going to have this kind of what did you call it augmented knowledge worker, is that right?

ARTEM: Yeah, AWA, Augmented Worker Intelligence

TIM: augmented work intelligence If we're just going to have, in a couple of years, something that's sitting over all the data sets, is there not going to be an advantage to actually having them in the first place, which you wouldn't have if you're an analog in the office kind of business? What do you reckon?

ARTEM: I think it's definitely a trend. I wouldn't say it's true for everything under the sun. There are always going to be businesses that, for different reasons, whether it be like sensitivity to exposure or the fact that they're very, like, reality physical I don't know, let's say like a lawnmower business or something like that. That they require things to be like not digitized, like literally require them to not be digital, like maybe some FBI CIA kind of stuff and whatever, so there's certainly going to be a segment of businesses that are that will require like analog activity, but I think the The trend, and probably the overwhelming fraction of businesses, will slowly realize the benefits of end-to-end digitalization. It's effectively one way to think about it, where we started this on recruiting and the kind of human capital management aspect. One way to think about it is that it's almost like this is very sci-fi, but it's also real. Imagine if there's a resource pool of workers, but they have special needs, so there's only, like, a certain kind of air that they can breathe. Let's say they're extremely smart. They're like super fast, like high energy; they can work 24/7, basically, and they don't need to get paid very much, and they're excited to do their job every day. Okay, but they need this special atmosphere to work, and that's a resource pool that's available, and it's effectively unlimited. You can hire as many of these people as you want. Now you have a choice: will you convert your organization to have this special level of nitrogen in your atmosphere so you can have access to this resource pool or not? And that's literally the situation we're in, except substitute nitrogen for you're an end-to-end digital organization because you have to be for these agents to be able to do effective work and substitute aliens for AI-like virtual Technology virtual agent and virtual worker technology

TIM: I mentioned then there's going to be some amazingly successful startups that get created quite soon that are almost like AI-native, purely digital businesses because it'd be so hard for a more mature company to go Hey, we're going to do this in a complete overhaul. We're going to have a complete rethink. We're going to have the right amount of nitrogen in the air, as you say, but for a new startup, if you get it from day zero, it's this is how we do business now, then I could just see some very differently run companies being formed right now with this different kind of mindset, which is exciting.

ARTEM: Yeah, absolutely. It's definitely revolutionary in many ways. This concept of the future of work is probably experiencing, I believe, the biggest impact since the PC was introduced. It might be more profound than the change with PCs because PCs were an extremely powerful tool, but with AI, enabled workers and augmented worker intelligence, also digital workers, they're not just a tool; they can take initiative and do things on their own behalf. and so it becomes like a very different kind of thing, and I think what we'll need to figure out is how do you manage organizational culture. how do you manage skill sets, and what are those skill sets in the face of the presence of this new kind of workforce that's an AI workforce, and how do you keep people engaged and motivated, so I think definitions of roles and responsibilities will shift over the next several years as a result of this new technology entering our workspace.

TIM: What I'm also seeing and thinking about now is almost like quite a divide coming up because here's one way to put it: how many companies in the world would currently even record meetings and do analytics on meetings? It's probably still a tiny minority, yet that's almost like for you guys, years ago, like you've already solved that problem. That's done, but most companies wouldn't even be at that point yet if I could imagine for the average big corporation if the one change they made was they started recording everything with your product and started using analytics; that one change would be like a step improvement if they actually used it the way that it's meant to be used yet. I feel like most of them probably wouldn't be within years of doing that, so you must then have some companies who are doing it at the end of their using your latest kind of functionality, yet others who for years might not even get to day zero. Is that divide going to be what's going to happen with that? Are these more traditional companies going to die or be outcompeted? Is this just all generally part of AI adoption, and the sooner and quicker companies do it, the more chance they'll have to thrive? How do you think about this kind of chasm?

ARTEM: It's a good time to be a smaller Nimbler company. It is true there's a lot of concern around AI. First of all, there are questions around how a company can use it because it's not a widely understood topic. Like we, you swim in this, so we have a really good handle on this, but from meeting a lot of people, I realized that you're absolutely right: meeting notes are like an amazing idea for them, whereas for us it's like a thing of the past. And so understanding how to use AI in the organization is a big question for these large companies, and then whether to use it because of the concerns about privacy, about liability, about what if something goes wrong—all of those kind of are valid concerns, but those are typical enterprise woes. A thousand questions need to be researched and answered thoroughly before an important kind of change in the rudder. is made to correct that humongous ship, that's the enterprise, into a new direction, and companies that do a better job at this will win, and companies that do a worse job at this will lose, and meanwhile this gives the smaller companies who can more nimbly understand and adopt AI in a way that makes their business x 10 X 20 X 100 more effective than the bigger competitors, they will have an opportunity to gain market share while the big companies are figuring it out.

TIM: This has got to be the dream moment to start a company if there are any sort of early twenties young, super-smart whippersnappers out there. Oh my God, this is maybe a once-in-a-lifetime kind of moment. If the technology is changing this quickly, you've just got a chance to really get in there and genuinely disrupt a bigger player if you can just move so much quicker and adopt, as you say, technology that's not making a marginal difference. This is like night and day almost, it seems to me.

ARTEM: Absolutely, yeah, it's fantastic. It's a fantastic time to be alive. It's a fantastic time to start something new and solve a problem in a new way, but not because AI is like, Not just so you can be like awesome T-shirts. AI, but which is what happened with com, like com came out, and it's okay; everything has to be com. So we're like, okay, like hats and gloves come, and suddenly, no, so you need to find ways that AI can really be a paradigm shift in a problem you're trying to solve, and then if you find that intersection, you can leapfrog a lot of the slower-moving competitors in this space. I'd say that that's absolutely true today, but I think in the near future we're going to be hitting a different problem, which is that because the barrier to creating solutions is lowered, you're going to have a lot of solutions, and so the problem is quickly becoming a problem of promotion, not a problem of technical realization. Because there are products out there like Builder AI, and there are others where you can be like, You just tell it what you want to create, hit go, and it will like Create the website, set up your apps. Create the application, write the code, deploy it, and place it like it will just do things. And you don't need to know; you hardly need to know anything about tech or how sustainable that kind of product is; that's a whole other question, but that is possible to do, so then if everybody does that, how do you stand out, and how do you make your solution actually dominate some reasonable part of the market so you can be a viable business I think that will be the next challenge to solve.

TIM: Especially as anyone who works in sales would have said, the last few years outbound have been very busy and noisy through any kind of email, calling, LinkedIn channels, or anything like that, as I guess pre-AI tools started to dominate, suddenly I'm really cheap and effective, and you could just bomb as many people as you wanted. So this sounds like it's just going to be, as you say, an exacerbation of that cause. Now there's going to be so many new products that get created. What's going to happen then? So yeah, let's say we have a, I don't know, a hundredfold decrease in the cost of producing something, and also not only that, it must be more people can do it. Like, suddenly, as you say, the barrier to entry is lower, so someone maybe with no engineering skills at all might be able to create a product that works. Now, that wasn't the case five years ago, so then, yeah, what's going to happen in two years? It's just going to become a pure promotion thing; it's just about building some kind of new network in a new way. Some kind of this is all going to be a viral growth loop. Yeah, what? How's it going to play out if you've got any magic words of wisdom?

ARTEM: If I had the answer to that question, I would be wearing a more expensive hoodie. I think that's a great question, and I don't know. I don't know. I think we will see. It's unclear to me because things are moving so quickly. We'll see how good the tech is. that's able to auto-build these products Is it true that not being technical, you can build a product and have it stay alive? It's hard for me to see that end to end because of my technology background, but I own the fact that I'm kind of a dinosaur now because when I started, we didn't have the cloud; we had to actually put servers into racks in big rooms and connect them to big power cords. So I have a little bit of a different, skewed kind of perception of what everything that's required to run a successful technology product Maybe those assumptions will be proven wrong, and maybe AI will be somehow Take care of a lot of things that I think are just very hard to solve. We'll see.

TIM: We shall see what about in terms of hiring, then, because I'm really excited to get your thoughts here. From my end of my view, I feel like hiring recruitment the way it's done is still so nascent. If I was thinking about hiring like anyone would think about marketing, I feel like we're not even at ground zero of being able to answer simple questions. If you asked a company Oh, tell me the average number of interviews each interview has done this week. Which interviewer is the best predictor of whether or not a candidate is going to be hired? What's the average time it takes for a candidate CV to be screened? I like that you can imagine basic BI questions like the concept of recruitment analytics almost doesn't even exist yet. like most companies would have no idea about those questions, yet then I'm seeing developments in AI that could automate away a lot of the entire thing or at least make it better than it is at the moment, so I feel like even from the current state of AI, we're just almost waiting for the application layer to catch up just to actually implement this stuff even with no more advancements. So I feel like there's going to be like a radical change in the way hiring is done with the only limiting factor that I'm probably being enterprises just catching up to being able to change their mind shift. What do you think? Where do you think hiring is going to go with the way that AI is changing?

ARTEM: I guess I should open with the fact that a lot of it is our platform, so Sembly is very popular among recruiting in HR, a bit to our surprise, if I'm being completely honest, because it wasn't built with that specific use case in mind. Like, we knew it would be useful there, but it wasn't something we particularly focused on, and we were surprised by the level of receptivity of our product in that space. We internally use it for recruiting activities for offboarding, onboarding, and recruiting, so you can imagine, for example, when we are offboarding someone, we have them record the things that they do and go through it, invite Sembly to the meeting, and quote-unquote show Sembly the things that they do, and what that creates is a video and like a guide. And they'll make a playlist of these recordings, but then the awesome thing is the new person that comes in, they can just look at all those recordings and see how to do it, but more than that, they can actually ask AI questions on how to do something, and it can use those recordings to answer those questions. So it becomes like this virtual onboarding guide that you can interact with for a new person starting up, so that part of the equation is very good on the recruiting side. Specifically, we use it during interviews, so we have to have Sembly participate in interviews, and then before it's passed to the next interviewer, that interviewer has a chance to review the interview that happened and also ask questions from that interview. So I can ask what the candidates say about this, and they might say they didn't talk about that, and I'm like, Great, that's something I'll ask, or, What do they say about this? They'll give me an answer. I'm like, actually, we don't need to interview this person because if this is the answer, then it's a no for me, something like that. So basically, you can interrogate and interview more than once with our product. I guess the important thing is you can interview the candidate without having to meet the candidate. It's weird, but it's really cool, and then there's also the kind of the process stuff, right? As a recruiter, you can ask me to list all the candidates and their key points from this week, and then we'll just build you a report, and then list me the candidates who are like the best at representing themselves on these specific skill sets, and then we'll give you a report. It's insane what you can ask, right? So the recruiters love the product, so your question was like, Where is it evolving? I think it's very important to include the timestamp with that question, meaning because things are moving so quickly, whether we mean in a year or two, three to five, or five to ten, it will make a big difference in the answer. I think in the long term you will have these digital worker interviewers, and they're going to get so good that it will be very difficult to distinguish whether you're talking to a real person or not, especially in the intake phase of the interview process, like the initial kind of candidate assessment. I think that will get really good with digital work, and it will be done in a way that it's not annoying and stupid like it's like a pet peeve of mine that people Oh, it's AI, like it sucks, yeah, today, but people will solve that. They'll make it so that it's not like these; they're going to be interactive and interesting. and you'll engage, they'll engage the candidate, like in a good way, and that won't be stupid, so the five to 10 years I think it's going that way is that you're looking, you're going to have digital, like digital interviewers, yeah, digital recruiters, and then the job of interviews and recruiters will be like to oversee the process strategically and vet the conclusions that the AI makes and then do exceptions, right? It's actually I actually really like this candidate, and for some reason I didn't like him. I'll override, and I'll just say what I'm going to talk to this person about because I see something like everybody has little hacks, like I have a little hack, like I look at the file name of the CV, and if the file name is like CV that PDF, I'm like, but if the file name is structured like first name last name, like role, I'll be like, okay, like this person thinks about little weird details. and that to me is like a little hint that may or may not be something that AI will pick up, and maybe eventually it will, so that's the long-term view, and then in the short term you're going to start to imagine, like, the recruiting process as just a big dashboard of a hundred kind of LEDs. and each LED is like a point in the recruitment process, whether it be sourcing candidates, engaging initially, sharing, like a job, like doing a preliminary screen, like doing an initial interview, like setting up an interview with the right… There's just 100 LEDs; that's what a recruiter does. and I think over the next 3 to 5 years, 1 by 1, these LEDs will light up green, and when they're lit up green, that means an AI can do it, and so you're going to, and it's going to be a little bit weird; like, it's not going to be like first this, then that; it's going to be like this silly, random LED here will light up green. That random LED here will light up green; that one will, so it will be like different parts will light up green, and then every so often there'll be like a consolidation where there's a product that comes out that takes all the greens and puts them into a nice workflow, and then more greens will light up, and then another product comes and takes all of those and puts it in a nice workflow. and then, so like, slowly but surely, like different parts of the recruiting process will get absorbed into AI, and I think in five to ten years, like, you might have a nice end-to-end recruiter-like AI recruiter experience just where you're just looking out for exceptions.

TIM: Yeah, I'm just, yeah, the mind boggles at the moment because I just feel like we're on for a sudden spike, a sudden improvement, a sudden wave of improvement. I wonder if you can think about which of those lights is going to turn green first because you're saying it might not be linear like in my head I was thinking. Oh, probably what would happen is it would be the initial kind of sourcing. Basic screening seems AI to do that if and on the current technology, whereas an in-depth final interview is like down to two people. Maybe that's a bridge too far at the moment. Maybe you and most companies might prefer to have that human touch anyway, so I thought maybe we'd go down the funnel progressively because as you get further down the funnel, it's like a higher value-to-time trade-off or whatever. But then there are things like the interviews themselves, like at the moment your product could summarize and give feedback and scoring, which most companies don't do at all, so then, yeah, maybe that's across all the stages as opposed to at one stage, and any other particular guesses around how it will develop based on which bits of the hiring process lend themselves most to the current state of AI.

ARTEM: It's hard to say most because I think, to some extent, every piece does. I don't think there will be hires without human involvement in any foreseeable future. There's going to be a human at some point plugging it in. Oh, I think, yeah, I guess there's like specialized and generalized. Like our product is more generalized. So if you're talking about specialized, there could be a specialized system for aggregating CVs and assessing those CVs for particular skill sets, and I think that's a very—that's a solved problem in many ways, and so that kind of initial screen I think is certainly not within reach like that exists like that. There's good There are good solutions for that already, so that initial screen piece, I think, is if you have a database of candidates, it's never been easier to just talk to your database to give you, like, who is most applicable to a particular position. By the way, creating job ads and defining positions is fully facilitated by AI today. If you're not using some kind of LLM to write your stuff, I think you're not doing yourself justice because it just does a fantastic job, like you, and it allows you, and it will pick up things that you might not think of, and it doesn't mean that you just copy-paste this result, but like you use it as a starter, and then you just tweak it. It's really good at that and, like, defining what the openings are and defining the kind of skill sets you can give it, like, a very broad description, and we'll do that for you, so that part already is very well solved during the actual interview process. I think that's a little bit longer term. there's a lot of products that attempt to facilitate interviews it's very early today and I think those products have a lot of challenges ahead of them but it's certainly going to be a problem that will be solved is just a little bit early for that but yeah applaud the companies who are in that space and moving that moving that ball forward And then the assessment of the interviews I think is a partly solved problem, so our product does that, but I think there's definitely more that can be done. Yeah, I think those kind of initial stages to me feel like low-hanging fruit in some cases really solved, and then it's just now it's becoming like a question of polishing the user experience with those solutions. I think the hardest one will be like a full-on interactive interview. That's a tough one to do.

TIM: I wonder, thinking about it now, if we'd see more automation in higher-volume jobs first because maybe the returns to automating it are easier, and the leverage the candidates have is lower, so imagine you're hiring, I don't know, grads, and you've got a hundred thousand applications, and you need to do a first screen or at least a few steps. I feel like companies are less pernickety than if they're trying to hire a CEO; obviously they're going to have a much higher touch process, so I wonder if we might see it in the other grad market first or in lower-skilled roles again, where I feel like the candidate pool maybe has less leverage at the moment. and so companies are maybe more willing to have a slightly worse candidate experience in some ways at the benefit of having a more scaled one; maybe that's where we might see the first real rollouts.

ARTEM: Yeah, that makes a lot of sense. It sounds very dystopian. It sounds like someone who is going for a more manual, not-so-high-skill position will have to deal with, Do you know how to do this?

TIM: Yeah.

ARTEM: In the very robotic voice, it sounds a little bit like idiocracy, but I think it's also reality. and it's funny because so much of what's going on in the world today sounds a little bit like the movie Idiocracy, but it's actually a reality, and I think this might be very much a part of it that that kind of lower, lower, lower-paid, lower-skilled positions will be handled in mass by AI to assess the candidates. Very likely makes sense; it could commercially make sense; you cannot

TIM: Time will tell. We'll see what happens over the next couple of years and beyond. I'm certainly excited, and I think there's just going to be some groundbreaking improvements in the way the world is run. I think I'm certainly optimistic. I'm a techno-optimist at the moment until I see evidence to be otherwise. Awesome! It's been a great chat today with you. I think it's been really insightful because you're just at the coalface of the cutting edge of this stuff, and so having you being able to talk about your products so passionately has been really awesome, so thank you so much for sharing your insights with our audience.

ARTEM: Thank you for having me. Tim, I'm biased. I'm an AI startup. I like what we're doing, and I'm very excited for the space, but I think I'm not the only one. As the song goes, if your listeners/viewers like what they heard today, they're very welcome to try our product. There's a free trial, so the website is www.Sembly.ai S E M B L Y, and we have a special Cyber Monday discount. I don't know when this will go live, but that's in place that you can get a nice big discount on all of our plans, and then if you want to reach out to me directly, it's Artem Koren K O R E N on LinkedIn. and I'm happy to connect