00:06:36 · Sebastian Johnson
[00:06:36] S0: Yeah. Definitely. This week, you you... The public launch of fillmore, you're sort of, like, opening up to to to everybody. That was everywhere. Right? And it was like very distinct. I probably saw like a dozen posts about it on LinkedIn and they were all... You could tell that they were all talking about the same thing just by glancing for half a second, which I think is like an amazing strength of brand. Yeah. But it was like... It was almost like a new activation, but it very instantly became recognizable and dominated. Tell me about that launch. Right? So so what have you launched, and and what's the impact been?
00:07:06 · Siadhal Magos
[00:07:06] S1: Yeah. So I'll I'll share I'll share some of the sort of technical aspects of what we launched launch first, and then we can talk about, like, the... Like, what does this mean for the market? So you can think about Metaview's core product as an AI native workspace. So you go you go to the workspace to do the work of recruiting alongside an AI, and it's still a big part of our future. We do believe people will continue to want to go somewhere to do, like, lean in work, and they want to be augmented by AI while they're doing it, and that's what Metaview's sort of... The core Metaview product is. The thing we launched earlier this week, fillmore, is almost a different form factor. Same thesis, but different form factor, and the form factor is more... It's like it's a coworker. So you don't actually go to it to do work. It either comes to you where you are, which is usually in Slack, or you sort of, you know, give it... Throw it a task while you're in Slack, it goes away and does the work itself. And who this is really great for, and the work, just to be clear, is researching where to find candidates, finding those candidates, engaging with them, I. E. Communicating with them to say, are you interested in this role and sort of following up and scheduling and all those things and booking new meetings. So you engage with fillmore because you wanna fill more roles and you want an AI agent to do it for you. And so who this works really well, the the sort of reaction this is is much as the workspace makes a ton of sense for this lean in work, the fact is AI has moved on to the point where you can trust it to do, like, long time horizon, multistep processes by itself. And, of course, if it's a multistep process and it's long time horizon, it's less about having a workspace. Right? You're not gonna... You don't wanna sit there and watch it the whole time. It... You you expect it to happen in the background. Just like when I delegate a task to a member of my team, I don't sort of watch them the whole time and, you know, tweak here and there what they're doing. I sort of let them get on with it, and you get, like, high... Much higher leverage results as a result of that because it can do far more work, than you could ever keep up with. And so that's what fill... Fillmore is.
00:09:00 · Siadhal Magos, Sebastian Johnson
It's a, like, highly autonomous sourcing agent whose sort of sole objective is to get you meetings with candidates that are fit for your role, and it has it's it's right for customers who give... Are are happy to give it a high level of autonomy in making that happen. Of course, it still understands guidelines. Still understands your brand. It still understands your tone of voice. It understands all those things, but it it can get super creative in the ways that it finds candidates and gets them in your calendar. And we've been sort of amazed even even as the builders of the agent because you leave so much of the reasoning to the agent itself. We've been amazed with some of the creativity that it's shown in in getting people booked in. [00:09:37] S0: How similar do you see the role of AI in how it's impacted engineering to the impact that it could have in recruitment? You know, everyone talks about, like, a 10 a... 10 x engineer. That's, like, the thing that everybody wants and how AI can make the very best even better. Yeah. Do you see the world of 10 x recruiters?
00:09:56 · Siadhal Magos
[00:09:56] S1: Certainly. It hasn't happened yet. I think fillmore is by far the closest. I I... If I'm gonna, like, put a number in it, I think you're... We're probably in the four to five x range with fillmore as opposed the 10. I know I know people don't necessarily mean these things literally, but it's funny how it ends up being actually quite accurate. If you look at the PRs by our engineering team, like PRs per engineer, it has, I think, at the moment, I think we're... It's either eight or 12 x what it used to be. So, you know, let's call it 10 x on average. But it's hard to say use... Because that job has been transformed. Right? That is now an agent orchestration job. Now the funny thing is... And the people that, you know, join Metaview as, you know, as members of the team, what they sort of get excited about is, obviously, people sort of put the the the job of engineering on a pedestal a bit because it's just, you know, it seems such a high leverage, like, highly... It's just like, it's a high status role, let's say, certainly in the tech world. The funny thing is it's actually harder to achieve that in in... Within the recruiting workflow. And it... That's for, like, a very simple reason that the level of trust you need to have for, truly... For a recruiting agent to do work on your behalf is much higher than the level of trust you need for a coding agent to do work on your behalf. Because if I get a coding coding agent to work on, you know, 12 features while I'm sleeping overnight, I can come in, realize that they're all bad, and bin them all. And, like, okay. I've wasted a bit of money on tokens, but no one's, you know, no one's been hurt. You know? Whereas if I'm ever going to trust a recruiting agent to do work on my behalf, it cannot do work unless it comes into contact with reality. Right? The work of a recruiter involves communicating with other human beings, And you can't just say, oops. That was a terrible way to communicate. Let me do it again. You've burned that candidate. Right? And so the level of trust is that much higher. And so I think that's why you have this slightly slower... That's why we haven't had it yet, but I think we're really getting there. Because now, a, the AI is smart enough to understand all of that nuance and context.
00:11:48 · Siadhal Magos, Sebastian Johnson
B, it can actually, again, do things over a long time horizon, like maintain a relationship or help you maintain a relationship with a candidate for multiple weeks or even months while you're trying to hire them. And just fundamentally, the the models are smart enough... Sorry. The models are smart enough such that they don't make some of the mistakes they've made in the past. [00:12:07] S0: You know, when you when you look at your product road map or designing or rolling out products, how much does that, like, sort of, like, high stakes nature of recruitment impact how willing you are to push things live or test things or... I don't know. And, like, how on it just to be watching all these interactions and checking because you're right. It's really high stakes. These are real relationships, real people that if you get it wrong or the agent get it wrong, that could be a relationship burn either for your, you know, yourselves or your clients. What impact does that have on the products you roll out and and when you roll them out?
00:12:38 · Siadhal Magos
[00:12:38] S1: It's always been... Because we've always... We... We're, like, very committed, and I think it's worked out really well for us. We're very committed to being on the bleeding edge, and you cannot say that and not incur some risk. Now within our... Literally how we operate, we then sort of segment our customers, usually in conversation with them about how they're, like, thinking about their AI adoption and that sort of thing. And some customers don't wanna be right on the bleeding edge, and that's totally fine because we have some really mature products that clearly are an appropriate and best practice almost way to use AI within recruiting. And then you have other customers who are seeking every advantage they can and are willing to take some of that risk with us. Now that doesn't mean we totally push risk to the edge and make it their problem, but, you know, there is tolerance in the market to to take risk with some of these things for the the upside. In this particular case, where essentially you have an agent, you know, communicating at scale on your behalf, There's a bunch of essentially calibration workflows the agent will go through. So, you know, this is designed by us. We we sort of... We... You can be so AI pilled as to say, like, hey. Like, we should all... We should, you know, we should let the agent decide if it should even bother calibrating. We we don't do that. We think people like to feel in control. People should feel in control of of the agent, but you need to give it enough breathing room to to work at scale. If you if you have human in the loop at every single stage gate, obviously, you're never gonna, like, reduce time to fill by 10 x because it's just... The human's gonna get in the way. And so we have this calibration process where essentially you go through a a flow where you get comfortable with two things, two... The two most important things with the agent. One, does it understand what good looks like for you, for a for a for a candidate profile? So there's, like, a very specific tailored flow we take you through to get that level of calibration that you feel confident this this agent is getting closer and closer almost to being a second brain. And then the second thing is, does it communicate like you? And, again, we have... There's a bunch of context you can... We sort of take...