Beyond The Prompt - How to use AI in your company

Reinventing Recruiting: Anvisha Pai on Using AI to Automate Hiring

Episode Summary

Join Anvisha Pai, founder of Dover, an AI-powered recruiting platform transforming the hiring process for over 500 companies. In this episode, Anvisha takes us behind the scenes of Dover’s innovative approach to recruiting, sharing how her team leverages generative AI to automate repetitive tasks like job description writing and email communication. From her early experiments with AI at MIT to building a fail-fast culture at Dover, Anvisha discusses the challenges and opportunities of integrating cutting-edge technologies into a startup environment. Discover how Dover fosters a culture of curiosity and innovation, enabling team members to turn small experiments into big wins. Anvisha also explores the habits and frameworks that help her stay ahead in the fast-moving world of AI, offering practical advice for organizations looking to adapt and thrive in an AI-driven future.

Episode Notes

Join Anvisha Pai, founder of Dover, an AI-powered recruiting platform transforming the hiring process for over 500 companies. In this episode, Anvisha takes us behind the scenes of Dover’s innovative approach to recruiting, sharing how her team leverages generative AI to automate repetitive tasks like job description writing and email communication.
From her early experiments with AI at MIT to building a fail-fast culture at Dover, Anvisha discusses the challenges and opportunities of integrating cutting-edge technologies into a startup environment. Discover how Dover fosters a culture of curiosity and innovation, enabling team members to turn small experiments into big wins.
Anvisha also explores the habits and frameworks that help her stay ahead in the fast-moving world of AI, offering practical advice for organizations looking to adapt and thrive in an AI-driven future.

Key Takeaways

Dover's Website: Best Hiring & Recruiting Software for Startups with Free ATS | Dover

00:00 Introduction to Anvisha and Dover
00:14 Early Fascination with AI
00:29 First Experiences with AI
03:05 Founding Dover and Co-Founders
04:00 Early Challenges and Iterations
04:49 Exploring AI Applications at Dover
06:34 Impact of AI on Recruiting
07:25 Developing AI Tools and Products
09:10 Company Culture and Innovation
16:20 Staying Abreast of AI Developments
30:38 Leadership and Management
33:23 Conclusion and Reflections

📜 Read the transcript for this episode: Transcript of Reinventing Recruiting: Anvisha Pai on Using AI to Automate Hiring |

Episode Transcription

[00:00:00] Anvisha Pai: I'm Anisha, founder of Dover. We are a recruiting platform for the world's best companies. We've helped over 500 companies hire amazing talent. We went through Y Combinator and have raised a Series A from Founder Fund, tiger Global. I've been fascinated with ai. We've been using it at Dover for a few years now. Excited to chat about that. I started my career in Dropbox, uh, feel really passionately about building product to make people's lives better. And especially how AI can contribute to making that happen.

[00:00:29] Jeremy Utley: Just start talking to us about what were some of your first experiences with AI? When did it come on your radar? And what were some of your early experiences that made you curious?

[00:00:40] Anvisha Pai: Yeah, that's a, it's a great question. Technically speaking, my very, very first experience with deep learning was in college. I went to MIT and I studied computer science and I did take AI classes. I was very interested in AI and it was very fascinating because we talked about like the AI winter that we were not in at that time, but had seen before.

And then after that, I think I took a few classes and machine learning was sort of the more interesting area at the time, because there was a lot of work going on in that space. We learned all about neural networks, but I didn't really pursue it because I was sort of like, that wasn't where the interesting work was happening, at that layer.

So that was kind of my first exposure to it. And it's, yeah, just looking back, I mean, it's interesting how like a lot of, People ended up moving in different directions because progress was very slow at that time. This was like in 2012 or something like that. I first came to know about OpenAI and Google Brain from, friends that were working there.

And so I was sort of exposed to it early. Maybe the first time that I really went like, wow, it's really changing and something crazy is gonna happen is when OpenAI started publishing stuff about GPT 3. Um, , so May 2019, 2020 is when that kind of came out. And because I had some friends who worked there, I was able to sort of get access to the playground and things like that.

And immediately, like, my friends and I were like, this is insane. Like this is really, really cool. Even at the, like, back in the time when it wasn't even chat, it was just completions. You type in something, you're like, you know, the quick brown fox. Command enter, just let it complete the thing, um, and , it was really weird and it would often at that time spit out a lot of garbage depending on what you set the temperature to and things like that, but at that very moment, we, my co founders and I got really interested in it and, um, we were like, hey, we actually can think of like already a bunch of applications of this technology at Dover, now we were already using machine learning, but, um, just seeing that we were like instantly like, okay, We should switch a lot of stuff over to this and we maybe got a little too ahead of ourselves because we're like, okay, this isn't like good enough to handle a lot of the use cases that we have.

But yeah, my friends and I even had these like hackathons where we would like create little fun, bots and things like that with GPT 3 so. That's kind of how it started.

.

[00:03:05] Jeremy Utley: And so tell us about your co founders that relationship. I mean, it sounds like a special relationship to go

we got to reinvent a bunch of applications here So tell us a little bit about who are your co founders and how did dover get what were you doing at that time?

[00:03:18] Anvisha Pai: Yeah, absolutely so dover got started in 2019. Maybe like very very early in 2019. My co founders and I have known each other for a long time, we were friends Max and I are friends from college and George is a really good friend of my husband's, so lots of kind of personal relationships there and they are both also really technical and so just to take a step back Dover's mission is to enable every company in the world to build an amazing hiring process and, We built a lot of technology to really automate hiring and make it such that a small company can operate with the same excellence as.

A much larger company with a lot of in house recruiters and resources without having those resources. So, , when we were building the MVP of Dover, we hacked so much stuff together. It was like, one of us was a customer success person going to talk to the customers. One of us was doing sales. One of us was building everything in the back.

So it was just this. iteration loop and we were just really close to each other, always talking to each other. We had this tiny little office in San Francisco. We called it the narrowest office in San Francisco because it was really thin and long. And we would just talk, uh, all the time and iterate that way.

And so, when some of the stuff came out, we. Um, might have actually been training in Tacoma times at this point, but yeah, we were very quick to be like, okay, this is going to help us because we were still very much an iteration and building mode at that time, like, really just building up more and more of the product. I would say that ethos is kind of. Stay with the company consistently till today.

[00:04:49] Jeremy Utley: You said we, we started exploring what are the applications we use it over in turns out we were a little bit early with some of those things.

, how do you define early? How did you know it and what'd you do when you realized that.

[00:05:01] Anvisha Pai: . One of the things about recruiting that's challenging is if you're doing a lot of Like, at the top of the funnel, there's a lot of email exchanges that happen you get applications, you need to respond to all of them, you need to reach out to candidates, they have questions you need to answer back to them, the very first time we saw GPT like, oh my god, this might actually be able to help us automate these Very kind of basic emails that we had to hire contractors to really help with.

And, we built this system where they were like effectively clicking buttons and going through this like logic tree, but it wasn't like, they still had to editorialize their responses and things like that. They couldn't be sending like templates back to people when they asked questions. So immediately we were like, okay, this can be used for that.

And we even rolled out, I think, a beta version of it, where they could click a button, it would compose a response and things like that, but it would just make up so many facts at that time, like, the most ridiculous things that we were like, okay, We can't really use this. Actually, funnily enough, our first use case for GPT 3 was a really weird thing.

It was cleaning company names. So, you know, when you get a resume or like a social link or something like that, it might say Google Inc. Or it might say Google LLC. Or it might say YouTube subsidiary of Google or whatever these are actually all the same company. So, we actually were able to use GPT 3 to do this, like, name cleaning and mapping stuff.

So that was the first use and so that's a lot more streamlined and simple in that way. And that ended up being a very early use case.

[00:06:32] Jeremy Utley: Can you tell us real quick, Ambisha, what's the impact of that? Like, for somebody who doesn't know your business, who doesn't know How much time was maybe spent doing that?

You say you clean company names. What's the kind of bottom line impact?

[00:06:46] Anvisha Pai: We had a massive spreadsheet that we would attempt to like programmatically clean up company names and dump it in here and then a person would go through every now and then and review it. It's a harder problem than you think it is.

Like sometimes you can't do with a simple rule based system because there's some weird, or what if someone has a typo? What do they say? G O G G L E. Out of Google, it's a harder problem than you might think. And so we actually had a person going through to, to check it, so this meant that a person didn't need to go through and check it anymore.

[00:07:15] Jeremy Utley: That's cool. I've heard Goggle is a growing company now, actually. Okay, . Then the next use case once,

[00:07:21] Anvisha Pai: oh

yeah, this actually might've even been the first one. I might've even gotten the order wrong, but one of the very first things we did was we were like, oh, we should build a job description writer. Using GPT 3, and so we, we built that, and at first I was like, we can't launch this because it was so, it would just. It would not really work,

And so we, sort of pivoted into like, hey, this is a fun tool, you know, put it on product hunt, like go play with it, I think I might even have the launch post and back then it was one of more like the first of its, or very early, maybe first of its kind to exist out there, we called it the job description rewriter, and there were these like YouTube videos from People in like Russia and stuff that were like going through and explaining it because it was so so novel at that point Like that there's this job description rewriter We'd get a ton of people going try to use it and then like they would write into us being like hey The output was like not what I expected and then we'd have to be like, yeah, we know it's kind of like a toy Just out there But now today we have an actual job description and career page writer.

That's built off of GPT for I think And it's in production and I hate saying this, but a lot of the customers just use the job description without even editing it. So today it actually works.

[00:08:38] Jeremy Utley: So talk, talk for a second. I mean, about the luxury of being able to put something out there that people write and say it doesn't work.

And you can, like, you're laughing this whole time, right? There, there are some companies who, if they heard their thing didn't work, it would be like existential crisis. You know, somebody's job is on the line. Can you talk about the role that, you know, maybe being an early stage startup played in your ability to kind of test and fail and take these kinds of things in stride?

[00:09:09] Anvisha Pai: . So I think that for us. It's not really been on the critical path so far, right? Because we built our business before generative AI existed. So it wasn't the moneymaker. Um, and it wasn't like it needed to work in order for us to make money. Like, we did all of it before, right, somehow, and we were still making money at that time and we were, I think, at that point in a pretty good place where we had, exceeded our first million in ARR we had traction, we had funding, we were in a stable place. The JD writer wasn't going to make or break the company. It was just a marketing thing,

[00:09:44] Henrik Werdelin: I mean, , I've been in that situation too, in companies that I've helped build, and if there's kind of three things that you need to think about when you think about adding AI to your organization and you roughly divide them in people, processes, and product, I realized that the product is probably one of the last thing you actually play with, probably because the technology moves so fast that it's very complicated. to put anything in production right now, because, you know, a new version of the LM will pop up, but what you seem to have done in your organization, which I find fascinating is to really allow everybody to feel they have permission to use it and not go default to. Oh, you know, where does my data and what will happen, like this, like very anxious kind of space or if I do this and you're the person changing goggle to Google, then, you know, like I'll lose my job because that is what I do.

Do you think that you were just fortunate that you. Kind of started around the time and so it just became something that everybody felt comfortable about. Is it because you guys are all engineers and don't feel scared about it because you understand what it actually is? Where do you think you guys are different from the millions and millions of other companies that realize that GPT 4 is out there but don't use it at all?

[00:11:01] Anvisha Pai: Um, I think the timing was good. So I think we had built up the company to enough of a point where we had product market fit. And we felt like we were able to experiment with this new technology. At that time, GPT 3 didn't really work out that well. Um, and It was still early enough that we were a small team.

Everyone was very technically minded. So, obviously, there were engineers and co founders who were technical. But even our, customer success support team, very analytical people we taught all of them how to use SQL and stuff like that. One of the things that we did a lot of is we have these company hack weeks.

So we do a one week long go somewhere, get a bunch of rooms in a hotel, or get a massive Airbnb, work on projects. So a lot of people worked on AI projects during those things. So everyone understood how it worked, and we also did workshops where we taught The non technical people in the company, like, hey, set up Repl. it, hit the GPT API make this basic web page. So we try to get everyone comfortable with it. But I think there's something else too. So I think that, like, with us, um, at Dover, we have this sort of, we like to try things a lot. And, accountability is important, but if you try something, In earnest and put your best foot forward and make an effort and it doesn't work out your head is not on the chopping block, I think you have to be okay with people failing or things being not 100 percent there as long as they went in with the best of intentions.

[00:12:38] Jeremy Utley: How do you know just on I love, you know, mechanisms like the hack week or workshops, things like that. How did you know whether that was a useful use of time? So six months from now, should we do it again? How did you answer that question?

[00:12:51] Anvisha Pai: I think for things like that, it's very hard to measure success, to be honest with you.

I think it's just almost a religious belief, I think this is why a lot of bigger companies don't innovate is because they feel like they have to justify everything with metrics. But sometimes you just know what's good. You know, it's good. Like if everyone knows how to use. One of the most significant technologies of our time like that's a good thing.

You subjectively know it.

[00:13:16] Henrik Werdelin: We've been fascinated by Kenneth Stanley's book why greatness can't be planned. He's an AI researcher used to be Florida then Uber AI labs and then open AI and now he has his own startup but he has this kind of great point, which is basically all systems and most innovations that have been really incredible word made in these open ended system.

And you have to pursue what he refers to as interestingness and that kind of will give you the stepping stones. I do find that's like such a fascinating kind of thing because most non founder led companies have a very tough time kind of getting their head around that principle. Which is ironic because that's often the thing that makes you do something incredible.

[00:14:02] Anvisha Pai: Yeah. And that actually reminds me of one kind of related thing. Like I think it was Paul Graham from YC who said every really innovative, really disruptive thing kind of starts out looking like a toy. People don't take it seriously. I think GPT 3 was kind of like that. It literally started out looking and feeling like a toy

[00:14:21] Henrik Werdelin: I remember that playground and because it looks so basic, it was just for people who were not in the early beta. It was literally just like a text field. And then, and when you logged into, that was all there was. And so the first many times you were kind of like just trying to go hi, run, and you're like, just see kind of what would come up for it.

And there was even like something fascinating that this very intelligent machinery was in this UI that could not be more basic. It was a text editing field.

[00:14:54] Jeremy Utley: Just to share a little nerdiness, I remember my first real like giggle moment was giving the special instructions that I want you to respond to everything as Mr. T. And then in the playground and just, I remember being in the playground going what can I say to make Mr. T mad? And it was just this, and I actually, I took photos, I remember, you know, taking photos and sending my family. I'm talking to Mr. T on the computer and it's hysterical, you know, and my family's like, you're talking to Mr.

T? You know, it's just, what? What are you talking about? But you're right, such a basic thing like a text field for there to be so much joy and delight that comes from that.

[00:15:32] Henrik Werdelin: One thing on the belief system, which I found fascinating, which we talked to your colleague about at one point was, that sometimes you built something and then the technology is not quite right, but then you wait a bit and then you go from GPT 3 to GPT 3. 5 and then suddenly the thing that you had in mind is doable, but you probably wouldn't have done it if you hadn't kind of like been exploring it early on. And so is that kind of principle of like, let's just. Try to kind of like come up with something do you have like a method of then going back when suddenly like , A new version of an LM model kind of pops out to think about what we should actually go and try this thing that was in a hack week a year ago.

[00:16:19] Anvisha Pai: . So, some of the people on the team, like our early, like our founding engineers, our CTO, myself We made it a point to just keep in touch, like, really keep abreast with new stuff coming out. And, a lot of that honestly was via Twitter. So I think once you kind of like entered into the ecosystem of generative AI, it just seemed clear, like there were improvements coming out.

I think Chad GPT was like kind of the crucible moment and GPT 3. 5. And then it just really started accelerating. And so we just saw it happen so many times that we were like, okay. And, I don't think that's anything that. Outrageous when it comes to software engineering, because there are always improvements, like the cloud infrastructure is a good example of this.

Google cloud is always coming out with newer, better improved infrastructure for us to use. And we've been really, good about looking at it and being like, okay, should we adopt this? Like, okay, there's this new version of Postgres. It has this thing, like we should probably get on that, you know?

So I think that it was a habit that we had developed and that it was pretty natural to. Latch onto it. I will say maybe the hard part is you don't know how good it's going to be or how good it's going to become. So I think that's a bit of a wild card and like a little bit of a guess that you have to make.

I wouldn't say that we're in the mode yet that we should like ship something and we're like, oh, we know it's going to get better. And we're just going to bank on that fact because you can't really do that. , I personally definitely have this belief that like we are going to build AGI and.

These technologies are going to get better and better, and it's going to get scarier and more formidable and more powerful what they can do. So, we are kind of building in that overall worldview.

[00:17:59] Jeremy Utley: Can you talk for a second about what your habit, you talked about developing a habit to kind of stay abreast.

Today, what does your habit look like?

[00:18:09] Anvisha Pai: For me personally, it's just been like following a lot of people on Twitter and trying to keep my Twitter like relevant professionally so I'm seeing the new stuff that's coming out and then it's also been going to events and Honestly, like even creating a little bit of a community here in New York, where I can meet and talk with people to see like what they're seeing, what they think is new and valuable, and then obviously kind of where you can being in touch with the folks at OpenAI whatever, being users of their products and trying to keep abreast of what they're releasing whenever they have a beta or a new thing where like, hey, we want to try this thing

[00:18:42] Jeremy Utley: and then so maybe you could even walk us through the last thing that you saw on Twitter. What happens? I would love to know the habit workflow, so to speak, reach out to the CTO. , do you copy the link? Like do you push yourself to think of what are all the ways we could use this? Like what is the habit? You expose yourself to new information. Then what, what happens next?

[00:19:07] Anvisha Pai: Yeah, I mean, I'm not the sort of person that reaches out to a lot of people to just network or Pick their brain, like I think I don't do as much of that. Maybe I should do more of it. But usually I see something, if I think it's interesting, read a little bit more about it share with our team.

So I'm like pretty good about when I see something I'll give you an example the other day I saw a new application come out that is really, really good at converting like a Figma code into react components. And I sent it to our designer and I was talking with him about, like, potentially using that and getting him up to speed so he can do a little bit of front end development.

So that's the conversation that we're having now. So I think it's just , bringing that idea to life a little bit by sharing it with someone and starting a conversation is typically my workflow.

[00:19:54] Henrik Werdelin: Do you manage the, maybe you guys are not at the size yet, , in the sense of there's probably endless amount of things to do and not enough hands to do them do you have any of them that gets concerned when suddenly you ask the designer to do some code and the developers are going, hey, wait a minute, what am I doing to, to get any of that anxiousness Is it still a company with the spirit of curiosity and just let's Try to get as many of the to do's done as possible.

[00:20:22] Anvisha Pai: I think we're still at the stage where. The engineers would be very happy if the designer could help with some of the front end component design, you know, like, I don't think they want to be doing that,

[00:20:37] Henrik Werdelin: A lot of the stuff, I guess, for many people is that the thing that the computer is very good at doing is stuff where the person is acting like a computer and most humans don't really enjoy very much, like the monotony of doing something over and over and over again,

at least that's been my experience, , you're not really taking the interesting work away. You're taking the, the kind of like coat monkey, uh, kind of work away.

[00:21:02] Anvisha Pai: As like a related story. So I have to credit one of our early engineers, Sid, with this. He actually ended up leaving Dover to start his own company. And he's building a LLM Ops framework called Vellum. I'm going to shout them out. Very cool platform. Sid got us to use this tool called tab nine. Which is github copilot before github copilot. So before that even came out, our engineers were using a code completion tool to be more efficient and I think that, you know, then we obviously switched over to github copilot.

Everyone has a license and they're using that. Um, but, Yeah, I think it's just been a culture of like, this is a good thing, be more productive and it's not just us, I think the whole industry is hopefully moving in that direction. Like, if you're an engineer that's not using GitHub Copilot, um, like, What are you, you know, why

[00:21:55] Henrik Werdelin: do you have areas where you think the copilots for X is not going to come anytime soon? I saw a tweet today about that had reduced. I think. 700 people's of like customer support. And I think every day, increasingly, we see like those stories. Do you have like areas where you think, yeah, we tried to look at something of that, but that is not going to get there. I guess, obviously believing in artificial general intelligence, like that's a hot statement, but like, as we get to that where do you think we won't see AI for a while, if any places?

[00:22:36] Anvisha Pai: Oh, I don't know if I'm qualified to answer that question. I mean, I can,

[00:22:40] Henrik Werdelin: you know, one thing we've noticed is that we've done these interviews quite a bit and everybody who's an expert, not in AI, but in applied AI, I've only been at for two years, right? Because until dbt three came around, there just wasn't really much you could kind of like easily apply this to. And so with that, I think you can qualify you. Nothing else like a policy expert.

[00:23:05] Anvisha Pai: Yeah. Well, I mean, I believe that AGI is possible. I think that there is probably nothing that AI cannot do at some point. The limiting factor will be pretty much things that we prefer for people to do.

It's like one example Today, right now, AI models are definitely good enough to conduct a first round recruiting interview with a candidate. You can use a combination of 11 labs and, GPT 4 or mistrial or whatever. You can have a conversation and run an interview. You can 100 percent do it.

Are candidates going to want to be interviewed by an AI? Are employers going to want to represent themselves with an AI? Like, they don't want to yet. So it's pretty much limited by our appetite, I think, really. So I don't think there's any technical limitation to like what is possible. It's going to come down to what we are comfortable with.

Maybe the medical profession might be another one where we might not be totally comfortable with completely interacting with an AI doctor, even though it's perfectly capable of providing that service to us. We might still want a human to make those decisions.

[00:24:13] Jeremy Utley: And I think if anything, from my understanding of the kind of Present research, we overestimate the degree to which humans are uncomfortable interacting with AI.

Like I saw a recent study where humans actually prefer AI results or responses. Humans prefer AI. Doctors . And so I think that our intuition is, Oh, humans are distrustful. But then when you start digging, it's like actually not here. And not in this other place. And so I, that'll be an interesting frontier.

I want to shift towards company culture a little bit. 'cause we've been talking about kinda what you do, your personal practices. How do you take somebody. And train them technically to where they can be proficient to see and execute opportunities. Maybe that's an interesting place to start, but I'm just curious about what are the broader kind of organizational mechanisms for seeing and seizing opportunities beyond your personal habits?

[00:25:09] Anvisha Pai: It's really interesting actually, because like I mentioned, in the early days we had contractors that were helping us with various. Recruiting tasks, responding to emails, scheduling interviews, things like that, because we hadn't really automated a lot of it yet.

And some of the people we hired to do that totally randomly turned out to be really, really good. And honestly, like what we just did was we like identify, like we recognize the talent and we were like, we should just give them more responsibility. And then they took on more responsibility, took on more responsibility, they came on full time., we just gave them more and more. . So it was really just like identifying talent and giving them the opportunity and then elevating them when, they're able to sort of do it. And so that's how he ended up with a lot of folks.

Like that on our team where we would just bring them in, like a lot of kind of younger people people going through a career shift we, we grew a lot during COVID, so that's why a lot of these folks were affected because their entire performance career was put on hold and then,, I think if you just give people the right opportunities and And it's not like we gave, we're a startup, like it's not like I had a ton of time to sit down with someone and teach them how to do SQL or something.

I just gave them some resources and was like, go and ask me if you have questions. And the people who are able to do that are the ones that thrive. And yeah, I think that that's pretty much all there is to it.

[00:26:23] Jeremy Utley: Can you tell us a story about someone? identifying an opportunity to bring AI into the product. Cause I think one thing that someone listening might think at this point is, okay, if I studied machine learning at MIT, then I can really be prepared for the next, technological breakthrough.

And I know that's not true, right? That there are that certainly your background and training. Um, and I think you're also cultivating an environment where others are increasingly able to appreciate opportunities. So can you talk through how those opportunities have worked?

[00:27:02] Anvisha Pai: Um, yeah, this isn't necessarily AI related, but the way that we've approached building product is we've been very hands on with our customers and we've done things that don't scale and then we identify problems and we're like, okay, five different customers have this problem, they're willing to pay for it.

We've done it manually. Here's a repeatable version of this process. We can productize it. And so we've had many people at the company over time propose and implement pretty massive changes to the platform. And this is not an AI example, but like one thing that comes to mind is one of our product managers a while ago built out an entire onsite scheduling product.

And he really drove that himself kind of had a team that he set up that was doing it manually. Then we, like, started building out a lot of it into the product, um, and it worked. And so, um, we were like, okay, let's. Let's go with it. Similarly, another person, not even in product and like business operations or something built out an agency management tool.

That was pretty interesting. A lot of customers were using it. We ended up actually, like some things don't work. So we ended up killing that because it wasn't universally helpful. But it. It was a really great experiment. , I think when you just hire people who are self starters and you give them permission to try things, then they will try things.

[00:28:22] Jeremy Utley: How do those, I mean, I realize this will be an over engineered example, but, Amazon's PRFAQ comes to mind of, somebody sees an opportunity, there's kind of a practice of, Pull together the press release and what are the answers to these top questions? Is there any kind of mechanism like you take either one of these examples, the deprecated one or the onsite scheduling tool, somebody has this idea, you've done it a few times.

What do they need to demonstrate in order to get resourcing to product ties?

[00:28:52] Anvisha Pai: Um, honestly just writing a doc, it doesn't have to be in any particular format, the. format that I like to see it in is like what's like the problem statement, like what's the customer problem, what's the problem slash opportunity for Dover, and then what are some ideas for solutions, like rough cut, like how much effort is it going to be what are the risks associated with it, if anything.

And yeah, just need to write a doc, talk to one or two people, and then it's sort of good to go. We don't have a lot of bureaucracy, unless it's something that's really risky or might have a big revenue impact, then we'd like to do like a review for it. But we're such a small company. I used to work at Dropbox and Definitely things were harder at Dropbox to get buy in. Like, you probably have to shop around an idea quite a lot. Um, and there, there wasn't necessarily, like, an Amazon type of, like, format. You could just throw something out there and it would, like, get looked at, you know? But it's so much easier in a small company. So

[00:29:47] Jeremy Utley: it can be more informal.

[00:29:50] Anvisha Pai: Totally. I shouldn't be more informal. Otherwise people are going to not do it

[00:29:54] Henrik Werdelin: you done both Dropbox and now you own, I'm obviously asking, cause I've built companies and we were three people and then we went public and then we were hundreds and hundreds of people. How do you think? You will maintain this as you grow because there are structures to find outcomes and there seem to be in most of the companies that I'm sure you and I have kind of been involved in, there seemed to be this kind of gravitational pull into kind of like the behavior that was more what you experienced at Dropbox than it was what you're talking about now. Have you done any thought on like, how do you resist that gravitational pull?

[00:30:36] Anvisha Pai: It's a good question. What I would do is like I admire certain companies like I think Uber in the Travis days, Stripe have really been able to keep that culture while they scaled of high performance, lots of self starters taking initiative. I really look to those people to sort of understand how they did it.

I do know that one very, very important thing is the quality of management and leadership that you have in the company. That's maybe the most important thing because ultimately the founders can only have that much of an impact past a certain size. So, making sure you're hiring or growing the right. management layer is pretty critical.

[00:31:17] Jeremy Utley: How do you define the right management layer? And how do you know if you're growing them?

[00:31:22] Anvisha Pai: Yeah. I think that there's like many ways to look at it. I mean, obviously there's a quantitative ways, right? You can always look at someone's team's performance and the objective. I think the thing that a lot of people miss, especially a lot of earlier stage companies miss, is if someone's team is quantitatively doing well, but qualitatively they aren't embodying sort of your culture or your values. What do you do about that?

And I think we've been in that situation before, I think it is very important , to make sure that someone aligns also culturally, not just like quantitatively that their team is performing well, because we have seen that go wrong , in the limit. And with the qualitative stuff, you kind of find out pretty fast if someone's not aligned. So maybe my advice there would be to sort of Especially as an early stage company, because it's so important, like, trust your instincts if you bring someone on and you realize that culturally, very quickly, they're not fostering the right set of values in their team, it may not be the right person.

[00:32:28] Henrik Werdelin: Just to add to that, at Bark, we had the rule until we were probably 300 people that you had to meet a founder before you got your job offer. And what necessarily that we were like master interviewers at all on the contrary, but I think it gave the people that were hiring kind of like a moment to pause and just think about like, how would this person fit culturally and we would stop candidates sometimes if we didn't think that was a fit because obviously, like, when you're busy and you have an open head count, you just want to fill that, roll pretty fast because then you feel that you'll be less busy.

So I very much kind of echo that ethos. It's something I think meant a lot to us. We're super, super grateful for your time.

[00:33:13] Jeremy Utley: Yeah, you're amazing. Thank you.

[00:33:16] Anvisha Pai: Thank you. I appreciate it. Take care. All right. Bye. Bye.

[00:33:21] Jeremy Utley: Henrik Werdelin. What'd you think? I, I know you loved her. I know you did.

[00:33:26] Henrik Werdelin: I mean, I like product people a lot and specifically somebody like her who is such a system thinker, but then at the same time really have very high. EQ, and so understands how to think both in systems, but also appreciate and value what humanity can bring to the table. And so, that's normally a hallmark of an incredible product person. So kind of obvious why the business seemed to be doing well. Yeah, really enjoyed the conversation.

[00:33:54] Jeremy Utley: Yeah, for me, the thing that struck me as a kind of a pragmatic takeaway, and this seems obvious for folks who are in the hunt, but for folks, maybe who are trying to dabble or explore applications of AI developing a habit. Of reviewing what's new and finding ways to stay informed, whether it's via Twitter or newsletters or podcasts like ours or whatever, finding a way to put yourself in a position to see new stuff and then take on the responsibility to share it and discuss it with folks you work with. I think that habit. Is essential to because the field as you've often said the field's changing so quickly If you don't have a routine of checking in and then sharing what you learn, you're gonna miss real opportunities

[00:34:40] Henrik Werdelin: Yeah. And I think just to add to that, like the whole classic input in, input out, you know, I think comes to this, if you don't consume what's new in this space, then you obviously have no idea of getting inspired and figuring out how you can apply it to your own organization. So, I'll very much echo that. The second thing that I think she mentioned, which, you know, I really enjoyed was that this idea of having a internal culture of not being too afraid to try a few things. And so they talk about the hack week and stuff like that, which, obviously is an obvious thing to do, except very few companies find the energy to actually then prioritize it.

[00:35:19] Jeremy Utley: It seems so hard, right. To take a week away from operations.

[00:35:23] Henrik Werdelin: Yeah. , but even if it's like a few days, that seems to probably be like super useful. Right. I think, and it is probably a earlier state startup thing, but this idea of finding very entrepreneurial, very interesting people, and then find a onboarding ramp that allow them to work with the company. And then if they show to be remarkable, find a way of accelerating them into the organization. Just as such a fascinating kind of way of thinking about hiring where. Obviously, as you become bigger and the company become more serious, then you have this job spec probably written by AI and then people have to fill in all these kind of different elements. And so for example, I love, love, love the idea of having staff that go and leave to make their own startup, because that is the ultimate proof of you having entrepreneurial talent in your organization.

[00:36:17] Jeremy Utley: And the fact that she celebrates it, she's not mourning it. I love it. ,she gave Sid credit for the amazing thing that he taught them. And then she wanted to hype his new company too. I mean, to me, it's like, that's, that's a wonderful ethos and embodiment of that spirit of , she loves entrepreneurialism so much that if it's inside or outside of the organization, she's going to celebrate it. That's it. Thanks for joining us for another episode of Beyond the Prompt. If you enjoyed this conversation, please share it with someone in your life who you think needs to know Anvisa and the tactics that she explored. And don't forget to give us a review, an honest five star review. Thank you so much.