CEO's guide to GTMmaxxing! Filip Kozera: lessons from YC & a $30M Raise transcript

Hello everyone. Hello. Hello. Hello. [ __ ] Russian bus. What's that? Just Okay. That's just foreign. Okay. All right. >> Yes. All good. Okay. Perfect. >> Okay. >> Are we speaking Polish or English? English. All right. Welcome everyone. I'm very sorry uh about the uh sitting situation. There's some extra blankets in the back. Uh, I'm not gonna trust this guy next time with counting people uh for location. >> There's just too much interest, man. >> Yeah.

Yeah. Yeah. Uh, cool. Let's get started. And there's some extra blankets in the back room. So, if you don't have anywhere to seat, just please don't be strangers. Come to the front and we will fit everyone in. And if I could ask the organizers to um like turn on the AC a little bit more. Okay. It's going to get a little hot. and mute this one. >> And could you just mute this one in the meantime? >> Hello. Hello. Hello.

Okay, I'll do it. All right, let's get started. Uh if anyone wants some drinks, uh just show your Luma uh invite and uh the bar is open, but I heard there's also no alcohol, so a bar is a big word for for that. Let's get started. >> No. >> Uh so thanks all for coming. Uh Philip, first question to you. uh your first uh AI company took like 100 118 investors where nobody wanted to invest. You got rejected with YC sometimes but then suddenly 30 million.

What changed you? What changed with your team? What was the story? >> All right. So, uh very long story short. >> So, it was your laptop. >> It was my laptop. Yeah. Okay. So uh very long story short, I've uh studied at Cambridge. I was a little bit lucky with studying uh what turned out in the end um changed itself into transformers. And I see some uh friends from the from the Cambridge lab. Um and straight after I went to San Francisco, I only had $3,000 in my pocket and big dreams, but not many contacts from um from anywhere really.

And the magic of San Francisco is that you still can get essentially over a coffee you can get an angel check. And at the beginning, you know, it's good to apply for one of the uh one of the accelerators. I applied, I think in the end, four times for YC and got rejected three out of four. Um and um at the beginning when you're trying to, you know, make it work, it's hustle. And um some people believe that when you're raising around, you just build a good product and it all works.

And actually these people, if they're not, you know, booming in the world, end up not raising. And there is a set of things that you should do when you're raising and you know you're not the hottest thing in San Francisco or your product is not instinct which is blowing up on um on X right now. Um and it comes down to process and also to building FOMO. I think a good process of of of kind of raising looks a little bit like dating actually.

So, uh, you kind of have to, uh, build up the, uh, the excitement, you know, and you shouldn't actually give away all the information straight straight away. Be a little bit mysterious and, uh, and and use one of the techniques. And so, one of the easiest way to to raise at the beginning is through angels. And never ever ever say the full amount uh, that you're raising if you are unsure you're going to be able to get there.

If you already have some offers for $2 million, then that's fine. You can say, "Oh, I'm raising $2 million." And that can climb later on. But if you have no clue, try to say a number that you're kind of like 90% sure you will be able to actually get to. So sometimes it can be as low as 100k. So you need to build up this thing. Everyone feels that they are getting a deal of their lifetime uh when they come to the meeting with you.

There is a very very simple technique of how to close them when you really don't have much. So um sorry am I going too detailed on raising? >> So we're gonna we're gonna play a little bit of a game here. Okay. >> All right. So, you're going to be an angel who is very uh who basically is not sure whether they're going to invest. We we're on our first or second meeting. And I think one of the the biggest things here that comes from dating is if you're on a party and you're never okay with rejection, you'll never get anyone's number, right?

So, you kind of need to be very comfortable with hearing no. And uh let's say the meeting went well. It's our first or second meeting. You're feeling the excitement, right? So, you're going to ask your angel, potential angel investor. I think you need a mic as well for the live or no? >> Oh, I Okay. So, can I hold it like this? Hopefully, they will hear. Um, so what's your typical check size? >> Uh, 50k. >> Okay. Um, so how are you feeling about this?

Do you think there's a potential for uh for us to, you know, partner on this? Uh who else is angeling? >> Uh it's a you know a combination of people but firstly before we kind of go through the list I would want to hear whether you would be actually somebody who wants to believe in me and without you know other people on board would you go for it? >> Uh I'm I don't know I mean you're a first- time founder. >> No of course.

So just tell me you know what would need to happen for you to say okay that 50k is a no-brainer. Um I don't know Daniel Gross on board. >> Okay. So what I'm hearing is that you would like some other angels to come on board that you value a lot. Are there any other conditions that you can come up you think about? >> Uh yeah I would like to see I would like to see like your team composition okay dynamics between you and your co-founder. >> Perfect.

So uh we can actually uh you know jump on another call but I'm going to keep this as a conditional. So, what I'm hearing right now is as long as you're happy on that meeting after we grab a dinner together with my co-founder and a persona and angel of the sorts of Daniel Grath is on board, you would be uh committed to 50k check. >> Yes. So you kind of want to go through getting rid of the objections and once you you kind of get there you say oh so what I'm here is you're committed on this and this and then very logical thing is would you mind if I share with other people that you are committed based on these um based on these conditions that you formed >> yeah no worries >> so and that's how if you have a shitty round you build traction because right now I'm going to spam everyone saying oh 50k committed out the 100k that I'm raising, right?

And you have to build a good technique for keeping people in the loop when it's not the hottest round in the world. And you then go to the next person and I say, "Oh, you know, oh, have you heard of Arthur? He's like, you know, a pretty big persona in the AI space in Warso." Yeah, he he's, you know, committed on 50k and would you like to, you know, I'm only raising this 100K, so it's going to go quickly. I only I haven't really started raising yet really.

And then they're like, "Oh [ __ ] there's only 50K allocation left." And you know, then you've got the 100K and you're like, "Well, my commitment, you know, I I thought I'm doing 100K, but there's so much interest that I think we should uh I'm going to extend it to 500K." And you kind of have to build up this machine of hype. And one lesson that I heard from a from a very very big CEO who is in a tens of billions of dollars is if you are not playing these games as CEO you're just not doing well enough is something that you should use to your advantage and if you're not doing that you just have less of a chance of success.

So that's kind of the more we can dive deeper into these kind of like very very practical lessons of how to raise your round. Um because the uh our latest round was the opposite of that. Uh we made people come to our office. We made A6 and Sequoia come in and if they said we're meeting at your uh at our offices, we're like well sorry. I think that's not going to work for us. And you play it in a very different way when you're doing it from a position of power.

But I think what's what's really important is that uh you know those things you can just learn these techniques and um I'm gonna share with everyone on the on the event a guide that I wrote with all of these kind of little little things of how you should run it. >> Uh okay. So we uh we have the dating component in in place but now I would like to get like you know before after after you start dating uh you need to have a mortgage together or something like this.

So then when you got people interested uh what mattered for them uh at the fundraising because you already have the story you already build the FOMO but what was true for your this 30 million rounds uh was it metrics was it uh viral growth look what really them interested in what they care about because story is like everybody knows you're playing games right so then you need some kind of a from other angels committed and other funds committed you need some numbers to buck it up.

What was that in your case? >> So, um I think in case of of uh of that round, we I think when you're starting as an entrepreneur, you often feel like there's going to be one event that is going to make all of this easier and everything will line up after. The truth is it's not like that. It's it's it's a the opposite of a death by a thousand cuts. You should just do a lot of things very very right. And in this case, you know, we um had the most upvoted product hunt launch ever.

We like racked up 9,000 votes. We had some one of our products that was built on our platform get to 10 million users within the first two weeks. And we also successfully managed to uh change some of that interest into real revenue. So we got to like 800k ARR uh with like eight days. Uh so you like it is you know I sometimes feel a little bit bad talking about this particular round because I never want to when I'm doing these kind of speeches or you know talking to entrepreneurs we all know it's a bunch of [ __ ] that you have to go through and I don't want to make it seem like oh yeah just everything happened right and I prefer sometimes to talk about the rounds where I had to do the 80 meetings or uh or stuff like this it just it turns out that uh you know throughout my 8-year journey of startups I managed to have one of them that went incredibly well.

Some people get you know uh other people you know they just get by and they use the right techniques and they just raise the money as they keep going. I think the the world has somewhat changed in the last year and a half. And the really really important thing right now is everyone of you guys, everyone of everyone in the planet and sign up for $100 on GPT um on GPT on codeex use Astra and build their product and sell it.

So revenue is pretty much a crucial aspect of every race nowadays. If you cannot deliver on that, that basically is a reflection of you as a founder. Uh it's not you can always get to something uh with very little funding which was not possible before. >> All right. So now we have uh we have the revenue but also behind your round as I understand talking with Sebastian from your team and there is a legend uh in Poland about wordware blitz and basically the the story is that like the guys has stacked every single investor meeting inside seven days period and the it's super important to to like have this FOMO do it in a well structured mannered and this is something not really common in Poland.

I hear a lot of people, oh okay, I've met with this one VC. I talked with some VC. Maybe in couple of weeks I'm going to talk with another one. Maybe I'll get a term [ __ ] How did you structurize this blitz? How to do it effectively and why this is so so powerful? So this is actually one of the practical uh aspects of this. There's a couple of different techniques for running these. Uh one of the most popular is called horse race.

It means that uh you are doing a horse race which means you have to open up the the gates for each horse at the same time. So that means you uh you should uh essentially have everything ready do the right research um have all of your intros ready and suddenly on all of these WhatsApp groups of all of these VCs there should be a storm of everyone talking about your round and in that way you know everyone is like oh yeah wow they they're everywhere you know they're going to close the round too soon.

Um and just for the practicality aspect of this, I'm going to cover one more technique. It's called the rabbit race. And uh the rabbit race basically means that you have one VC uh that you know very well. Sometimes you don't really want to take money from them, but you know they're going to give it to you. And that essentially means that you start the process couple days before and only then everyone is chasing the rabbit.

So often you don't have to say the name of the firm but you hint that you are much farther in process with one of these um one of these players and that essentially means that everyone has to get their [ __ ] together very very quickly which means that they have to you know do their due diligence um very quickly. So uh when we were finishing up um when we were finishing up YC uh we were about two months and a week in a whole program is three months and normally YC advises you to raise only in the last week before the demo day.

We didn't want to do that because we wanted to get to know um the people who were uh you know we are going to partner. We knew that our round is going to be pretty hot. We didn't actually know how hot it's going to get. Um, and we basically had a list of 25 people. So, not you know the 200 that you should be aiming for if you are doing a kind of know a round that you're unsure about the um outcome of it. say 200 PCs, 200 PC and angels I would say with good intros to each and ideally multiple intros.

Um, and you know it's sometimes good if you know like some people with money then sometimes get a small check and it's the best way to make sure that they are going to do a good job introducing you to all of their fun all of their funds on their cup table and so on. So like I don't know if this if this works in uh in Poland. Is there is there enough angels in in Poland? Okay. So maybe yeah the best way is to get other founders on your cup table before everyone else.

They're going to help you and they're going to introduce you to their whole cup table. I think there is there's people who are coming in this. Um and so coming back to the story we we knew we had a round. We knew that it's going to get pretty cool, but like we didn't know what's going to happen with it. We timed everything for these people to come to our offices over the next seven days and we employed some tricks. So when Sequoia sent us pastries or we're getting pastries in the in our office every day from from Kofla from Sequoia, I for example made sure that on the box from Sequoia pastries I put it in the conference room and I wrote myself uh with love from Sequoa.

So every time an investor would be opening the the box with pastries uh you know they would be like oh oh okay. And I'm like oh that's so awkward. I'm so sorry. So, uh, you know, you have to play the tricks, uh, somewhat. I think, uh, I've never actually said this story, so hopefully this, uh, where is this going? >> Uh, it's only live on YouTube. >> On YouTube. Okay. So, it's not on LinkedIn. Okay. So, because I don't want some people to see this.

Uh, oh [ __ ] Uh, >> all right. Let me let me take >> No, I mean I like what if I would be speaking Polish, I would be able to say this. Uh >> yeah. Okay. Uh very very quick story in Polish. I'm going to come back to English in a second, but this is just um Okay. Um, Okay, let's let's talk business. Uh, what are you guys gonna offer? New Okay. Are there any side letters or are you okay with safe? Okay, I'm gonna >> uh essentially the clue the clue is that uh you often don't know your valuation.

So don't say it too early um because the market sets it. So every time you're asked how what's the valuation you know just leave it empty be like whatever you're raising okay if it's preede you say 1 million you say oh you know as this is like a typical preede I would like to give away 10%. You know these investors are not dumb. They can multiply one by 10 and get to 10 million, right? But the really important part is that you never say it and you get the percentage worked out first.

You know, if sometimes you have to take what I say with a pinch of salt because if somebody will be pressuring you and Polish um like Polish real life is a little bit different to Silicon Valley like don't be an idiot. Don't like you know just say Philillip told me never to say evaluation. amount of I'm in trouble but most of the time it's very obvious like you say 1 million we want to give around 10% away but if there's a lot of interest in your round that number can go up and if you said oh at this round we want to give 10% away and you're suddenly raising 2.5 that valuation just jumped from 10 to 25 and nobody even but an eye because now you have a hot round and uh suddenly you know like oh yeah 10% oh yeah but now we're raising 2.5 so it's 25 everyone just remembers is the 10% uh figure. >> Uh cool.

So you got this Okurva moment and uh you realize that there is there is a potential. So what happens uh in such a short period that you did it? Uh what happens from the moment you have the first term sheet to finalizing this valuation because as you said it's going to increase and increase and increase if it's hot. Uh but what are the step from you know being committed to actually setting up that this is 30 million on the X valuation.

When do you commit? When do you finish the fundraising? Uh because you cannot do it like forever, right? Because the the FOMO was going to out and you need to focus on product. So essentially after we got this offer uh we actually went to slightly different investor um the spark capital which uh did um the seed uh deal for Anttopic strip and we really like the partner there. So uh we actually had a list of preference of which partners were there.

Felic is also great. We love working with Aastasia, but this guy basically had a little bit more relevant um PLG motion experience. So, you know, at this stage, you're starting to play chess, right? You know, okay, part of this is money, but a big part of it is also exactly who you're partnering with and on what terms. And you know, as a as a founder um above 100 million, you are also trying to take secondaries. So um this is something that people don't talk about but you know we are risking a lot and it's important to try to take some cash um off the table to kind of manage your own risk and your own personal life situation.

You know sometimes you know I'm I'm turned 30 recently and you start thinking like [ __ ] I've been in this game for a while. I'm not like I I need to make sure that I can buy a house at some stage. um and you start negotiating these things and there's multiple things that you can negotiate as a priced round. So if you have a hot round, you're basically get a chance to rewrite some of your company setup while you're doing the priced round.

So if you have a hot round, try to do a price round. If you don't know what's going on and you just want the money in your bank, try to close it on a safe. We actually had an offer to close 30 million fully on a safe, which would have prevented a month of due diligence. So price round if you have a hot round first of all you're going to convert all of the previous saves into real equity which is important for your investors as well you get something which is called QBS which is like tax-free amount but it's also cleaner like you don't end up having multiple of these saves that are just randomly hanging out and you kind of need to bite the bullet at some stage and convert all of them.

Um, if you're not sure what the hell is going on and you just want to close this money, you can do a rolling safe. So, don't try to close it upon one event, but just keep closing them as you go. So, don't build commitments. At some stage, start asking for real signatures. Um, for us, what that meant is that we got the first offer and we went to talk to a bunch of um other investors. We received one more um offer to beat this valuation from a very very very large fund uh 15 billion fund.

Um and they are normally trying to put in money uh to uh kind of have information rights. So you can essentially when they are giving you an offer they hope you're not going to take their offer which is you know a little bit of a a dance because they they hope you're better than that essentially but you know, gave you this money and for them 30 million out of a 15 billion fund. Who gives a [ __ ] It's pennies. Um and uh we then went to the investor that we really kind of, you know, valued for their product sense and we told them we've got this offer and he actually um put out an offer which was significantly lower. uh and then we negotiated all the terms still ended up being lower than that valuation we got offered first.

So if we would have optimized just for valuation, we could have probably pushed it all the way to maybe 300 uh which was the double of the first valuation. Um we ended up not wanting to do that because whenever you take like a high high valuation, what ends up being is your next round. Yeah. you know, like even if you take a we took it and in the end like the main investor was 140 and then immediately after there was a bunch of investors who came in at 210 million.

Uh but we could have pushed it way higher and it's it's the game of hey like I'm going to be in the [ __ ] if I like right now we can still do like a 500 million valuation round and it will be fine but really it should be a billion. Um and um you know sometimes it's good not to abuse your position of power uh too much. um investors get really hot and you know they honestly sometimes they don't give give a [ __ ] about the valuation that much because if you're going to have a huge huge outcome which you know every AI company is aiming to have a trillion dollar company right now doesn't really matter if it's 200 or 300 they want to do the deal and very few of them nowadays are principled enough to kind of you know put on this they are hot they just want it they just want it you know and we you you're talking ARR to valuation, right? >> Yeah. >> Hell hell if I know uh like uh you know who's heard of Instinct.

Okay. Very few people uh they went from uh they are on zero revenue. They are doing a personal assistant which is we're kind of similar but we're for work. They are for personal assistants. They went super hot on their um kind of launch, but wasn't really the launch. Um and they take all of your privacy, but they give you the service for free. In the last 6 months, they went from 200 million valuation on zero revenue to now 14 billion and still zero revenue.

So, what is the multiple there? I mean, infinity. uh normally I see anywhere between 10 and 150. So it really depends what the story of your company, what is your pedigree and how what is the the you know the the real metrics. Um for example for our category uh W2 retention is is u probably like the the north star. So if somebody signs up, right, and they get to activation, what is it going to how many what percentage of them are going to be still around on week two?

Um, so those numbers if they are above like 65, you're very good. Our retention is like 78 right now or even 82. So the investors really love that. But the absolute amount of revenue doesn't matter that much. Okay. So we went through like insane valuations but first um in before you start rising you need to have this revenue you need to have this traction. So if you can move more to the go to market part uh right now if you could walk us through the early traction of boardware how did you build the initial traction?

What was the what was the funnel from launch video or like you know peers and friends launches uh true activation through first paid clients uh all the way like what whatever happens before you started asking for money uh how you got your first clients before you became hot. So I'm going to mix the stories a little bit because uh I I don't know if if people knew but we slightly pivoted the uh the um kind of customer that we are attacking with kind of similar back end but uh right now SA.AI is our main product mix the lessons from both.

So I think when you're talking about go to market it actually really really changed in the last um year and a half. um ads and performance marketing used something that you do if you have zero PMF and it was like almost lame uh in in in Silicon Valley at least and now it turns out that like my story about this is LLMs have a great EMF and nobody knows what to use them for. So, as long as you are able to package some of the models in a way that is useful, you're going to get PMF.

But then your biggest problem will be how the hell do I tell the right people about what it can do for them? Like I swear to God, like if you take CHT and do a good job um you know, putting it into a 500 person company and you're going to run good orals, lessons, and you're going to like nail it down, the impact on their business is going to be huge. and it's just judge PT, you know. So, um, everyone is competitive to everyone and the go to market is becoming like a a big puzzle that you need to solve and there's two different things that you should be aiming for.

Um, first of all, constantly test your distribution channels and your ICP should also be defined about who can you reach um and what messaging resonates with that person and what ends up being the customer acquisition uh cost for that particular persona because you might be actually really really popular with founders but it turns out that you're competing for so much attention with these people that it turns out hey I can't reach this customer.

So maybe you should be targeting you know HR uh individuals giving them somewhat similar workflows and only expanding in their companies through that persona because you can reach them. Um so it's a mixture of things that should just continuously work and I think as a CEO I didn't put enough effort into that. I last couple of months I really spent focusing on the product and I didn't build the right systems uh for performance marketing which is essentially you know a bunch of this stuff combined with SEO with AEO you kind of have to be doing the the grunt work I get bored easily so I didn't hire the right person to do it early enough I picked it up like months ago and it just takes time for the engine to start working meta odds is a great way to do this.

Uh you should firstly try static copy. So just try images and I'm going to get a little bit um tactical here um rather than strate strategical um Andromeda project is a new uh meta algorithm which basically turns out ignores most of the uh audience setting that you have and uses a vision model on your static ad to determine who is it going to be best for. So some tips and tricks on that is try to create an ad set which is uh kind of mimicking parts of like your what works for your um for your competitors.

Uh you can do that in ads library. You can literally type in your competitor's names and it will come up for you with a sorted list of where they've got the most views. The most views means most spent means lowest C which means that you just have to take a bunch of these create a bunch of ad sets. Okay. uh which will um have similar six ads and you'll have like let's say six batches and what you do next is you add a little bit of like text on top of the ad saying looking for small startups or looking for HR professionals and this algorithm by Facebook maybe you know will change in a week or two weeks but right now it's looking at that and trying to put you somewhere in the audience bas on this stupid LLM just reading what you wrote there And then you can figure out what is your cog for each one of these uh personas that you think are your ICP.

Who can you reach the fastest? And only once you have that and you have a little bit more of like an intuition around static ads, you should invest in video ads. Video ads are actually you know you it's hard to create engaging uh videos and as founders you should be part of uh the uh the creators for quite a long time. It's like your new job as a CEO is like come up with a very enticing story about these things and then this is the lesson from the previous company.

You can start employing uh kind of these you know you're trying to grow like this and then you can do you know viral marketing and you can do press and you can do launch announcements and all of these things are going to be kind of like spikes in your usage but they will not deliver the kind of steady growth. So try to start with the steady growth then add on launches then add on all of that stuff and obviously in the end of the day your product is going to do the most of the work if you've done a good job with product which is word of mouth and the uh golden metric is the virality coefficient which means how many people will each person bring to me um and you trying to create viral features inside of your product like we put maiden sauna underneath people.

The end of the day that you know the real exponential growth where you're doing steady doing spikes and for each business is going to be different and in the end you're looking for the exponential. >> All right. So uh you you talked about virals also and who of you here uh have seen Philillip half naked in a sona in the launch video. All right quite a bit and who of you has seen uh his co-founder in a right quite a few.

Okay. So you you have this story of you know launching this you said that this is only a spike uh and the most of the work happens on the constant basis trying to make fix the basics do the ads and so on. Um so what is the what is the main reason of doing those uh those videos uh if it's not growth? >> Very interesting question. I would say that that video with me in the sauna did not work. Uh it was too long. It was 3 minutes.

You should be aiming at around minute 20 if you're doing a launch video. Also, the channel of um the channel of X launch videos is so so overpop populated right now and the quality bar is so incredibly high. Honestly, I would say don't do it. Uh it's you know we ended up like if you have a bigger team as a CEO you often end up wanting to kind of push for something tangible uh so the whole team can kind of align behind one goal. some of these launches to be honest running a team of like 25 people it's it's more for me to just say like guys we need to push and it's really good to pick up you know these real things but still they end up being deadlines so if you need a deadline a long video can be a great deadline uh but I think X videos if you don't have something that you can continuously do and it's like a series and you can keep doing them with low cost.

I would say it's too much work. >> Okay. Uh so, uh I think we're going to we're going to bridge >> Can I actually say something more? What I would invest instead is if you have customers who really love you, um get them to do testimonials, have them put it on LinkedIn or um or act and boost the [ __ ] out of it. So, um it's a really easy way of getting these people some ext. faces are going to pop up on a million people's screen.

Uh and at the same time for you it sometimes can be better than a launch video because they are seeing a real problem with most of AI it's the problem is how the hell do I use it and if there is a real person who is telling real stuff then actually that might be better than launch video. So um do more customer testimonials. >> All right. So it's more UGC than than PR stunt videos. Uh but then uh what is your take on building public?

Because I haven't seen a lot of uh Sona uh I haven't seen a lot of demos whereas at the board where previously uh you've been all over recording videos of use cases. I remember we met on a call like two years ago uh when basically Philip jumped in to solve my client's problem and you started just doing this live and you did a lot of recordings of it. So did anything change when it comes to building public in in the recent year.

Um I think as a as a founder you often end up oscillating between some two versions of you. So often it's how much deep tech I'm going to do compared to traction. And you know sometimes you do some stupid idea gets a little traction and you're like well I don't want to do that. um and you go back to deep tech and you kind of oscillate and I think with with word uh we did a bunch of public stuff but we didn't focus enough on like the core core product and it ended up that vibe coding was a um headwind for us.

Uh so we even though we had a good pathway to 10 million ARR it was just not enough and uh with with SA I think that's my mistake of not focusing on growth early enough. I went into CPO mode. I was on product every day talking to users and kind of focusing just on that. To be honest, I should have split my time a little bit more. And we're now going to do a bunch of marketing things. Uh especially in Poland. There's a spot coming that's just focused on Poland.

So hopefully you guys going to enjoy it. >> Uh cool. So uh last question before we can see you doing your GTM work inside Sona uh to get practical. Uh last question because in Poland there is a lot of technical talent. Some of you standing in here even asking me a question like should I get a business co-founder uh should I hire my GTM engineer like what is the moment when founder itself isn't enough at what scale would you recommend somebody thinking about outsourcing your founder mindset into into the growth if if ever and you said that you made some mistakes in hiring GTM uh so I would combine this question on when to start and what are the reasons behind starting this hiring or co-ounding and how do you pick the right person? >> So, uh first of all, I wouldn't say I'm amazing at it yet.

We're very actively hiring and we can sponsor visas. So, if anyone wants to uh we're looking for a head of growth, we're looking for engineering talent and we might be having a polar office. So, it will really depend on like an incredible person that I can trust with the right pedigree to help us with that. Uh so with growth it's really hard because these guys are much better at selling themselves than engineers and engineers you know they will kind of play it down you know they like yeah that's a project and you can see all of this stuff and really um I would say what changed on growth it's even harder with sales people right because they will sell the [ __ ] out of themselves and with growth actually it became a little bit more of a science so a lot of the things that I hardly ask about is um like sometimes it's just how AI native they are and and you would not believe how like it's easy to catch people um like I honestly ask just a series of questions of like hey what's your favorite model and then if they can't produce an answer I'm like okay that's end of their interview and it happens more than you think who's the CEO of open AI and then there's a who's the CEO of entropic and people fall off you know like I'm not joking and then if they are in like AI video making ask them what's your favorite um video model and you know they sometimes like say Higsfield and it seems like you know they might have produced something but they were never curious about the inner workings of like the AI there and then I ask them what is um what is that one loop that you built uh in your personal or work life where the AI improves itself.

So I think every probably every system is a series of iterations and fixes and the moment when you are able to close the feedback loop for that system to improve itself is is extremely important and if they have no clue about what a skill is like how to you know make sure that your agent doesn't flop at the next API call it's going to make um that's really really tough. Uh so how to and also part of your question was when to do this and I would say it's surprisingly late nowadays and um I think we have this scare in Europe of tech like it's it's almost like you know there will be one co-founder who rocks up like I've uh I have this thing like founders in blazers really I hate it I don't know it's like it it means like that person like whenever They rock up to Weiwork, they think about themselves as like I'm the dude, like I'm the CEO, you know?

I I don't I can just I don't have to do it. And then they like they're always looking for some guy and they're always looking for some dude who is ideally we have this word right. And I find this ridiculous. It's like uh for English listeners, it's like a basement guy who uh is supposed to just code there and sit silent and not not you know be a problem for the CEO and and that's bollocks. I I mean if if as a CEO or founder you're not coding at all like it's it's ridiculous nowadays it's sometimes honestly for me it's easier to put a PR which I know that will communicate what I'm trying to do better than me trying to talk to a person.

It's like, you know, I leave a lot of comments. So, like if I [ __ ] something up, they will be able to fix it, but they at least know precisely what I want to change. And uh like often, you know, like there's no comments right now because GBT6 Astra is freaking amazing. Um and yeah, like it's much later than you would think. I would say the first MVP of the product can be built by you and your yourself alone, even if you're not technical. >> Cool.

So now uh to the to the meat I would say uh I think now is now is the moment where you can uh show us in practice how you GTM max how do you do GTM maxing how you ka how yaktmaxing by yCO now is the the live the live part of it uh tell me whenever you're ready >> and I'll put you here on screen uh as you can see I haven't invested in uh in IT support yet as a founder. All right. Can you see my screen? >> Not yet. >> Uh, remove other media.

Okay. All right. Now we can see your screen. You'll see it in a second. Uh, I'm going to come a bit closer and hold your mic. Okay. >> Oh, thanks for organizing, Carter. All right, let's get let's get started. So um let me just start with a little bit of a uh with a little bit of a backstory of what SA is. Uh in the simplest terms SA is your AI teammate that uh AIT man for for work. It has its own computer that works 247.

So you can reach it from uh from iMessage. You can reach it from Slack and uh it can spin up up to 500 500 sub agents uh which means your computer is not slowing down. It has 3,600 different tools uh which we are quite opinionated about. Um so you can kind of see uh 4600 actually. So you can look through all of them. We've got people search, we've got enrichment. Um and uh you can also use this and sign up on um iMessage.

So uh some of the times you know you're just asking post hog to give you uh give you a set of um data but other times you're actually creating content and that means that you know you can ask sona um how can it be used by recruiters and we have a very good hickfield connection which uh then together with hyperflow actually uh hyperflow hyperframes sorry hyperframes can create um explainer videos really quickly. So, um actually I did try to get Sona to um to give me um to give me GT video session links.

So, here there are a couple of um very long sessions where I was trying to work on a uh on a video and essentially Sona has a built-in give me a second because I don't think I have internet here. Um so what Sona can do here is uh it was able to take frames from within your product and this is kind of a a you know I'm sometimes I'm going to be demoing Sona but sometimes I'm just demoing AI essentially which we've uh baked into the product with skills.

So, Hyperframes is an open source project from Hen uh which is the avatar stuff and that essentially means that it can animate HTML by just putting the HTML screens together and including motion in it. So, uh if you give access to the code base of your product, it is actually able to take the screenshots of your product and put them into uh videos. So here is a great example of a video that was produced based on uh uh based on uh our um our codebase actually.

So uh you can use that then to get um a lot of you know just everyday putting content and what I would uh say as well is if you go to uh sona/brand this is a kind of a place where your agent can grab all of these assets from. So I would really say that you have to be building these companies in a way that is understandable by day. So let's start all the way at the beginning. Um this is Sona. You are able to use it from many different place and it's an AI assistant for work which is also a chatbot but also can create software.

So it used to be that you are using a chatbot every time from zero. Everything becomes a chat and I actually think that's not going to be the end way of interacting with AI because a lot of jobs are actually quite reproducible. So sometimes you need to send a bunch of emails and your agent will end up writing the code every time from zero. So in a way you kind of want this sea of context to live um in between like the agent and some of the applications that we're going to see in a second.

And um one second we have a little bit of a delay. And when you kind of think about these this stuff, you can kind of think about every little interaction that you have should be writing to your memory files. And SA is wired up in a way that it captures all of your relationships, all of your events, every little preference and creates this sea of context. And in the middle of it, there is the main harness of the agent.

But this agent can pull from multiple of these connections and it through to different uh you know different uh applications that you wrote. So as for me as a CEO you know I have a bunch of applications that are quite simple to be honest. It's for example a CEO daily prep which shows you know I have today with Dr. Vala and essentially I can check my item I can check my mentions I can check different different stuff uh like the deep dive with him means that the agent never has to write this but the agent wrote its own API calls.

So this is how the world is going to change slightly. It used to be that everything that you write is a skill and now the agent can actually create these applications for itself. Um, it even though you can still chat to it on the left, it sometimes ends up using these buttons and you know like we thought about mimicking it clicking on the things but it's stupid because it end up that every one of these applications is actually an API call and it doesn't have to rewrite all of this complex code every time which means that it can be as much as 10 times cheaper or sometimes 100 times cheaper depending on the um complexity of the job and much quicker.

So uh as we kind of move forward with uh with the num with the people that we want to uh create for example we are going to start a session and say hey based on my recent um based on my recent meetings come up with five different ideas for a LinkedIn post and the sona as an assistant should know you well enough to make this uh to kind of you know make this quite interesting and in the meantime when this is progressing I actually did some research on all of you.

So uh we should have had around 140 people I think and uh we did research based on our tooling that we've selected and you don't have to wire anything of any one of these tools yourself and you know for example someone in the room has already sold a company for 120 million can raise their hand I mean I know who it is I just don't know if I should open it uh it's that there is hey why not indeed build the world's first sign language messenger and still in high school.

Do we have that person on? >> Okay. So, what I did with Sona, I threw in a long CSV and I essentially said, "Hey, uh using sub agents, which then it spinned up 370 sub agents to do this and I did this when I was in a taxi at uh well I was here actually so 619 and I uh said, "Hey, research all of these people in the quickest way possible." And with most of the cloth work or other agents, this would just not be possible to spin up 370 sub agents to do these uh jobs.

So we've got somebody there's a a Cambridge PhD in the room who was supervised by Zubin Gamari and Bernard. I've got I've got that guy there. Uh >> I also see that I came with somebody who has the biggest tech job board and he's in the audience. Yes. >> Okay, there we go. >> Agreed. >> And he just revived our Laapolska. Well, hello. And uh even though this is quite simple, you can see how you can actually grab any list of customers and make it so that you are able to reach them quite easily with before you would have to wire up Hunter uh then sign up for one of the reach out like loops or something then manage your so you end up paying for like four different suses even though you know your agent can build software for you which is very very personal.

So, um I for example write uh run all of my hiring on these outreach campaigns and my sona built for me uh the right let me just show you outreach. Here we go. uh build for me a uh basically a whole page which then creates uh searches for me the right candidates and then I get to kind of swipe on them Tinder style and uh from the swiping it can uh research their email address and then take each one of these people and we for example have Ian Sansa and I can get all the links that I would like to and I already know that my AI I actually guessed uh his email and I can start pinging him emails at scale.

So similar techniques can be actually used to uh to research for example cold sales. So this is what I what I built uh essentially when I was on the way here and I researched 154 of you and you know we have artwork. I shouldn't probably showing show people your emails. Uh but this is all the people. I'm just going to stop at one at least. Sorry, are two of you here? >> Uh, all the people in the audience here and I've actually got all of their uh emails and some of them I got phone numbers and I actually wanted to whip up a demo where it calls one of you.

Uh, so let's see if uh uh can you call one person uh right now? Um and throughout all of this uh it actually LinkedIn post ideas uh through just knowing me uh the uh I actually drafted five LinkedIn drafts. So the price of intelligence from the abstract argument with Robert I have been saying the same thing for two years. The models gets 10x cheaper every 12 months. Every told me two or three. We just checked the model running on our $29 prime are better than the frontier models we built for the 299 product.

And this basically runs on top of mic granola transcripts. So you can kind of see how the agent together with software gives you both the the flexibility and it gives you an ability to um build more stuff based on the software you already use. So another one that I will show you is mule. It's my version of granola and I started using whisper flow recently and that basically meant that uh I have a bunch of meetings in whisper some in uh some in granola and some of them I couldn't send a bot so I just created a service I asked sona for this now a bot can you know just join these calls even though granola and nor whisper has that functionality and ide additionally I'm a little bit scared of that one actually uh But additionally uh I build a uh we say a lot of like inappropriate stuff in our meeting and I really didn't want to not to be part of our due diligence at some stage.

So I created a reduction b which never saves any of this. So here we can see that u sensitive legal damaging speculation about headsp space founder payouts and then one passage about crude bodily reference users emphasis which are actually wiped from this uh from this transcript. Uh so here we can see that what is this? Well we don't know. Um and in this way you can kind of build these applications that can do stuff on um can do stuff uh at a scale which was not possible with AI because it was too expensive before.

And uh I will just maybe show one more that I use for growth. Uh we have a bunch of customers that we know are kind of not well like engaged. I have privacy on. So uh this bot particularly is checking for people which we don't think are activated and gives me an ability to email all of them and uh we can see that you know we've I've been emailing with this person that they are possibly at risk and this basically gives the both me and the agent the right buttons to do this at scale.

So suddenly I can, you know, have Sona spin up 10 different email addresses because she has access to my Google dashboard and then use these email addresses without all the [ __ ] about, you know, warming up or warm them up um agentically, which is better than some stupid service that charges you, you know, basically 10x of what their costs are. And uh then just use that to have an ability to send an email. Um and this this this is pretty much it.

I think in the end of the day you should be building good knowledge with the right skills. You can kind of see how many skills Sona creates for me automatically. You should u basically forget about privacy and connect pretty much everything that you have. Uh like it's so overrated like what like when everyone says like oh I won't give it access to my email. I'm like you're not a serial killer neither you're that important. like chill.

Then uh you basically want to um build these skills and have an ability to share them with your team as you on board them. You're probably the only person that is AI native in your company. Um and then uh what I do is I give access to multiple people and to my sauna. So uh we can kind of see that you know even my best friend has access to my sauna which means sometimes I don't need to be in a meeting room to give opinion about something and I'm going to give you real examples and I might uh regret this but I'm going to filter for shared sona so we can see oh [ __ ] mikron asked me about philip sona about friendship.

Uh I'm probably not going to do that one. Uh but here we've got some stuff from the last uh couple days where my chief of staff is for example asking for explainer video uh sources and this message is basically filtered by my sona to give him just the right uh data and my sona will also protect my privacy. So essentially what needs to happen in all of your organization to really GTM max the this and you know you can use Sona I'm going to give away free access in a second but you can also use any other um agent uh we just we think our agent is the best uh create the right skills make sure that the skills um improve themselves so at the top of the skill it should say whenever you're done with using the skill make sure to uh learn all the lessons from the MCP calls from the API Guys, it's that easy.

You just add a sentence. Um, if you are doing something at scale, you should be building applications to do this. You shouldn't just be uh, you know, believing that the agent will do this efficiently every time. It won't and it will cost you a lot of money. And when you're sending suddenly a thousand emails a day, it can really rack up costs. Um, you should ask Sona about what tools we have wired up and it can even steal that IP. you know, it's like in the end it's whisa, it's Hunter, it's bunch of these things we're opinionated about.

Um, then once once you have it set up in this way, connect your iMessage. So, um, have whenever you're going around and you have an idea of something you could do, don't think don't think about it. Don't write like write it down to like, okay, I'm going to take care of this later. give it to AI and just say ask me questions and try to do as much work as possible u so I can review it it's literally it's the biggest thing that will change your life is oh [ __ ] I can ask AI to do this it's that simple and it's still I get lost in this you know I'll be like pro like prolonging a task I'll be like ah you know putting it away and away and then I just ask Sona and it turns out she's got all the context for if just to do it for me.

Um and then from there try to onboard your team. Share your um share your skills and share your sona with them. They should be asking quite often like what did what did Arthur do and what are his skills on dealing with Hicksfield and then you know his sona is going to go here is going to look into Hicksfield and is going to look through a very you know detailed uh skill that even has the right code snippets and you're going to be able to pick um right of his uh trials and what didn't work for him and then try to build your whole team uh try to introduce your agent through Slack, you can then start uh mentioning these uh like the agent, what do I think about it?

I literally write that sometimes. It's a little bit annoying, but I'm the CEO, so I don't do a [ __ ] And uh I can just say at Sona, what would I say about this? And it produces basically my opinion about this. And we start realizing how Sona is approaching it very differently to for example Victor which is also a great company but we are saying we're going to give you superpowers firstly as an individual and then we're going to scale each one of these people to be the best version of themselves which means I have time to coach my employees.

I analyze my transcript and try to figure out how can I be better. It means that every hiring uh call gets the right notes that I get to learn over time. And then when you get that superpower into all of your employees hands, it turns out that you can work as a Fortune 500 company and be 15 people. I still think people are extremely needed to take responsibility and give taste to whatever you're doing in life. Um so that's why we are not approaching it as like one coworker.

It's no it's everyone gets their own AI assistant that is helping them but they are them in a way. Um and I would say the next part is if I could we literally just had an incident on this because I use this as a way to push my team. Uh but uh let me see if that incident is improved. Uh I'm reverting the login separate from this go to prevent. Okay, I think it's done. So, if all of you can go to sona.ai on your phones uh and this should look like this.

I've turned on fully free access uh for you guys. So, you can have Sona for until I really uh regret this decision for free. Uh and uh just click try on iMessage. This is going to ping you to um an iMessage I think and just send that message. It runs through iMessage or RCS if you're on Android. Um and this will basically mean uh Sona will respond to you. Then you uh login and it's going to do go through a couple of proactive tasks and you can keep using it and ideally you can give me feedback on Loom uh on my email. >> Okay.

And if it doesn't work here with internet connection, uh we'll just send over links afterwards uh on the blast for everybody who's in here. So don't you worry. And we're going to send a feedback form together with it. >> Yes. Cool. >> Everyone everyone from here, everyone from here should receive, I think, uh $25 of credits and this is enough to build your first application. If you would like to do that, just make sure like you change the model to smart. our uh cheap press model is not yet good enough.

Um but you can also run through a bunch of tasks but what I would suggest is like really connect the things that annoy you. So um typical candidates is HubSpot, Salesforce, um Ashby and any uh any software that you dread of going to Stripe Post Hog and it can really do quite complex things. just try to give it a job of like, hey, give me what's going tell me what's going on in my post talk. Um, dot dot dot. Oh, and yeah, I'm going to give you just phillipword.ai or philip sona.ai and I personally read all the feedback.

So, if uh if you run into any problems, just get an incident. So, hopefully this works. Uh, but uh give me give me feedback and I would love to to learn. >> Cool. Uh, thanks Phillip. Thank you. Uh, I think it's a moment for an APPLAUSE and and I think given given the fact that most of you are sitting on the floor, I would keep the Q&A to like five questions maybe. And uh, >> I have time I think >> and but then we're going to mingle around.

That's why we're in the cafe. That's why we're in here. So it's like casual. So uh, four, five questions and and we're gonna go drink coffee. Are you planning launch a mobile app for Sona? >> Uh right now we focused on on iMessage to be like the closest to customer. I think something that is great about the mobile app is like you can send notifications but I almost want people to like reduce the friction to zero. We're working on WhatsApp right now.

It's going to be live in like two days. Um, and really make sure that like when I review work, I go on my computer because it's just I need to read the little [ __ ] or I need to see the results of it. It's only when I delegate work to the agent. I don't really need this. Like I would say often when you have that idea, just leave a voice note like we will deal with it. Don't even like use the Apple stuff for voice notes because that really sucks the transcription. just leave a voice note of like a bigger idea and then when you're getting to sit down and like that's how my day works.

It's like I'll give it bunch of jobs throughout the day and then I'll sit down to it like 8:00 p.m. and kick off some like big big jobs for the night. So I thought AI is going to mean that I'm going to work less, but it ends up I just work more and I'm more anxious about not working. So welcome to the new year. >> Just say it out loud. Victor and all these are coming up um codeex they have some of these things um particularly Google's the new anti gravity 2.0 you know, kind of looking at the interface and kind of what it has.

My question is you probably make some markup on the credits just being fully transparent. If I know how to use AI, what would be compelling for me not to just build it myself and ask for >> minus 100 margin profits 100%. So, we don't make any markup. We lose money on every dollar that you spend. We lose a dollar. Uh so, essentially, it's a battle. uh it's a it's a land grab. You need to be raising quite a lot of money to be giving people things away for basically losing money.

So, first of all, that's the first thing. Second of all, um I think you get to benefit a little bit from just us as a big team trying out a bunch of these things. Like, who the hell knows if to try people search there's like 30 different tools. We just chose the right one. We wired it through and you don't even have to see the API of it. We handle the billing ourselves. So in the end, when you look at all of these agents, the pie of knowledge work is absolutely like behemoth.

It's huge. It's bigger than coding. So there's going to be a bunch of multi-t trillion dollar companies doing this. And when you're trying to compete in any one of these things, turns out that more general agents keeps on scooping it. And the main thing that should keep you um sane and I'm not maybe doing greatest job because every time I go to um Twitter I get the slight anxiety but essentially what prevents uh me from going crazy is knowing that taste in product and curation is actually extremely important because in the end of the day nobody knows how to use AI yet.

And if you tell people how to do it and wire up the right tools and do it properly, it's already that that's kind of your mode. And it's it's gotten ridiculous. It's Yeah. >> I don't know if it's a bubble. >> Well, no. I mean, like if you work with AI, you tend to hang out with people that work with AI and you all think you're behind, but then you're like, "Oh my god, we're so behind." But then you're like the other people that don't really think. >> Yeah, I think I think it's gotten crazy.

Like it's very hard to to sleep at night and like know how quickly the competition is moving. Um the one thing that's kind of good is that there is an escape hatch. Like you can sell for like I don't know a couple billion now which used to be not possible. Uh and so it's better to aim for the stars you know just [ __ ] it. >> Question over here. What was one of your initial assumptions when going to market some that turned out to be like completely wrong? >> So we had fullon open claw before open claw and it worked really well and I thought uh [ __ ] this will never work because you want to be doing background work and nobody is going to keep their computer open. um open claw has kind like so I should have been launching earlier essentially and I think uh it's a very simple lesson everyone always said says this but I think the part of raising a valuation that is you know 210 million or whatever it's it ends up being that you always question and like will this get me to a billion dollars billion dollar is like a lot of zeros there's a big difference between a million and a billion I think as humans we don't really comprehend like you know how many orders of magnitude of more that is and I should have been launching sooner launching more.

Um yeah so that's the main lesson. I think we got a lot of things right the first time around which means it's even more um frustrating when we now seeing people not use MCP and we're seeing that you know like people are saying okay you should just use you know the code and that's it and yeah I think that's the biggest lesson. Here's the question. >> I'm not that big of a bozo. Maybe one day I'll be able to answer. I think that I think on every every stage the techniques are like the psychology of humans stay stays the same.

But I'm I'm sure that these techniques are getting crazier. Like you know these guys from Instinct like my god hats off huh like they are they just raised at a billion everyone thought it's ridiculous and now they were ra and then like nobody is raising nobody knows this yet but like in valley everyone knows that they've got offers at 14 billion valuation 14 billion six dudes and zero revenue like that's pretty impressive so I guess I can maybe answer this question in the future question over here. >> Yeah.

We see actually wrote articles like seven two years ago about that describes everything as I understand after launch with some times by the web traffic and the amount of hype product. Uh so wondering is there any learnings or things done differently to continue with this hype? I think in the end of the day it's relatively simple and you have to have high belie like we essentially went for a go to market pivot right and I think it's very simple in a way that if you have extreme conviction in your product you're going to be talking about it and if it's working you're going to be talking about it we ended up feeling like we cannot get to 100 million ARR and it was extremely difficult decision to say hey we're going to start talking a little bit less about this and start building again.

So, I I would say uh I don't know. I don't think I learned much from this. Um keep on putting yourself out there, but make sure you've got the right engines. And if you don't believe in your product, like you're not going to be wanting to talk about it. That's kind of the rule. So, I maybe we could have been doing this, but uh it ended up that we basically had too much headwinds, I would say. >> Time for drinks and questions.

It's it's really it really depends because then I can't really like I can help with some uh extra kind of tactical things. So if you really like even if you just want to ask a question about your particular situation that ends up being like more useful overall rather than like going through with this but people have lives so if anyone needs to leave you know but okay cool I'm happy to take more questions. Uh I think that works a little bit better than uh just going around in the room because then there's a circle of 50 people. >> Go ahead. >> Yeah.

So I love I love the idea with all the recent hype with J plus models. Do you think it's going to disrupt and make new opportunities for like assistance and it's going to stop or just make it cheaper in some cases? I think Antropic and Open AAI are going to start milking people after uh the IPOs and there will be a new window for people to enter the race as uh hopefully uh Europe nor America is going to ban Chinese models.

Uh our main um like we are already using um a lot of like fine-tuned Chinese models. Jeff, >> what is the Jeff model? >> Oh, okay. >> Okay. So, that's not based on transformers. Yeah, Facebook is like my orders