YC Partners Answer Your Questions | Office Hours

0:00Two of the really hard questions you have to answer as a founder when you're getting started are who am I selling to and how do I get their attention? And those are like the two big magic tricks that every founder has to pull off. When I came to YC, I think I had to untrain myself for a couple of years of like all the learnings I had and how they were not applicable to startups. If you have a lot of time to think about this question, it's probably too early.

0:27Welcome to another episode of Office Hours. Today, we're gonna respond to questions from the YC community, starting with several about AI go-to-market advice. Here's the first question we're going to take a look at. If you're building an AI company in a legacy industry with a long-term vision of fully automating everything using agents or LLMs, but you can't deliver that on day one, what's the best way for you to go to market as a startup? I think that there are three types of companies if you're going to bring AI to your legacy industries. Let's take, for example, the accounting industry. So there are three ways you can do it.

1:00You can either build an AI software company that you sell to accountants. The second one is you can start your own accounting firm and it's sort of full stack or does everything. Or third, you can try to buy an existing accounting firm. There are pros and cons to all three of them. The most common one is the first one. This is how most YC companies do it. They will try to understand the world of accounting. They would try to figure out what are the areas within accounting that are most valuable to go after when you're building AI software. that is also reasonable to build in the first, I don't know, couple months or first six months of the time of your company.

1:36And then they try to sell that service to the accountants and they are not supporting all the other features and all the other things that accounting firms are doing, but they're doing that thing really well. And that tends to work pretty well as long as the thing you're doing is valuable enough for them to buy. The second option is to start a new accounting firm. There are a bunch of ways to do that too. The biggest challenge is the lot you have to do as a company to do that. You have to probably do taxes and closing the books and doing a bunch of less common things, but you still have to do if you want to take on that role.

2:08The challenge here is you probably have to have an accountant on staff or maybe several to be able to do all this stuff. And you'll have a lot of manual work. So if you do it this way, the thing that you would track is the percent of the work that's automated. You want that percent to go up over time. The third way is to buy an accounting firm and then try to ingest AI there. The good news is you get customers to already exist. The difficult thing is you're changing the culture of an existing company. If the bigger that company is, the harder that will be. I'm not sure I've seen a lot of ice companies try the third thing.

2:42The most common one has been the first one and the second one is like a closed second. With the second one, you mentioned tracking how much of the work is being automated. Do you have any thoughts on what they should aim for or how they can force themselves to do it? Because what I've seen is when companies have tried this, they get so bogged down in just the execution of all the work that comes in and being a successful accounting firm in this example, that they never get around to shipping the automation or becomes the second thing that they have to think about every day. I would say this is where software founders sort of like are the most powerful because they can see at all, look at all the works that you do and they can try to figure out what of this work is going to be easiest to be automated like right away.

3:26versus someone who, say, is an accountant funder who's not a software funder might not see that in the same way. So it helps to be a software funder in that category. The second thing I would say is you just have to find a metric. And the failure mode I've seen here is basically it works and you try to scale up revenue too early. So let's say you automate 20 % of the work and 80 % is still manual, and then you're trying to scale up the company. and you're hiring like 20 accountants and then 30, and then you're actually manning an accounting firm with some software. That's just not recommended.

3:58And it's going to be difficult many, many times. And the more, like at Airbnb, we had this metric, which was percent of technical people that work at the company. And the reason we had the metric is at some point you have too many non-technical people. All they do is request things from the technical people and then you can't get anything done. So you need a certain, I think, I forgot the percentage, 30 % or something like that was counted as technical. It's a framework that I find helpful. because then you could make sure that the number of technical people in the company is always enough. You can continue working on automation while you're also trying to do the other stuff.

4:28How about creating a forcing function? You have one accountant and you cannot hire more. Yeah. Then you get more business and you have to figure out how to scale with AI. I think your point about having a number, I don't know how they did exactly at Airbnb, but that's apparent and visible to people, maybe that helps kind of put a foundation on the culture of the business of like, this is really important to us. All of us are looking at this. We want this number to go up. And if I was to say a CRSA investor, and that would mean one of these companies, I actually care more about the trajectory of the automation rate than the overall revenue.

5:00Because like, yeah, there's many accounting firms out there. If you start an accounting firm, you make a lot of revenue. Like, does that mean you can raise a CRSA by a software investor? It's like, not really. That's not exactly the thing you're proving. You're proving that you can actually write software to automate a bunch of the tasks. So I'd rather have a slower growing firm or company where there is just more automation and there's a clear track record of them doing more automation or using more software every month. I worked with a company in the previous batch, the Spring Batch, which was just now wrapping up, called Vessence that are building software for lawyers.

5:32And neither of the founders has a legal background. And so the way that they got started before YC was they found a large law firm in Stockholm that was so excited about the idea of using AI software that they let the founders work out of their office for several months. And that's how they built their MVP, which seems like a really great path for founders that are excited to work in a legacy industry, but don't have as much personal exposure to it. Yeah. I think also if you're in the first category, when you sell to, you sell to, say, an accounting firm or a law firm, you want to find law firms and accounting firms that are with a founder or decision makers really, really bought into helping you and getting things done.

6:12and sort of like they themselves are empowered and incentivized to increase the use of software in their companies. And it's hard. Like that's not easy to do for your first customer. It's like the stage even before the early adopters in the crossing the chasm model. It's like the ones who will adopt it before it even exists. Yeah. I think a common thing here is just doing really good. So like as you're reaching out to customers, have like a lot of pre-qualification or qualifying questions to sort of qualify someone in or out from being that person. Like, do you know how someone who's going to be highly incentivized, kind of excited about doing new software, an early adopter, can you figure out how the pattern of those people look like early on?

6:51I think that's pretty important because otherwise you can drag into enterprise sales. So drag into sales, they'll take forever. And you actually don't know if these people are going to be adequate or not. All right. Let's take a look at another question. Enterprise AI plays like Viva or Palantir can take time. There's not that many buyers. The sales cycles can be really long. Adoption of new products can take a great deal of time in those markets. Meanwhile, investors are impatient for growth. They want to invest in companies that are growing very quickly. In today's AI gold rush market, should companies start out in the mid-market?

7:25Should they think about that being the initial place to go where maybe the TAM can be a little bit larger, maybe where the sales cycles can be faster? How should they be thinking about time to grow versus long-term defensibility? I'll take a swing at this one. So I think early on, and this is assuming that the company asking this question is like fairly early on, the most important thing is the pace of learning. How quickly you're learning what the customer wants, what the user needs, what their problems are, what the really pointed pain is versus the more dull pain that they're just kind of tolerating.

7:58And so oftentimes when I meet with a company that may want to go super enterprise from the get-go, which these enterprise software companies are not so different from like the moonshot space things we fund sometimes, where, you know, putting a satellite in orbit is not so different from landing like a half million dollar ARR contract with a big company. And so if you kind of think about it that way, the advice to the space companies is always, well, let's find something smaller that we can wrap our hands around and be really successful with. It's the same for a software company because it'll help them learn a lot faster, get better feedback, and also put themselves in a better posture as a startup to make changes, iterate, ship product, talk to users, and so forth.

8:40Whereas if you go straight after the big guys, unless you have some special in from the get-go that'll help you short circuit that sales cycle, you're just not going to learn as quickly as the other companies that are interested in that industry. I would put maybe a caveat here. I think there are two categories of companies. There are the companies like even my company, Algolia, that kind of like that kind of co-host over time, starting with the long tail of customers. And then as you, the product matchers, as you can address more use cases, you go up market. And you end up with these multi-million dollar deals.

9:09But on the other hand, sometimes you have a company whose the problems they are trying to solve is only an enterprise problem. And if they are trying to sell to their batchmates or small startups, I mean, they don't have the problem. So they don't have a choice. They have to go to start high. Now there is like difference between mid-market and enterprise. That's very fair. And probably you want to go after the smallest company that has the problem you try to solve. Or sometimes it's reducing the scope dramatically. Maybe you can land with like one or two users that work at the enterprise and get something that's useful.

9:41It's just a much more narrow product maybe is the way to get in and have a shorter sales cycle. I would say in addition to picking the right segment, which sounds like mid-market is probably the better one or more promising one, is to qualify the people you're selling to really well. A lot of part of industries are trying to buy ASR for now. Enterprise, mid-market, small companies, everybody is interested in this. But you still have to qualify the individual. selling to a person or a team. You have to know that they are empowered to make the decision, that they have incentives to make these decisions and the consequences of buying the software.

10:14And you have to be able to meet them and kind of talk to them and make sure you have an actual interaction with them. A lot of founders should think of segment before qualification. And I think sometimes just the right person is more important as long as they are empowered. I've seen buyers at mid-sized companies move really fast if you find the right person. But enterprise, the challenge, like you said, is like feedback cycle is slow. They make slow decisions. And then founders can get bogged down in like a multi-month sales cycle. And you actually don't know in the end if it's working or not.

10:45Thank you so much for that question. For our next question, the founder actually submitted this over video. So let's take a listen to that. Hey, should I hire a growth hacker, a communicator, a sales, an SDR, etc.? How should I try to replicate those employee by AI employee? I know there's a lot of solutions, but I'm still not sure what's the best approach here. All right. What do we think about that question? So I'm pretty bullish on AI sales software, but I will say that AI SDRs tend to work well when they're plugged into a sales process that's already working well. And where I haven't seen them work well is when founders sort of turn to an AISDR as the solution of last resort, where they're just totally unable to sell their product.

11:35And they think maybe AI can solve this problem for me. And I haven't seen that work. I think the hard work of figuring out how to sell the product is still very much on the founder. When I came to YC eight years ago now, I spent the last five years working on growth at a consumer company. and I think I had to untrain myself for a couple of years of like all the learnings I had and how they were not applicable to startups. And I think a lot of the tools people are building, they fit great into like bigger companies since things are working. And most of the growth advice is the same for consumer companies and they don't really apply to startups.

12:12So if you haven't figured out things are working, if you don't have hundreds of customers, like it's unclear if it will work. And if it's working, it might be working in the wrong direction, sort of like it's sort of like You do a bunch of progress that actually doesn't teach you anything. Two of the really hard questions you have to answer as a founder when you're getting started are, who am I selling to and how do I get their attention? And those are like the two big magic tricks that every founder has to pull off in sales. And once you know the answers, it's a lot easier to point an AISDR or an agent to help with that, right?

12:46There's a lot of schlep work once you've figured those things out to find these people and to get their attention. But actually figuring out how to do those things the AI has not actually been that helpful with yet. The fun thing is that you can actually adapt that advice for the AI, SDR companies. In the sense that if they go after the people who are not able to sell their own product, there is a little chance that they can do better. And these customers are going to churn. A lot of revenue fast, but mostly churn. And I think that has been the learnings, actually. The ones that I know, that we fund to some of these people, is that they will probably tell you that when we sold to startups, trips, they were churned.

13:22It's a real learning that they have. For them, the goal is to find the people who have a good product, who can sell it - Who figured out the magic trick. Exactly. And then they can scale that with AI. It's funny. It's not so different from the advice I think we've been giving founders forever about hiring the first salesperson, which is it's almost always too early to hire them unless the founder has already figured all this stuff out, how to get attention, what are the objections we're going to get. And so the same advice of like, you want to hire a salesperson when it's an execution focused experience and they're just kind of running the playbook that you've already put together.

13:54It's probably that times 10 with these new AISDRs and some of these new roles. And I think that founders should be curious enough to learn all of these jobs before they scale up or really try to hire these teams because VP marketing is like notoriously high churn job. It's not because the VP marketing folks aren't good. It's because the founders have the wrong expectations of what those people do. And they haven't been curious enough to learn about that job. And it's like, well, harder to measure than engineering or something else. So I'm a huge fan of like founders being curious and really trying to learn the job first before they hire a bunch of people.

14:29That's a great question. Thanks for asking that. Okay. Here's our next question. Should we startup founders spend money aggressively now to gain a temporary edge or wait for the next model leap and try to make what we're doing free and more accessible? That's a good question. I think there's maybe two aspects to the questions I'd like to address. The first one is these founders, they should ask themselves, am I doing something, building something that's going to be irrelevant once GPT-5 is released? Or am I doing something that's going to become much better once I can leverage the new AI models?

15:06And so of course if you are just building something that's solving for the pains that GP5 is not yet solving, it's probably a bad idea. Now let's say you are in that second category and the improvement of models are just going to make your own product better, should you wait or should you invest and work on it? I would argue that if you do invest on it, you are going to learn a lot from the process and once the models are going to be ready, you plug them and your product is going much better day one. So indeed you have maybe wasted, I don't think that's the right word, but maybe overspent a little, but the learnings are worth it.

15:42That's why you do it. We had these experiences this year with Cloud Sonnet, right? When that model came out, a lot of companies that were building say internal tools were like suddenly working and they weren't really working before. So we've already seen this before. CodeGen is a great example. Like all these CodeGen tools, like they were barely working and with the new models. Magic. All right. Thanks for that question. All right. For our next questions, we're going to talk a bit about pivoting. We got several questions from founders about that. First up, when should you consider pivoting if you've got some traction?

16:14A lot of times when we think about traction, it's, oh, things aren't working. You're stuck. We need to pivot. But what if you have some traction, but it's not strong enough and you wonder about that? When should you consider pivoting in that situation? Probably the most difficult situation you can be in right if it's working it's working easy not working obviously you have to change something i can see of one of my companies who went through that exact journey uh fire call so fire call is a way for companies to uh simple open source product that can help a lot of companies to extract data to extract information from any website super successful right now a lot of ai agent customers and so on but before that they were working on another product called mandable And I think that when they actually pivoted, they already had hundreds of thousands of dollars of AR.

17:01So significant traction. Not like just 200 bucks. Actually real customers. And big logos too. At that time, they were doing some Q &A on top of documentations. And I think in their case, what they saw is that their growth was relatively slow. They were seeing all of this need. from the market on the side and what happened is that as part of building Mendeable they ended up working on the crawler because they couldn't find any tool out there that could solve their own problem. So they built it for themselves and as they were chatting with other founders they realized that everyone, every AI agency company needed that crawling function and they ended up realizing that their product, the niche thing inside the bigger Mandable thing, was actually way more valuable than Mandable was.

17:58And so they experimented a little. It's not like they moved from idea A to idea B overnight. They experimented with that component and it took off and rapidly they decided, okay, that's the company. Big leap of faith, like hundreds of thousands of revenue, then you start from scratch. but it worked out so well for them. Was there like a moment or an observation that they had that led them to observe that this is the subset of our product that we should bring to the fore? I don't think it's a formula. I don't think there is an algorithm. If A and B then do that. I think it's more like a deep conviction you build for yourself coming from a lot of conversations you are going to have.

18:42I mean, in a way that happened to us at Algolia when we started our very first product was an SDK that people could embed directly in their mobile app on device. And we got some revenue, nothing compared to Mandable, but like a few thousand dollars a month. And we kind of like get that feeling that it was very hard sales, like things were slow, people were not really valuing the product. And at some point we realized, okay, we could build a lifestyle company, but that's not what we want to do here. Let's change something. That was easier for us because lower revenue, But for Mandable, I think it's more like it's a company, it's founders that are very well connected with a lot of peers.

19:21And I think that's what gives them that internal conviction to actually change something. I love something that you said, the question of like, are people really valuing the product? I think that times when I've been in office hours and maybe sometimes we fund people and they already have a little bit of traction and they come in. And as we talk to them and kind of learn about how they talk to their users, we find out that they're not talking to their users or they don't really know much about them. And so there's kind of a gentle urging the founders, maybe not so gentle sometimes, to go and find out how much they value the product.

19:52And oftentimes, they don't value the product very much. And so that's great fuel to encourage people to dig deeper and find a better version. An example that I worked with was Greptile from the Winter 24 batch. When we funded them, they had a few thousand dollars of MRR, and they were feeling awesome about this, right? They went from zero. They had dozens of customers paying them. The numbers were going up all the time. And I just kept urging them to go talk to the users and try to find out which ones really valued the product, like you said. And I remember they asked straight up in an office hour, shouldn't we just grow this number?

20:27Like, let's just make number go up. Why do we need to do all these interviews? And after they talked to enough folks, they realized, oh, like, no one person is saying the same thing about our product. It's just kind of a disorganized relationship between what we're making and how people are interacting with it. And we need to get it more organized and really focus on one or two of these. And they've been able to do that and grow, but it's about, are they valuing the product? One more thing I think about pivoting is that it's a very vulnerable step in a company. It's one of those moments where companies fold and they just give up.

21:01They're like, we've tried two or three things. We're just going to give up. Nothing is working. and when you are asking yourself the question should we pivot you have to also be pretty certain that you have the energy to do the pivot because the pivot is a hard one because you have some existing stuff that you built to pour your months or years into something it's not working and now you have to do something else so you have to first be like okay I have the energy to keep going and trying to do this and it'll be an uncertain time for a while so you have to build a conviction you have to have the energy and I think there's frameworks that we can help with on how to do it.

21:33But at the end of the day, it's sort of like, it's easy to give advice in the pivot, but you're literally starting from scratch. And it's very hard for us. A lot of founders come to us and be like, is this a better idea or this is a better idea? And I'm like, yeah, if you give me seven ideas, I will give you my subjective opinion of which one is better. But I don't know. I'm not the customer of any of them. And you should go and validate the ones that you think are the best ones. The harder one with pivoting on the framework side is to be like, here is a new idea and they come with one idea. And if I say, no, I think here are all the reasons why it's a bad idea, then founders get demoralized and be like, well, they don't want to work on anything because the one idea we had wasn't good.

22:08You don't think it's good. Our user Gustav doesn't think it's good. Yeah. Or just like the feedback we got from people is that this isn't good. So it's much better when you're pivoting to have a range of different ideas exploring. So you can find conviction around something and be fine with throwing out a few of them. So that's one framework It relates to sort of like finding a better idea. It also relates to sort of like finding the motivation. I worked with a company in the fall batch that spent the entire batch sort of wondering, should we pivot or should we not? And they had a few thousand dollars in MRR and they've been working super hard on sales and it just wasn't turning into anything real.

22:42And they ended up pivoting right after the batch ended to a like open source billing framework called Autumn. And they have far less traction now than they did in terms of dollars. But what they do have now is conviction and you can just hear it in their voice. And that's like, it feels like the higher order bit where the actual leading indicator that maybe you should pivot is you just stop believing that what you're working on is gonna work out. All right, thank you so much for that question. YC's next batch is now taking applications. Got a startup in you? Apply at ycombinator.com slash apply.

23:14It's never too early and filling out the app will level up your idea. Okay, back to the video. Okay, for our next question, a founder asked, when should you kill a good startup idea to find a great one? And furthermore, like what is a great startup idea anyways? It's a hard question to answer because you don't actually know that whether it's good or great in that moment, usually. Like you don't know if a great startup idea is great until you've truly gotten people to give you the feedback to say that it's great. So it's just hard to know that. And I think it's more like a hypothetical question, assuming you have all the answers at the time.