Transcription
I investigated the last 500 YCI startups from winter '23 batch to all the way to summer '25 batch, since YC is clearly and heavily investing on AI at the moment. In 2023, 44.5% of YC startups were AI-related, and in 2024, that jumped to 72.5%, and in 2025, that's about 80%. So, yeah, it's just a matter of time before it hits 100%, and we will be living in a dystopian and controlled world, being enslaved by robots. Nah, I'm just kidding. I hope so.
We're going to go through this list to identify what patterns are emerging, what kind of problems the startups are trying to solve, and most importantly, where is all of this AI innovation is actually headed to. I got all these data from the YC startups directory, since it doesn't have the filter by AI category option. The file originally had 1,000 lists. The filtering out was a long and kind of a lengthy process because I needed this list to be as accurate as possible. We will talk about this later.
Okay, now let's talk business. That is business-to-business, because around 85% of the startups are out of this 537 startups. Yeah, there are actually 537 startups that are B2Bs, and just 11.2% are B2Cs. One of the reasons is there is a higher revenue potential per customer in B2B, whereas in B2C, the per user is typically much lower, and you often need millions of users just to build a sustainable business.
These B2B startups are focusing on solving a very specific, high-value, and often inefficient problem in a single industry. For example, Amber, this Amber startup building an AI revenue cycle management, makes the explicit promise to reduce claims denials by 55%. And if you've ever dealt with insurance, you know that's a big deal. And next, we have is Linewise, building an AI to help tackle manufacturing line issues. It solves a very high-value problem where downtime can cost millions for companies.
And on the other hand, in the B2C space, startups are focusing on problems of personal productivity, education, or niche marketplaces like Martin, an AI personal assistant like Jarvis for managing a user's digital life.
"Jarvis, delete my browser history."
And Sonata Sun, a tool Sonato, which helps users create songs with AI.
[Music]
"That was pretty good."
Next is Mosaic, a platform to create and run your own V-rating AI agents. I might need this.
Finally, the third category, which is unknown or hybrid, like Studio, an all-in-one AI creative platform, is tagged as both B2B and B2C. And next to is Waffle, an AI game builder.
Now, let's talk about AS. Of course, that means Agent as a Service. Out of this 537 startups, 151 are building AI agents. And out of this 151, 44 alone are from the recent spring 2025 batch. And if you haven't heard, 2025 is officially the year of AI agents.
We can categorize all these AI agent startups in four categories. Category one, startups building agents to automate the entire sales funnel: lead generation, research, outreach, and CRM management, like Proxy and Orange Slice. This is by far the most crowded category, and most of the AI agent startups are currently around this.
Category two, these agents tackle internal back-office tasks in a company, like managing data, processing paperwork, and all of this. For example, Socratics AI provides AI workers for fraud and risk teams. Novaflow deploys AI employees that automate medical operations for clinics.
And next, we have is Category 3, which are agents that automate the work of software engineers, like Delta Engineer, that can architect software.
And now, the last category. Just like there's a line that goes "sell shovels in the gold rush," these companies are building the tools to help other companies build, manage, and scale their own AI agents. Companies like Agent Hub, an evaluation engine for agents, and Beexel, which calls itself the AWS for AI agents.
In the future, every large company will deploy thousands of these autonomous agents into their workflow, and they would need an entire ecosystem of tools to build, test, and manage and secure these agents. The game is trending from knowing how to use a specific app well to knowing how to manage a team of AI agents well. As you can see, all of these companies that are building AI agents are not just selling software, but also digital employees at scale. This is one of the reasons why many companies are just focusing on B2B in the AI space.
Speaking of digital work, there's one area AI made a really big improvement over one year. Yep, you are absolutely right. That is coding. Out of this 537 startups, 98, that is 18.2%, are focused on AI coding and developer tools, because there's a huge demand, and AI assistant coding is becoming the new normal, and it's creating a snowball effect. Like, YC funds a bunch of these companies making these tools, those tools get used by other YC startups, and those startups then build their own product faster with smaller teams and for way less money. So the loop goes like this: more AI tools, faster building, more startups, and even more AI tools.
So if you are a talented product builder and know a specific industry inside out, you can just jump right in. For example, Lora is making an AI workspace for lawyers. Both AI is handling insurance certification and billing for dental clinics. Their ages knowing their industry better than anyone else.
Now let's talk about some weird and wonders. Here are some of the unique startups that I found which are quite interesting. First is Artificial Societies. The name is exactly what it sounds like. They create simulated societies of AI agents that test how real people might react to your product, your marketing campaign, or your social media content. It's like The Sims for marketers. And to be honest, this is a quite a genius idea.
Next, we have is Maya. Their website headline is "Stay together using AI." You have probably guessed it right. It's an AI-powered couples therapy. Jason Wong wasn't kidding when he said that AI will be in every industry imaginable.
Next, we have is Goda, building an AI-powered physics engine that gives engineers and scientists simulation-quality results for their experiments.
Okay, now let's zoom out. So after all this data and analyzing, here are the three big areas where YC is placing its bets. First up is vertical AI for legacy industries like healthcare, finance, legal, manufacturing, and even construction. These industries are huge, old, and honestly kind of stuck in the past with slow, paper-heavy workflows. YC has a ton of startups here: 41 in healthcare, 21 in finance, and 26 in industrials. Even tiny, efficient gains here can mean billions in value.
Second is the AI infrastructure layer: developer tools, infrastructure, and data tools. Basically, the platforms that help other people build and run AI apps. YC is betting big here, like 189 startups in dev tools, 21 in data tools. The idea is that these tools could be as foundational as AWS is for the cloud.
And finally, third is the agent economy. This is where the autonomous AI systems step in to handle tasks we usually give to people, like sales, customer support, recruiting, all these AI-powered agents. We are looking at a foundational shift in how work gets done.
So that's it. I hope this video was quite helpful and valuable, and let me know in the comments. Like and subscribe, and see you in the next.