The Decision to Apply to YC
I applied to Y Combinator from the UK, where I had been building Clarm for about six months. By that point we had early customers, real results starting to come in, and a clear thesis. But I also had a strong sense that I was operating at the wrong speed—that there was a version of this company that moved faster, thought bigger, and was surrounded by people who would push it in directions I could not see from where I was.
When Clarm was accepted into YC X25, I made the decision to move to San Francisco without much deliberation. It was the right call. I say that now without any reservation.
Going to Silicon Valley changed my life as a founder. It is like the F1 of the world. People are going so fast and thinking so much bigger—there is so much to learn, and so much of it is learned not from lectures but from proximity.
What Changes When You Are in the Room
There is something that happens when you spend time with founders who are operating at a genuinely high level. Your baseline shifts. The things you thought were ambitious start to feel like reasonable minimums. The things you thought were impossible start to feel like problems others have solved and you just have not figured out yet.
I had a conversation early in the batch with another founder who was making 100 cold calls a day. Not as a one-off sprint—as a daily practice, consistently, for months. I had been congratulating myself on doing 20. That conversation recalibrated something in how I thought about what was possible.
You are very influenced by the people you spend your time with. Who you share an office with is really important. If someone is doing 100 cold calls a day, you start to believe you can do that as well. It does something to your brain. Your internal model of what normal looks like adjusts upward.
For founders who cannot make it to San Francisco: the principle still applies. Find the right people in your city. Spot them early. Surround yourself with them. You can get to some level of Silicon Valley in the place you are, if the people around you are operating at the highest level they can reach.
The AI Revenue Proof Problem
The most important thing YC taught me was something specific to building an AI company: the difference between proving that your AI works and proving that your AI generates revenue.
These are not the same thing. They are not even close.
Before YC, I was spending a lot of energy on capability proofs. Look how accurate the responses are. Look at the latency. Look at how well it handles edge cases. These are genuinely important. But they are product metrics, not business metrics. A technical founder defaults to product metrics because that is what we are trained to optimise.
YC pushed me, consistently and relentlessly, toward revenue proof. Not “the AI answered correctly”— “the AI caused a customer to pay.” Not “the AI saved time”—“the AI saved $X in costs or generated $X in new revenue.”
The question that changed my framing was simple: “What would the customer have done without you?” If the answer is “done it manually,” the next question is “and what would that have cost them?” If you can answer that clearly, you have revenue proof. If you cannot, you have capability proof.
How This Changed the Way I Measured Clarm
Before YC, I was measuring Clarm by conversations handled, response accuracy, and support deflection rate. Good metrics. Useful signals. Not revenue proof.
After about six weeks in the batch, I rebuilt our measurement framework entirely around revenue attribution. The question we now answer for every customer is: what pipeline did Clarm surface that would not have existed otherwise, and what did it convert into?
That reframe produced numbers that were much more compelling—and much more honest about what we were actually doing. GiveLegacy went from $0 to their top inbound revenue channel in 90 days using Clarm. 25.2% of conversations showed buyer intent—that is a conversion rate, not a vanity metric. It answers the question “what would they have done without you” with a percentage anyone can benchmark against.
It also changed how we talked to prospective customers. Instead of leading with accuracy rates and response times, we led with pipeline outcomes. Instead of “our AI handles up to 94% of support questions automatically,” we said “one customer went from 760 email inquiries to 4,624 conversations with 25% buyer intent, and it had a 25% buyer-intent rate—from a channel that was producing $0.” Those are fundamentally different conversations.
Three Things YC Changed About How I Think About Growth
1. Revenue is the only proof that actually transfers
Technical proofs are convincing to technical buyers. Revenue proofs are convincing to everyone. If you are building an AI product for a business audience, your goal is to accumulate examples where a real customer paid for something and the outcome was unambiguously better than the counterfactual. Everything else is a proxy.
2. Do things that don't scale, until the lessons transfer to things that do
This is old YC wisdom, but the AI version of it is worth being explicit about. In AI, the things that do not scale are the high-touch customer deployments where you are personally involved in every decision. The lesson you extract from those deployments—the pattern of what causes a conversation to convert, what question predicts a deal, what signal your customers care about—that lesson scales.
I spent a lot of time in early deployments personally reviewing conversation logs. It felt like it was not scaling. It produced the training data and the pattern recognition that made our intent detection actually work. Do the unscalable thing first.
3. The company that knows its unit economics wins
YC is relentless about unit economics. What does it cost to acquire a customer? What is the LTV? How does the ratio change at scale? For an AI company, the additional question is: what is the compute cost per outcome, and how does it move as volume increases?
Knowing these numbers changes how you sell. When your CAC payback is three months and your LTV is measured in years, you sell with confidence. When you do not know these numbers, you undercharge, over-discount, and make decisions that look fine at one customer and catastrophic at a hundred.
What I Would Tell a Founder Applying to YC Today
Get the revenue proof before you apply. Not a promise of revenue. Not letters of intent. Real customers who paid real money because your AI did something that changed a measurable outcome for them.
YC can help you accelerate almost everything. But it cannot give you the fundamental insight that only comes from a paying customer who says “this is worth it.” That is what YC is looking for. That is also what you should be looking for, whether or not you apply.
And if you are building something in the AI revenue or inbound space, I am genuinely happy to talk about what we learned and how we approach this. Reach out directly.
Related Reading
For the specific revenue proof we built with early customers, read the founder origin story behind Clarm. For the tactical side of building inbound revenue without a team, see why founder-led sales breaks and what to build instead.