May 26th at 9PM PST was the latest day when YC could’ve invited us for an interview. We lost any hope a week before that date because it was already almost 2 months since we applied, so we were pretty surprised to receive an invite at 3PM May 26:
We started to prepare immediately. We have gone through various useful resources, from YC’s guidelines to personal interview experience blogs [1] [2] [3] [4] [5] [6]. From these materials, we realized that it is going to be a 10-min high-pressure call with a large number of questions and digging into weak sides. Here is also some things that we marked as important for the partners:
The list of ~30 main questions came out naturally. We needed to practice them and identify weaknesses, so we were lucky to find 4 YC alumni who agreed to help us with that. It is interesting that YC themselves advice not to mock, but we found it very valuable in terms of adjusting to the format and highlighting problems. Here is what we’ve learned:
With that in mind, we’ve reiterated all of our answers several times. We divided the areas of responsibility and practiced more. The interview was scheduled for 5:30 am at my local time. I was quite nervous and managed to get only 1h of sleep.
We were let in 2 mins late. There were three partners, which were randomly throwing questions at us. Somehow we forgot to record the interview.
Here are some of the questions that we’ve remembered and our approximate answers to them:
What are you building?
ChatGPT capabilities for customer support and knowledge management providers. On-premise, secure and customizable.
Where do you face this problem yourselves?
This winter I (Sergei) was snowboarding in Kazakhstan, and I broke my leg. On my way back to the UK I spent an enormous amount of time searching Turkish Airlines FAQ how I fly and what docs I need. I realized that I just want to ask the question and get the answer.
How this relates to the copilot for customer support agents which you are building?
We thought about who might be benefiting from automatic answer generation the most and converged to agents because customers tend to skip self-service portals and “smart” chatbots to reach agents faster.
What interesting have you learned so far?
Agents don’t utilize knowledge bases that much, they tend to ask their more experienced colleagues or supervisors. We fit perfectly there because we use knowledge from previous chats so our solution acts exactly like a head of support.
Do you have experience building in the customer support domain?
Yes, in one of my previous jobs I was involved in a development of a tool that automatically categorized incoming tickets and helped search for similar ones.
How big is the market you are going after?
The knowledge management software market is $570B in 2022 and growing at 20% a year. What we’re focusing on now is small-middle customer support software providers, who have users and money, but not AI expertise. One client is estimated to bring around $100k a year, and there are 100-200 of them in the market, so we’re aiming at $10-20M a year within 2 years.
How are you going to expand?
We plan to go after libraries, archives, and other types of knowledge management software because we won’t have to change core technology much.
There also was some additional back and forth on the market and scaling strategy, but we don’t exactly remember. It took 18 mins and ended with mixed feelings. After a few hours of waiting we received a rejection letter:
We were so drained from the preparation process and the interview itself, so we weren’t even sad so much. What was interesting was that there were no questions about the team, while we considered this as one of our strongest parts.
Indeed, we don’t know exactly how to make this giant af. There is so much attention from big players on the AI topic right now that it will be almost impossible to sell to them in the future. Another issue is that our initial idea is not something unique and unexpected, and YC already have similar companies in their portfolio even from the previous batch, so the entry threshold for someone like us definitely goes up.
But it is not the end of the world for us. We will continue to do outreach, establish pilots, and try to convert users to paying ones. We will continue interviewing customers to find the unique set of features that will bring the most value to them. I believe that with the development of a product, we will understand the currently missing parts.
Continuing to work hard!