Challenges Faced By Summit Intelligent Systems

Summit Intelligent Systems is going to be the biggest AI company in the world, but for right now we are a group of 6 high schoolers chasing a dream of providing free AI tools for local businesses in our area. This is easier said than done. The biggest challenges we face right now are:

Customer Acquisition Getting Past the Free Tier Barrier to deploy our AI services.

Customer Acquisition: This is a big one. It has been a month since I started this startup, and the biggest challenge we face is getting customers. Right now we have three major customers, obtained through cold emailing and cold calling. In the initial phases, this was very fruitful, as demonstrated by the customers we landed. However, in recent weeks it has been harder to use these methods to get more customers. Cold emailing is becoming slightly more cumbersome, as people on my team complain that they need to write an entire email on top of all their schoolwork, which I completely understand. Cold calls are another problem: we only have one or two really strong communicators on the team, and they are often busy. By the time school ends, most businesses have closed, so cold calls aren't always possible. Furthermore, actually finding these businesses is also proving to be a challenge, because it requires countless hours of searching for potential customers and local businesses. To combat this, I built an autonomous agent that searches for local businesses, scrapes their websites, and drafts an email, and at the very end, I review and approve it. This custom pipeline has enabled us to send far more emails than usual and has allowed more customers to discover us. So far, through this system we have gotten one response, a "no", but even that shows the system can work, and it will work in the future. To keep this running continuously, I set up a background process that runs the pipeline at 8:00 AM every single day.

Free Cost of AI Services: For context, I am a high schooler with no money, so in order to access these services and deploy tools for our customers, I have to build and deploy everything in a way that stays within the free tier. For instance, for one of our customers we needed to build an AI email chatbot. One of the early problems I ran into was that Render would not deploy the service because it exceeded the memory limit, and Render's free tier doesn't give you much to work with, so this was a major problem. To fix it, I had to lazy-load some of the embedding models and use an API for creating the vectorized database. Another example: for a different client, I'm building an AI chatbot that was initially deployed through Streamlit, which handled all the backend. However, Streamlit turned out to be a very finicky platform, so I had to transition it to a full-stack application. Not wanting to run into the same memory problems again, I turned my old Windows laptop into a server, and let's just say it sounds like a jet plane at night. It's running on overdrive, so hopefully it doesn't crash anytime soon. Hopefully, once I get more paying clients, I can upgrade these services and deploy everything on a proper server.

That's it for this blog. See you later!