Very first thing I realized? AI hackathons are positively not only for engineers.
This shocked even me — a few of our most spectacular tasks got here from the least technical crew members. Yoni, a producer at Aleph with no technical background, used Cursor to construct a Chrome extension that crawls LinkedIn to search out related individuals to interview for our podcast. Erica, our Chief Communications Officer, constructed a customized GPT that completely captures her writing model and may generate social media posts based mostly on our podcast transcripts.
Ron, who filmed every little thing, used AI instruments to create an extremely artistic recap video of the hackathon; he even used an AI reproduction of me to spare me from filming and re-filming myself (a lifesaver)! See right here:
The non-engineers did want some assist, although. That’s the place having “floating” engineers who can help with choosing the proper instruments, debugging, and deployment made an enormous distinction.
For instance….
Our Finance crew wished AI to automate emails, financial institution statements, and legacy methods. Inside hours, we hit a wall.
I had one mission throughout our hackathon: assist our Finance crew construct one thing actual. I promised two hours of my time, however ended up embedding with them for the whole occasion.
Noa and Shani (each non-technical) got here ready with concepts for coordinating emails, processing financial institution statements, and updating legacy methods with out APIs. Good candidates for automation, proper? “That is easy,” they thought. “Doesn’t even require any actual intelligence.”
Effectively, seems these issues are nonetheless not realistically solvable inside a 24-hour timeframe.
Your first concept is (in all probability) flawed. AI isn’t magic. Be able to pivot quick.
By mid-afternoon, we pivoted to one thing game-changing-yet-achievable: AI brokers that monitor Gmail and draft responses mechanically based mostly on configurable prompts. The brokers might be custom-made to deal with particular forms of emails and entry our CRM when wanted. Particularly, we wished to automate away the duty of dealing with an investor asking to vary their financial institution particulars, or their listed deal with. These will not be as uncommon as you might suppose, and contain infinite forwards and backwards emails to acquire all the mandatory documentation required by regulation to facilitate that — resembling proof of deal with, proof of banking, private particulars, and many others.
The lesson? Most tasks fail earlier than they begin — as a result of the concept is simply not a very good match for a hackathon. Have engineers accessible to information non-technical groups towards possible options to issues value fixing.
And on that word: right here’s methods to run your hackathon.
