I’m stealing borrowing the idea for a writeup about Thunk, our startup, from my friend and business partner Binara. You can read his article by clicking here.
Let me get you up to speed real quick about what we do. Thunk is an intellligent AI assistant that helps anyone inside an organization get quick and easy access to business data, using natural language, without having to go through a BI Analyst or some BI dashboard and trying to figure out where things are. So, Business Data -> Simple English questions -> Results. Got it? Good. Now this is how we started out.
Binara was reading this book Building a Second Brain by Tiago Forte, back in 2022.
He was very much engrossed in it, and he forced politely urged me to read
it as well. In one of the discussions we had about the book, we realized that
there’s no tool out there that does everything the author lays out for a
personal knowledge management system, so why not build one! That’s how we
started off on Thunk. It was a note-taking app, a Personal Knowledge Management system we liked to call it.
After spending a little bit of time with the initial idea, we realized that we’re just reinventing the wheel: almost all the features we were proposing were either supported natively by similar platforms, or could be enabled with available extensions. That led to our first pivot. Why not hook into all that and just go to where the data already is? So then Thunk became a knowledge management hub where we’d connect with a user’s Notes, emails, calendar, etc and act as a search engine to find and pull out any bit of information.
By this time, I was considering leaving my then employer oDoc, but I did not have a solid plan on when to leave, how to leave, or what my next steps are. While I was mulling this, my now girlfriend Dinithi urged me to take the leap, so I built up some savings and took the plunge. I am very grateful to her, for that—among so many other things.
So I left oDoc, and started working on Thunk full time. And Binara was doing the market research side of things + anything else I didn’t wanna do. We quickly realized the amount of people that might be willing to pay for something like this is fairly small, and that we need to change. This is when we decided to pivot towards business users. Again, we slowly realized that this is a bit too niche and not as attractive of an offering as might’ve hoped.
During this time, Binara was working full time as a Business Intelligence Analyst at o2 Store, and he realized some of his pain points can be solved with an AI powered solution. So he went ahead and pitched it to me and said that he believes that is the direction Thunk should go in. Back then, I was very much an AI skeptic, and I believed that this will all be a bubble that will soon pop (it still might, but it sure does have some real use cases). So, I was rather annoyed by Binara’s proposal. But one of the things Binara is really good at is persuasion, and he managed to persuade me to “give it a try,” and “we don’t have to do it if you think it’s bad” 😒
The long and short of it is I ran an experiment to see how well an LLM could handle the tasks required to retrieve data and do some analysis on it. And it could, surprisingly well. Sure, it has its moments and does unrelated things sometimes: stuff like following instructions very literally and producing results the user did not expect, or doing unnecessary calculations, and so forth. But, it is reliable enough for our use cases. So, I took the plunge and YOLO’d it.
That’s where we are. As of writing, Thunk is in the hands of one client and we’re in talks with a few other potential clients as well to see what we can do for them. I’m excited to find out where our adventure will end up. 🚀