The urge to build
I first met my co-founder Nadila Kulanatha in school at the Royal College Computer Society. We had a lot of fun building apps and doing other projects. I found a sense of fulfilment in being a part of a team that was building stuff and making things happen. We each played to our strengths and covered up for whatever we lacked, and at the end of the process, we had something that someone found valuable. By the end of my tenure in the society, I knew this was one of the few guys I respected and trusted enough to build something with someday. We were itching to build stuff. So I, Nadila and a few of my other friends from the Society would go on to start a design and development studio. After that stint, I ended up at o2 store, my first and current employer. Fun fact: Nadila forwarded the vacancy at o2 Store to me.
Unpacking the process
I joined o2 Store as a Business Intelligence Analyst, where I now handle everything from creating reports to developing interactive dashboards for our team and clients. The founder, Sanjeevan Jeevaratnam, recognised the need for more data visibility in the procurement process. The lack of easy access to data leads to complex workflows, wastage, and a lack of transparency. To solve this problem calls for robust data infrastructure.
Business users, such as sales and operations executives, constantly have data-related questions that inform business decisions. However, during my time at o2 Store, I realized that regardless of an organization’s BI infrastructure maturity, there will always be nuanced questions that require impromptu analysis.
Business users face dynamic situations at work and often need answers to unique ad hoc questions to tackle their work. However, the answers to such questions aren’t always readily available. So they reach out to me for guidance on how they can find the answer or they would require a new customized report, or just one-off answers. Depending on the complexity and uniqueness of the request, the best-case scenario is that I can generate the report or get them the answer within 10 minutes. In the worst-case scenario, I have to put my other tasks on hold and prioritise their request, ensuring they receive the necessary insights or answers they need within the hour. The ensuing, follow-up questions and feedback also add to the time spent on this process. This single workflow takes away time from both the business user and the analyst.
AI skeptic turned believer..?
Sanjeevan is a big advocate of AI and never fails to remind us about how it could make work easier for everyone. Back when I had just joined o2 Store I hadn’t used AI much, and I got frustrated to the point where I thought if AI is so great why hire us? However, I soon felt the benefits of AI myself each passing day and saw the potential for it to augment my work. Having realized that he was onto something I began to use LLMs even more which made my work somewhat easier, smoother and faster. However, there were problems with that as well. Generating accurate SQL for our organization’s unique data structure & schema was still a struggle. I’ve used ChatGPT, Gemini and Claude to write SQL with limited success and a lot of hand-holding, and spoon-feeding to the AI to get exactly what I need. Most of my complex queries that demand more nuanced logic and a holistic view of the business, have forced me to go back to writing SQL manually. One day I thought there should be a better way to do this. So I decided to work on a tool to make this whole process more seamless. The same tool would also allow business users to query the database in plain English, making my job and everyone else’s jobs easier.
Now here we are building Thunk for one of our first potential customers; o2 Store. Nadila himself was an engineer at oDoc and had left to be able to work on Thunk full-time. He was the first person to hear my crazy ideas and believe in me on this venture. I hope many more will believe in us and join us in our journey.
Love the insight the article provides! 👏
👏🏽