Helping sales folks find case studies in seconds

Datasparq had 70+ case studies chaotically sprawled across various drives and folders. Sales folks wasted hours hunting for the right examples whenever prospects asked "Do you have experience with X?".

So, I built a chatbot that actually knows where everything is. I used Django and Python to connect our Google Sheets database to the Anthropic API. I designed a simple, but elegant interface—type your question, get case studies with metrics. Plus, access previous conversations.

The real challenge wasn't the tech—it was stopping the AI from lying through its teeth. Turns out LLMs are people-pleasers by default, happily inventing case studies that never existed. "Supply chain for aerospace manufacturing? Of course we have that!" No, bot. We don't.

Spent way too long wrestling with prompts to force it to stick to actual facts. Even now, outputs need checking, but it's remarkably honest for an AI.

What happened:
- Sales team finds case studies instantly
- Case studies are written in a consistent TOV
- Works live on client calls to find real-time relevant examples
- Everyone's stopped asking if we have relevant experience because they can just check

It's become the company's external memory.