Hasta la vista, spreadsheets! With technical advisory from Modulai, Mediatool, now Camphouse, built a chat-based data visualization tool that lets users extract and summarize marketing campaign data through natural language queries. Instead of building reports by hand, customers can ask a question and get the data they need to make a decision in seconds.
Challenge
Camphouse's users struggled to locate and summarize data across the large datasets behind their marketing campaigns. The existing system meant either navigating through existing reports to find the right summary or building a new one manually. The company wanted to streamline this by serving data visualizations as direct answers to natural language questions.
Solution
Building on a feasibility study from Modulai, Camphouse used OpenAI's chat models with few-shot learning to generate the complex JSON structures needed for automatic data retrieval and visualization. The chat interface lets customers query the platform in plain language, turning a manual reporting task into an immediate answer.
Tools
The project used OpenAI's GPT models with LangChain for rapid prototyping. The models proved effective at in-context learning from a few examples, producing JSON-structured output that Camphouse's data summarization API could consume with minimal post-processing. Fine-tuning allowed flexible adjustment of the input and output schema, improving both accuracy and latency.
Value created
Modulai's feasibility study gave Mediatool a validated technical path forward, showing that few-shot learning with LLM's could reliably turn natural language questions into the structured output their platform needed. The approach went on to inform Camphouse's product direction, which today includes an AI assistant that lets users ask about campaign performance in plain language instead of digging through dashboards manually.
Learn more
