LLMs are trained to interpret language, not data. Bridging the gap between AI and data means obsessing over context, semantics, and performance.

Tinybird is a real-time data analytics platform for building fast and scalable data pipelines, offering tools for data ingestion, processing, and visualization. From this blog users can learn about stream processing, data warehousing, and analytics techniques to unlock insights from their data and make data-driven decisions.

Tinybird

LLMs face significant challenges when working with analytical data, struggling with tabular data interpretation, SQL generation accuracy, and complex database schemas. The key to successful agentic analytics lies in providing comprehensive context through detailed documentation, semantic models, and sample data rather than expecting perfect SQL generation. Building query validation loops with error feedback, using LLM-as-a-judge evaluators, and focusing on business understanding over technical perfection enables more reliable analytical insights.

Why LLMs struggle with analytics

Some of the problems marketers don't talk about

What we've learned while building the Tinybird MCP Server

Finally someone addressing a practical problem with agentic ai.
I have been thinking about a solution, but the limitation of data adds up to the vagueness for LLM.
Maybe synthetic data generation solves it to some extend.

add a analytical tool then pass the output to the LLM

Anything to do with numbers I just provide the LLM with the data schema and what the goal is and ask it to solve it with python code, then run the code locally. They’re good at writing code, but can’t do math at all.

Llms are always confident even when they are wrong. I always ask if you fully understand the task and if not ask me 3-5 clarifying questions. Trust me your response is going to be changed.