The thinking behind better data decisions.
Written by the people making them.
Written from inside the problem, not above it. Real decisions from building a governed AI analytics platform - what the data team bottleneck actually costs, why most AI analytics tools skip the foundation, and what it takes to make data trustworthy before you query it.
Anonymizing column names before sending schema to an AI is standard advice. What most teams miss: the AI still has to understand your data somehow, and anonymization changes what that costs.
A 2026 benchmark across 37 LLMs found hallucination rates between 15% and 52%. Single-model analytics pipelines inherit that variance. Here's why a two-stage architecture - Classifier then Synthesizer - produces deterministic, trustworthy answers.
49% of organizations don't trust AI-generated insights. The reason isn't the model - it's that the AI has never been told what your columns actually mean. Here's how to fix it.
80% of data governance initiatives fail by 2027, says Gartner. The reason isn't the tool - it's ungoverned data. Here's what governance actually means and how to fix it.
Anthropic's engineering team published the internal architecture they built to make AI analytics trustworthy. The four decisions they made are the same four decisions Edilitics is built on.
The NCeG conference. The second AI reckoning. Building AskEdi. The decision not to launch. Five years later, what it actually means.
In early 2024 I wasn't looking for a co-founder. I didn't think I was in a position to ask. This is what happened in the week Mihir Sanchala said yes anyway.
Two and a half years. Two interns. One question nobody had answered. Then ChatGPT arrived and I spent three months wondering if any of it still mattered.
A decade in retail and growth taught me how to read a business. It didn't give me access to my own data. That gap became Edilitics.