Blog
Notes from the unglamorous half
Engineering write-ups on parsing difficult documents, enforcing permissions where they belong, and keeping answers reproducible.
Connect Supabase, PostgreSQL, and MySQL to DataFact
DataFact now supports Supabase, PostgreSQL, and MySQL as structured-data sources for governed analysis.
Bring DataFact Agents to Slack and Telegram
DataFact now delivers governed Agent conversations through native Slack and Telegram channels.
AI Document Q&A With Page-Level Citations: A Practical Guide
See how AI document Q&A with page-level citations should preserve tables, figures, and source locations—and how to evaluate the evidence behind each answer.
AI Report Generation From Excel: A Controlled Workflow
Learn how a reusable template, worksheet mapping, task status, review, and download turn Excel data into a controlled AI report-generation workflow.
Document RAG vs. Text-to-SQL: Which Does Your Question Need?
Compare document RAG and text-to-SQL for enterprise questions, learn where each approach fails, and see when a multi-source AI workflow needs both.
How to Evaluate an Enterprise AI Data Analyst Before Rollout
Use this practical evaluation framework to test enterprise AI analytics for answer quality, permissions, evidence, reproducibility, and operational readiness.
Governed Text-to-SQL: What Enterprise Teams Should Require
Learn what governed text-to-SQL means, how it differs from a demo chatbot, and what to evaluate before giving business users natural-language access to data.
On-Premise vs. Private Cloud vs. SaaS AI Data Analysts
Compare SaaS, private-cloud, on-premise, and isolated-network deployment for enterprise AI analytics using a practical security and operations checklist.
Reproducing last quarter's number after the definition changed
The report said 41.2 percent in April. The same question returns 39.8 today, and nothing is wrong with either figure. Whether you can explain the gap in ten minutes or ten days depends on one architectural choice.
Row-level rules belong in the query, not in the prompt
If the only thing stopping an AI from returning another region's salary data is an instruction telling it not to, you do not have access control. You have a suggestion.
Why merged cells break retrieval, and what to do about it
A merged cell that flattens into the wrong row does not produce an error. It produces a confident, wrong number that looks exactly like a right one, and it is the most common silent failure in document AI.
Analysis rules are knowledge objects, not prompt text
Suppress cells under n=30. Control for age before reporting anything by marital status. Every senior analyst carries a hundred rules like these, and almost every AI analytics tool asks you to paste them into a prompt and hope.
The two halves of enterprise data AI, and why almost nobody joins them
Document AI reads your contracts but cannot query your warehouse. Analytics AI queries your warehouse but has never read the contract. The questions that actually matter to a business sit across the join.