Spec-Driven Development with Claude Code: My Setup
How I use spec-driven development with Claude Code to build a real site: the six docs, the rules that keep chat out of them, and a real session.
AI for Analysts
Where AI genuinely helps an analyst's workflow, where it fails, and how to validate what it produces before it reaches a stakeholder.
AI Transformation Starts With Your Data, Not the Model
Why most AI transformation programmes stall, what the research says, and a practical data-readiness checklist analysts can run first.
Claude Code for Data Analysis: A Beginner Setup
How an analyst can set up Claude Code to work with CSV files, SQL and Python safely. Project folder, context file, permissions, Git and a first session.
Building a Simple Data Analyst AI Agent With Claude
What separates an AI agent from a chatbot for analyst work, a minimal worked example, and the guardrails to put around it before it touches real data.
Text-to-SQL Tools Compared: What Works, What Doesn't
General chat assistants vs BI copilots vs dedicated text-to-SQL products. What each does well, where each breaks, and how to evaluate one.
Why LLMs Hallucinate Numbers in Analytics (and How to Catch It)
LLMs predict plausible text, not computed results. See why they invent numbers in analytics work, the patterns to watch for, and how to catch it before it reaches a report.
What Is a Semantic Layer, and Why AI Analytics Needs One
A semantic layer defines your metrics once so people, dashboards and AI all use the same numbers. Learn what it is, what it changes for text-to-SQL, and how to start small.
How to Validate AI-Generated Analysis Before It Reaches Your Boss
A practical checklist for checking AI-generated analysis: reconcile totals, trace every number, test edge cases and label what is still an assumption.
AI for Data Analysts: Which Tasks to Hand Over, Which to Keep
A practical split of analyst work into tasks AI does well, tasks it can help with under supervision, and tasks that should stay yours, with examples.
Prompting an LLM for SQL You Can Trust
How to get reliable SQL from ChatGPT, Claude or Copilot. What context to give, how to use examples, and the checks that catch wrong joins and made-up columns.
Git and an LLM for Power BI: a new way to build DAX and M
Save Power BI as text, version it in GitHub and connect an LLM to write DAX and M. Why this workflow changes how analysts build, review and trust their models.
Why AI analysts fail on bad metric definitions
AI can write valid SQL in seconds, but it can't know which definition of revenue your CFO approved. Why a governed metrics layer matters, with benchmark data.