A nine-part series for analysts: what AI can and cannot do, tokens, prompts, evals, RAG, agents, privacy, documentation, and a personal ops checklist.
- 1 What AI can and cannot do for analysis AI is a sharp intern for analysis, not a signed-off analyst. Part 1 of Practical AI for analytics people maps assist versus replace, where liability for numbers still sits with you, and when to refuse a generated answer. Scheduled · September 15, 2026
- 2 Tokens, context windows, and cost (intuition) Context is a suitcase, not infinite storage. Part 2 of Practical AI for analytics people explains tokens, why long pastes fail, how cost scales with schemas and history, and when chunking beats cramming. Scheduled · September 16, 2026
- 3 Prompt patterns for data work Good data prompts are contracts, not vibes. Part 3 of Practical AI for analytics people covers role, spec, output shape, and refuse rules, plus SQL-then-explain patterns that block invented columns. Scheduled · September 17, 2026
- 4 Evals for humans AI drafts look fine until a join or filter is wrong. Part 4 of Practical AI for analytics people shows human evals: spot-checks, golden questions, and a 10-prompt regression set you can rerun after every model or prompt change. Scheduled · September 18, 2026
- 5 RAG in plain English RAG means retrieve company docs, then ask the model with that context. Part 5 of Practical AI for analytics people explains the flow in plain English, when it helps, and when it will not save a messy knowledge base. Scheduled · September 19, 2026
- 6 Agents, tools, and harnesses An agent with tools can query, write, and break things faster than chat alone. Part 6 of Practical AI for analytics people maps harnesses, guardrails, and human-in-the-loop so automation stays useful instead of reckless. Scheduled · September 20, 2026
- 7 Privacy when pasting data into chat tools Pastes into chat tools are processing decisions. Learn a risk ladder, what never to paste, and how synthetic samples keep AI helpful without shipping real customer rows. Scheduled · September 21, 2026
- 8 AI for documentation and data dictionaries Use AI to draft data dictionary cards fast, then verify grain, formulas, owners, and sensitivity before anything is certified. Draft is cheap. Fiction with a badge is expensive. Scheduled · September 22, 2026
- 9 Building a personal AI checklist Close Practical AI with a personal ops board: expand AI-SQL checks to charts, Python, and stakeholder email, run a fifteen-minute weekly hygiene pass, and recap Parts 1 through 9. Scheduled · September 23, 2026