Hello ,
This week’s newsletter is different. No SQL code. No database queries.
I lead a team of 30 data professionals in my day job. We use tools like Claude and Copilot to get oriented, draft, review, explain, and move faster through routine parts of analysis. The pace of change is real. The useful lesson is how to use these tools while keeping the judgment your team relies on.
That is what I mean by Agentic Analytics: a practical way to use AI tools like Claude, Copilot, and ChatGPT as part of your analytical process. You give the tool useful context, check its work, and keep responsibility for the decision.
That last part matters to me because I remember building an early pipeline report from SQL that a teammate had to run for me. The Excel report was useful, but I could not yet explain or improve the logic underneath it. Learning the database fundamentals changed that. AI gives you a much better starting point today. To use it well, you still need enough understanding to direct the work and stand behind it.
Over the next 13 weeks, I will share the methods and mental models I am seeing become useful in real analytics work, using safe Summit Adventures examples you can inspect and adapt. Today is the map.
How Most Analysts Use AI Today
If you’ve used ChatGPT, Claude, or Copilot for work, it may have looked like this: paste something in, get something back, and decide whether it is useful.
Sometimes the first draft is useful. Sometimes a query runs but misses an important group of customers. Without a consistent way to give context and check the output, it is hard to know which result you have.
This series is about building a working method: clear steps, Summit Adventures examples, and a habit of checking.
The 5 Levels of AI Leverage
Here’s the ladder we’ll climb together. Be honest with yourself about where you are today. There’s no wrong answer, and every level builds on the one before it.
Level 1: Learn with AI. You use AI to get unstuck. It explains a query you inherited, decodes an error message, or clarifies a concept that never quite clicked. You still do the work, with a useful starting point.
Level 2: Draft with AI. AI writes the first version of the query or the analysis, and you check it before anything ships. This is where most of the time savings live, and it only works if you can read what AI wrote. Next week I’ll give you the exact method I use for this.
Level 3: Analyze with AI. AI becomes a thinking partner. It explores data with you, suggests angles you hadn’t considered, pressure-tests your interpretation, and drafts the executive summary. The decisions stay yours.
Level 4: Automate with AI. Your best recurring analyses stop being things you re-do and become workflows that run on a schedule. The Monday-morning revenue summary writes itself, and you review it instead of producing it.
Level 5: Monitor with AI. You design an alert around a decision-relevant threshold. For example, an alert can flag dormant high-value customers and include the next action for the person who owns retention.
You may recognize yourself in more than one level. The point is not to score yourself. It is to choose one useful next step.
Why You’re Better Positioned Than You Think
Everything this newsletter has taught you about SQL is exactly what makes this ladder climbable.
AI can draft a query in seconds. But a drafted query that looks right and runs fine can still be wrong. Maybe the join silently drops the customers who never booked. Maybe a filter excludes the rows that mattered most. The person who can’t read the query can’t catch the mistake.
You can read the query. That gives you a way to catch a plausible-looking result before it reaches the business.
The habits you’ve built here still apply at every level: the Mister Rogers exploration process, the Empty Stadium instinct for finding what’s missing, and the So What Framework for interpreting the numbers. AI can speed up that work. You still use those habits to direct it.
Where This Series Is Going
Over the next 12 weeks, each issue will teach one method you can use on its own, with examples against the Summit Adventures database (the fake adventure tourism company I created to help people learn business analytics). When I show an AI exchange or an automated artifact, it will be a reproducible Summit example or clearly labeled as an illustration. Some highlights:
The Draft-and-Check Method: the four-step way to let AI write SQL without losing control of quality
The briefing document: a reusable way to give AI useful context
AI-assisted exploration, executive summaries, and data quality audits
A reusable workflow, automation design, and monitor design
You can use each issue on its own. You do not need to install anything unusual or build an automation to benefit from the first methods.
This Week’s Action Item
1. Think of a recurring report or analysis you produce, something you re-do weekly or monthly.
2. Finish this sentence: “If this ran at Level 4 or 5, it would ___ without me.”
3. Write it down, along with the decision it is meant to support.
Hold onto it. It gives you a real use case for the methods ahead.
Until next time,
Brian ([say hi on twitter!](https://twitter.com/briangraves))
—
P.S. Next week, I will show you the first method in the series: how to let AI draft a SQL query while you keep control of the logic.
Want More SQL Insights Like This?
Join thousands of analysts getting weekly tips on turning data into business impact.