Live course on Maven · 5 weeks · 10 sessions

Agentic Analytics: Build an AI Analyst

Leave with a working AI analyst on your machine, a test suite that scores it, and a package you can show your boss.

10
live sessions of two hours
5
weeks, starter repo to scored system
0
lines of code required
Shane ButlerHai GuanSravya Madipalli
Taught by Shane Butler, Hai Guan and Sravya Madipalli
$2,500
★★★★★4.9/5
78 reviews on Maven
Next cohort
Nov 2 to Dec 4, 2026
Then
Jan 11 to Feb 13, 2027
Sessions
Tue and Fri, 2 hours, plus office hours
Enroll on Maven
Ten or more seats get 30% off. Team pricing
This course is for
Analysts and data scientists who want to build the system their team uses
Product managers and engineers who want an analyst that shows where every number came from
You need a terminal, a Claude subscription, and about five hours a week.
This course is not
A prompt engineering course
A coding course
A demo of what AI can do
A certificate for watching videos

What changes in five weeks

Before
An AI tool that answers differently each time you ask
Numbers in a deck you cannot trace to a query
Waiting on the data team for every question
No way to tell whether a change helped
One model, chosen by default
After five weeks
An AI analyst with written definitions it reads every time
A number with a source behind it, every time
Connected to your warehouse and publishing to your team
A test suite you re-run after every change
A comparison of two models, with the cost and accuracy numbers

What you build each week

Two live sessions a week, Tuesday and Friday. Every session ends with something built.
WEEK 1

Foundations of agentic analytics

Build one skill on Tuesday and a whole small system on Friday, and check it still works after its memory is cleared.

Welcome to Agentic AnalyticsDesign and build your system
You leave with
A working starter AI analyst with your first skill and a small agent system
WEEK 2

Design the system, connect MCPs and a warehouse

Move your work into the full system, point it at a warehouse, and publish to a workspace your team already reads.

Expand developed systemsMCPs
You leave with
Your analyst connected to data and publishing to your team's tools
WEEK 3

AI evals for AI analytics

Tuesday works without an answer key: does it agree with itself, is there a source behind every number, do four methods agree on one question. Friday adds the answer key: questions with known answers, and an AI judge you calibrate.

AI Evals I: trusting system outputAI Evals II: measuring system improvement
You leave with
A set of questions with known answers, a scored test suite, and an AI judge you calibrated
WEEK 4

Context engineering and management

Run the same metric three ways, then write down what a contested metric means so the analyst reads it every time. Change one thing, re-run the tests, read what moved.

Context Management I: context storesContext Management II: the improvement loop
You leave with
Written metric definitions your analyst reads, and one measured improvement
WEEK 5

Open source models and leading an AI-native data team

Decide where a model should run for your company, run the course tests on a second model, and reconcile a disagreement. The last session is a workshop on the system you built.

Open source models and CodexLeading an AI-native data team
You leave with
Your Analyst v1.0: the system, its scorecard, its definitions, a two-model comparison, and a one-page README

What you get

Two live sessions of two hours each week, one homework exercise of 30 to 60 minutes, and an optional office hour. Homework is never a prerequisite for the next session.

A complete working AI analyst repository with its configuration
A test suite with an AI judge you calibrated
20 hours of live instruction across ten sessions, plus weekly office hours
Step-by-step written guides for every session and permanent recordings
Alumni community and a free retake of a later cohort

What students say

From the 78 public reviews on Maven.
★★★★★

“Best course I've attended so far. Fully hands-on, with three instructors who were friendly, real, and not afraid to say 'I don't know.'”

Anurag
Vice President, AngelOne
★★★★★

“Loved it, gained actionable skills, would recommend to any analytics team.”

Diana
Director, Data Science, L'Oréal
★★★★★

“Especially helpful for beginners looking to explore how AI-enabled workflows can accelerate insights without compromising rigor or accuracy.”

Sudarshan
Data Scientist, Activision
★★★★★

“This course completely changed my idea of data science work of the future. I have a new direction for self-directed learning and a whole world of new ideas for my career.”

Chris
Senior Data Scientist, L'Oréal USA
★★★★★

“Practical, engaging, and a strong introduction to AI analytics. It was especially useful to see how Claude can support real-world business thinking and workflow design.”

Karan
Executive Director, UBS
★★★★★

“You'll come out of the course with tons of ideas for agents and skills you can create to make your company's data org operate more efficiently.”

Nico
Senior Analytics Manager, GoDaddy

Your instructors

Shane Butler
Shane Butler
Co-founder, AI Analyst Lab
10+ years in product data science, causal inference and AI evaluation at Stripe, Nextdoor and Ontra.
Hai Guan
Hai Guan
Co-founder, AI Analyst Lab
Ran data at Nextdoor, LinkedIn, Pinterest and Meta. 16+ years teaching product people to decide with data.
Sravya Madipalli
Sravya Madipalli
Sr Manager, Data Science at Superhuman
14+ years building data science teams at Microsoft, eBay and Nextdoor.
Taking the course as a team, or expensing it?

Ten or more seats get 30% off. Private cohorts run on your schedule. Maven issues an invoice you can expense, and we can send a short note for your manager on request.

Questions people ask before they enroll

Do I need coding experience?

No. Everything is built with markdown and natural language. You need a terminal, a Claude subscription, and about five hours a week.

How much time per week?

Two live sessions of two hours, one homework exercise of 30 to 60 minutes on a guided path, and an optional office hour. Homework is never a prerequisite for the next session.

Can I use my own company data?

Every exercise runs on NovaMart, the course dataset, so you are never blocked by access. You leave with a guided path for your own data, applied after the course under your company's rules.

What if I miss a session?

Every session is recorded and every session has a written guide. You also get a free retake of a later cohort.

How is this different from AI Analytics for Everyone?

AI Analytics for Everyone teaches the analytical thinking: the questions, the metrics, the decisions. This course teaches you to build the system that does the work and to prove its numbers are reliable. Many people take both.

Can my company pay for this?

Yes. Maven issues an invoice you can expense. Ten or more seats get 30% off, and we can send a short note for your manager on request.

Who is this course not for?

People who want a prompt-engineering course, a coding course, or a certificate for watching videos. Every session is hands-on building.

Next cohort starts Nov 2.

$2,500 · ★ 4.9/5 from 78 reviews · Or come to a free workshop this Wednesday. Register free
Enroll on Maven