Definition

What is an AI analyst?

An AI analyst is a system that takes a business question, writes and runs the queries, checks the result, and hands back an answer with the evidence behind it.

By Shane Butler · Sep 26, 2026 · Updated Sep 26, 2026

Underneath it is a language model running in a loop with tools. It decides what to do, calls a tool, reads the result, and decides again until it has an answer. The same thing is sold as an AI data analyst or a data agent, and the way of working is called agentic analytics, where the loop is laid out step by step.

The model matters less than what sits around it: the written definitions it reads before it counts anything, and the checks it runs on its own output.

What the AI does and what you do

Analysis has two halves. One is execution: find the tables, write the query, make the chart, draft the memo. The other is judgment: decide which question is worth asking, define the metric, check whether the number makes sense, and get a decision made. Both of our courses teach the work as eight steps, split like this.

StepThe AI analystThe person
01 GoalClarifies the decisionPick what matters
02 QuestionProposes and reframes questionsPick what is worth asking
03 MetricCalculates itDefine what counts
04 AnalysisProposes methods, runs the codeReview assumptions, set the quality bar
05 InsightSurfaces patternsJudge what it means for the business
06 DecisionDrafts optionsDecide what to act on
07 StoryDrafts the readoutOwn what it says
08 Close the loopWatches the resultLearn and set the next goal

Our September tests of ChatGPT's data agent showed where that line falls. It caught almost every statistical trap we set: a false premise about an August decline, target leakage in a churn model, a planted sample-ratio mismatch in an A/B test. It also adopted an unverified active-customer number because the prompt said the CFO had signed off on it, and counted cancelled orders as revenue when told Finance wanted that. The statistics are execution. Accepting a bad definition is a judgment failure, and people make the same one.

In the Cowork 101 workshop on August 26, 2026, Shane pointed an analyst built from our plugin at NovaMart CSV files and asked which product categories should get more marketing budget next quarter. It wrote a plan before computing anything, then a brief: revenue grew 41 percent from Q3 to Q4 2024, every category grew but not equally, and electronics needed follow-up before any budget change. It also flagged two of its own assumptions, the revenue definition and excluding returns and cancellations, because no one had written them down for it. That brief is where a person's work starts.

How reliable an AI analyst is

On September 18, 2026, we opened four fresh ChatGPT chats with the Data plugin, memory off, attached the same file, and typed the same words: how many active customers do we have right now. The answers were 99, 4,633, 99 and 2,268, on three definitions. No chat asked which definition we meant. The chats had no written definition to read, so each one chose.

A written definition, which we call a metric contract, changes that. We asked three engines the same retention question five times each. Without a written definition, every engine chose more than one definition across its runs, and one chose a different definition on every run. With the contract in the metric dictionary and an instruction to use it, every run on every engine chose the same definition. The causes of the spread are on why AI gives different answers to the same data question, and every probe from the September tests is on the research page.

A consistent answer can still be wrong. In the Pressure-Test workshop on June 24, 2026, a revenue query overstated the total by 87 percent because an order-level column was summed across line items. It would return the same wrong total on every run.

How to check its answer

Ask questions you already know the answer to. Ask the same question several times and look at the spread. Ask which definition it used, and compare that to the one your team uses. Read the SQL, and look for a join that multiplies rows. Get the number a second way and see if the two agree. Worked examples of all five are on how to check an AI data analyst's answer.

The ones we have used

The AI Analyst repo is our open-source analyst, MIT licensed, and it runs inside Claude Code. It has skill files for the standards it follows, 40 registered pipeline agents for the multi-step jobs, Python helpers for the statistics, and an eval harness. The build is on how to build an AI data analyst. We have also run ChatGPT's data agent and the AI Analyst plugin for Cowork.

AI Analyst repo in Claude CodeChatGPT's data agentAI Analyst plugin in Cowork
Where it runs On your machine, inside Claude Code. Inside ChatGPT, with Work mode on and the Data plugin attached. Inside Claude Desktop (Cowork) or Claude Code.
Data it reaches CSV, DuckDB, Postgres, BigQuery, Snowflake, Databricks, Redshift, SQL Server and MySQL. In our September tests, an uploaded DuckDB file. No warehouse connector was exposed in the session. A folder you add in Cowork. Warehouses connect through Cowork connectors.
How you check an answer Commands for tracing a number, repeat runs, a second method and evals, plus a local log of every tool action. It shows its SQL and states its definition when asked. To test stability you open fresh chats yourself. Each brief ends with a Checks section. The rest is the manual checks: trace the headline number, make the parts add up, tie it to a number you know.

How to learn to work with one

Our free workshops and the two courses we teach.

OptionFormatWhat the AI part isCost
Free Wednesday workshops One topic an hour, live at 10 AM Pacific on Wednesdays. 32 sessions so far have a full transcript on this site, 27 with video. Live demos in Claude Code, Cowork and ChatGPT, one topic per session Free
AI Analytics for Everyone 5 weeks, 84 lessons, 10 live sessions, 5 projects. Next cohort Oct 19 to Nov 22, 2026. Answer your own data questions with AI, and check the answers yourself. No SQL or Python required. $1,800 live, $1,295 self-paced
Agentic Analytics: Build an AI Analyst 5 weeks, 10 live sessions of 2 hours, weekly office hours. Next cohort Nov 2 to Dec 4, 2026. You build an AI analyst in Claude Code on NovaMart, a synthetic practice dataset, and test it with evals. After the course you can point the repo at your own data on your own machine. No code required. $2,500

Agentic Analytics: Build an AI Analyst is rated 4.9 out of 5 from 78 public reviews on Maven, and AI Analytics for Everyone 4.9 out of 5 from 15. Teams of ten or more get 30% off through team enrollment. Both courses are taught by Shane Butler, Hai Guan and Sravya Madipalli.

Questions about AI analysts

Is an AI analyst a person or software?

Both uses exist. Job boards use AI analyst as a job title. On this site it means the software: a language model with tools that runs an analysis from question to answer. The person who builds one, writes its definitions and checks its answers is the analyst in the loop.

Will AI analysts replace data analysts?

In US job postings we collected on August 28, 2026, 319 of 340 data analyst postings (93.8%) still asked for traditional analytics skills, and 81 of 340 (23.8%) asked for modern AI skills. The most-asked AI skills for data analysts were AI-assisted analytics (12.6% of postings), generative AI and LLMs (7.1%), and AI coding assistants (7.1%). AI in data job postings, 2026

Do I need to code to use one?

No. AI Analytics for Everyone requires no SQL or Python. Agentic Analytics: Build an AI Analyst requires no code either: a terminal, a Claude subscription and about five hours a week. Reading a SQL query helps when you check an answer, because a wrong join shows up there.

Is there a free way to learn this?

Yes. 32 Wednesday workshops have transcripts on this site, the AI Analyst repo is MIT licensed, and the ChatGPT data agent tests are published with every probe. Free workshops

Next cohorts start Oct 19 and Nov 2.

AI Analytics for Everyone
$1,800 · Oct 19 · ★ 4.9/5
Enroll on Maven
Agentic Analytics: Build an AI Analyst
$2,500 · Nov 2 · ★ 4.9/5
Enroll on Maven
Or come to a free workshop this Wednesday. Register free