Workshop

AI Agents for Economists Beyond the Chat Window

This workshop introduces AI agents through the perspective of real economist workflows. The focus is not on abstract demos, but on what changes when an AI tool can directly read files, run commands, and work across the software people already use.

Audience: Economics faculty and graduate students

Prerequisites: No prior programming experience required

Overview

What the workshop is about

Most researchers use AI by copying material into ChatGPT, asking a question, and then copying the answer back into their workflow. That works, but it turns the researcher into a middleman moving information between the AI and the tools where real work happens: Stata, LaTeX, and the file system.

AI agents remove that middleman. Instead of chatting through a browser, you give an agent a plain-language instruction and it acts directly: reading files, writing code, running analyses, and producing output. That is not just faster. It also makes new kinds of multi-step work possible.

The workshop walks through one AI agent hands-on, but the concepts and skills generalize across the current generation of agent tools.

Topics

What we plan to cover

Chat vs. Agent

What changes when AI can directly access your files, tools, and terminal instead of only responding in a browser window.

Concepts, Skills, and a Practical Tour

The core vocabulary, the handful of skills worth learning early, and a practical tour of how an AI agent fits into real research work.

Live Demos

Examples from research, presentations, and teaching workflows that show where agents are genuinely useful.

Reusable Skills and Workflows

How recurring tasks can be turned into repeatable workflows instead of one-off prompts.

Opportunities, limits, and what changes for economists

The workshop will close by discussing what these tools genuinely improve, where they still fail, and what kinds of work they are most likely to reshape.

Outcomes

What participants should leave with

By the end of the workshop, participants should be able to use an AI agent for concrete research tasks, understand when agent tools outperform chat interfaces, and identify at least one recurring task in their own workflow that could be automated.

This session will also serve as a pilot workshop ahead of the Macedonia-Illinois Workshop in May 2026, so feedback from participants will directly shape how the material develops next.