#  Beyond the Chatbox: Faculty Move From Chatting With AI to Building With It 

 





October 08, 2026

 

 

     ![Madeleine Woods talking at a table with students](/sites/g/files/omnuum6756/files/styles/hwp_16_9__480x270/public/2026-10/20260903_AILab_stills_07.jpg?itok=Lw6CQsZ_) 

 



 

Most faculty know AI as a chat window in a browser. This fall, the Bok Center's Learning Lab is offering "Beyond the Chatbox," a ninety-minute hands-on workshop that moves them a step further. Faculty work in the Claude and ChatGPT desktop apps, which can read and write files in a folder on the faculty member’s own computer. The session condenses the four-day workshop series the Bok Center ran for faculty cohorts over the summer.

The session is not a tour of finished products. "What we're teaching you today is a meta tool, a tool-making tool," Madeleine Woods, the Learning Lab's Assistant Director for AI Initiatives, told participants at a September session.

Participants start by downloading a set of example projects from GitHub repository into a folder set aside just for workshop materials and future AI work. Woods and her co-teachers have faculty create this folder as a first step. The AI models can read and change *only* what is in that folder. Participants also use their Harvard accounts on these desktop apps and learn about the associated privacy agreements and approved data-security levels.

Next, faculty are introduced to the Lab's working heuristic of inputs, operations, and outputs: gather the materials, write out the steps, and define the content and form of the final product. The biggest shift is moving where the work occurs and lives. In an AI chatbox, the conversation is the main record and artifact of faculty work. This linear, dialogic form can be limiting. In this workshop, the faculty's main record and artifact becomes a system of folders. Faculty keep their materials, saved instructions, and results there, and any new chat can pick up where the last one stopped.

In one of the session's worked examples, participants start with a disorganized collection of photographed cards from a teaching fellow (TF) training activity. Each table of TFs had arranged cards with academic work tasks along an axis of low to high comfort with AI taking over those tasks for either students or teachers. Working with an AI model in the desktop app, faculty renamed the inconsistently labeled photos, read the cards and their position into a spreadsheet, and turned the spreadsheet into an interactive web page. Participants then checked one AI spreadsheet against the original image of the table, card by card. The same steps work for a common classroom task: collecting and reading students' handwritten in-class work.

The session also helps faculty understand how AI tools work "under the hood." A tokenizer page shows how a model breaks text into pieces. Printed copies of a system prompt for the Fable model, published by Anthropic, sit on the tables for close reading. These are Anthropic’s instructions to the Large Language Model (LLM), which load before the user's first message and never appear in the chat. Woods goes on to explain where this system prompt fits into the context window (the amount of text a model can work with at one time), and what happens when a long conversation or work process fills this window. "AI can feel like a frictionless process, because the companies want it to feel like a frictionless process," Woods told participants. "What we're going to do is reintroduce friction, to get you to the level where you are making decisions again."

Participants leave with the desktop apps for Claude and ChatGPT installed on their own machines and the beginnings of a project from their own teaching or research, plus a folder of example projects to keep building from, including a makeup-exam generator and a manuscript-transcription project. Ultimately, "you know what's important for your discipline: what can be used with AI, and what should be maintained \[for humans\]," Madeleine told one group. "We're just here to support you in that project."

Faculty can [register for an upcoming session.](https://bokcenter.harvard.edu/generative-ai-events) Anyone at Harvard or beyond can [join the Bok Center's community Slack channels](https://bokcenter.harvard.edu/getting-involved) to stay current on AI in higher education tools, research, and case studies relevant to Harvard instructors. Alternatively, sign up for a weekly email digest of updates.



 

 

 



 

 

 Share on:- [     Facebook ](#)
- [     Twitter ](#)
- [     Linkedin ](#)