How to Use ChatGPT Projects: 5 Practical Use Cases

TL;DR

ChatGPT Projects automate repetitive work by combining project-wide custom instructions, uploaded files, and organized chats. The five demonstrated uses are document-grounded mentorship, weekly PDF-to-chart dashboards, PDF-to-CSV data entry, customer feedback action plans, and negotiation emails. Because instructions persist across the project, some workflows require only uploading a file with no new prompt; read on for setup steps, limitations, and practical examples.

Transcript

If your job involves tedious, repetitive tasks,  then stay tuned because today I'll show you how to use ChatGPT Projects to do those tasks with  consistently high quality, like making reports, manipulating data, tuning up your email quality,  and much more. Let's dive in. Let's start our journey on ChatGPT.com. You can click the card in  the top ri... Read More

Key Insights

• ChatGPT Projects are a paid-plan feature. If the projects section does not appear in the left sidebar of ChatGPT.com, the account is on the free plan and shows an upgrade plan button in the bottom left corner instead. OpenAI has promised projects for the free plan but had not shipped it.
• Custom instructions in a project are scoped to that project only. Whatever is written in the add instructions box is respected and considered by ChatGPT in all of its responses for chats inside that project, and nowhere else, which keeps different workflows from contaminating each other.
• Uploaded files become context for every chat in the project. In the mentorship example, two documents containing principles on building great products were dragged in through the add files button, and the instructions told ChatGPT to tie each piece of advice back to the original principle it came from.
• Once instructions are set, a file can be submitted with no prompt at all. In the weekly dashboard, data entry, and customer feedback examples, the file was dragged in and submitted directly, and ChatGPT produced the requested output because the instructions already described the expected input and output.
• Good custom instructions are built through trial and error. The chart instructions were refined by asking for the charts, then adding instructions telling ChatGPT not to do the things it did wrong, plus a final instruction to stop it from ad libbing and keep it on track.
• Output consistency must be specified explicitly. When the same report workflow was run a second time, the bar widths in the generated charts differed, and the fix is to put the consistency requirement into the custom instructions rather than expecting identical formatting by default.
• Custom GPTs are limited to one model while projects support many. Custom GPT chats run on GPT-4o, whereas projects offer more options including the reasoning model and some legacy models, making projects better when model choice matters for the task.
• Projects cannot be shared, which is their biggest disadvantage. A set of custom instructions meant for a team should be built as a custom GPT instead, because a GPT can be shared using its link while a project has no sharing mechanism at all.

Questions & Answers

Q: What are ChatGPT Projects, and how can they automate repetitive work?

ChatGPT Projects organize chats while adding custom instructions and uploaded files that provide context across the project. Once the expected input and output are defined, repetitive workflows such as report creation or data entry can run by submitting a file with no additional prompt.

Q: How do I create and organize a ChatGPT Project?

Click the new project button, enter a name, and click the create project button. You can drag past chats into it or use the three-dot menu beside a chat, select add to project, and choose the project. A new prompt submitted while the project is selected becomes a chat inside it.

Q: Do I need a paid ChatGPT plan to use Projects?

At the time demonstrated, Projects required a paid plan. If the Projects section was missing from the left sidebar of ChatGPT.com, the account was on the free plan and showed an upgrade plan button in the bottom-left corner.

Q: How do custom instructions work in a ChatGPT Project?

The add instructions box defines how ChatGPT should help in every chat within that project. Those instructions affect only the project’s chats; the demonstration requested bullet points, a sixth-grade reading level, and a fun emoji at the start of each response.

Q: How do uploaded files provide context in ChatGPT Projects?

Files added to a project become context for every chat in that project. In the mentorship example, two documents containing product-building principles were uploaded, and the instructions required advice to be tied back to the principle from which it originated.

Q: How can ChatGPT Projects create charts from a weekly PDF report?

The project instructions can specify a PDF input, total weekly revenue, a horizontal bar graph of revenue by store, and a vertical bar graph of revenue by product line. After saving those instructions, uploading and submitting the report produced the total and both charts without an additional prompt. Formatting consistency, such as matching bar widths, must be requested explicitly.

Q: Can ChatGPT Projects convert PDF data into CSV format?

Yes. The demonstrated data-entry project used one part of its instructions to define the PDF-to-CSV task and another to specify the CSV format. After the files were attached and submitted without a prompt, ChatGPT returned the information in the requested table structure for CSV download.

Q: What is the difference between ChatGPT Projects and custom GPTs?

Projects are simpler to create, support multiple models, and keep their instructions and uploaded files within a project, but they cannot be shared. Custom GPTs can be shared by link and can use actions that integrate external services, but they are limited to one model, and their chats appear in the main feed rather than being movable into projects.

Summary & Key Takeaways

• ChatGPT Projects live in the left sidebar of ChatGPT.com and require a paid plan. Free-plan users see an upgrade button in the bottom left instead. Creating a project takes a name and a click of the create project button. Past chats can be dragged in, or added through the three dots next to a chat name by choosing add to project.

• Beyond organization, projects have two extra features: custom instructions and file uploads. The add instructions box answers the question of how ChatGPT can best help across every chat in the project, and its contents are respected in all responses within that project only. The add files button attaches documents that become context for every chat in the project.

• Five use cases are demonstrated: a mentorship project grounded in principle documents, a weekly dashboard project that turns a PDF sales report into revenue charts, a PDF-to-CSV data entry project for insurance plans, a customer feedback project that produces an action plan, and a negotiation email project written in the style of Chris Voss.

• Projects differ from custom GPTs. Projects are simpler to create and support many models including the reasoning model and legacy models, but cannot be shared. Custom GPTs are more powerful through actions that integrate external services, can be shared by link, but are limited to one model and their chats appear in the main feed and cannot be moved into projects.

• Chats inside a project do not share context with each other. Every chat is independent, and one chat cannot reference data from another chat in the same project. Persistent context comes only from the project's custom instructions and uploaded files, not from sibling conversations.