ChatGPT Projects: How to Build Persistent AI Workspaces That Actually Remember Your Business

What You'll Learn

• Why the "re-briefing tax" is the biggest hidden cost in how most executives use ChatGPT
• How Projects works technically, and what it can and cannot remember
• A concrete setup walkthrough for 3 high-value executive use cases
• The custom instructions that make Projects genuinely useful, not just organized
• Common failure modes and how to avoid them

The Scene

A CMO at a 60-person SaaS company uses ChatGPT every day. He has been using it since GPT-3.5. He opens it, types his question, gets his answer. Repeat.

The problem: every conversation starts from zero. He re-explains his company, his product, his team structure, his preferred tone, and his current priorities at the beginning of almost every session. He estimates he spends 5-10 minutes per conversation on setup context before he gets to the actual question.

Five conversations a day. Five minutes each. That is 25 minutes a day of context setup, not AI work. Over a year, that is more than 100 hours of time spent briefing a tool that forgets everything the moment the conversation closes.

ChatGPT Projects, launched December 13, 2024 for all plans, solves this problem. Not with more memory features, but by giving each work context its own persistent environment: its own files, its own custom instructions, its own conversation history. When you open a Project, ChatGPT already knows everything you have told it in that context.

The re-briefing tax disappears.

How Projects Works

A Project is a named container inside ChatGPT. Each Project holds:

Files: Documents, PDFs, spreadsheets, code files you upload stay attached to the Project. Every conversation inside the Project can reference them. Free users get 5 files per project; Plus users get 25; Pro/Business users get 40.

Custom instructions: A persistent set of instructions that apply to every conversation in the Project. These replace the need to re-explain context. You write them once and they apply automatically.

Conversation history: All conversations inside a Project are grouped together. You can search them, reference them, and the model can draw on them.

Visual organization: Projects can have custom names, colors, and icons so you can navigate between multiple work contexts quickly.

One critical note: Projects does not have long-term cross-conversation memory in the same way the Memory feature does. The custom instructions and uploaded files persist, but the model does not automatically recall every specific exchange from every prior conversation. Think of it as: the Project remembers the briefing documents and standing instructions, not the conversational history automatically. You can reference prior conversations manually by opening them, but they do not auto-populate into new conversations.

Setting Up Your First Three Projects

The right way to think about Project setup is: what are the three most distinct work contexts where you use ChatGPT regularly? Each context should have its own Project.

1. Client or account management

If you manage relationships with specific clients, each major account deserves a Project. Upload the account brief, recent communications, contract terms, and your notes on relationship dynamics.

2. Board and leadership communications

One Project for everything board-related: your current board deck, the last three board updates, your leadership team's OKRs, and any standing context about board member priorities.

Ready to master AI workflows?

Get full access to our prompt library and deep-dive strategy guides.