If you’re using ChatGPT, Claude, or even a self-hosted tool like Open WebUI, there’s a good chance you’re still doing things the old way: opening a new chat, re-explaining who you are, uploading the same documents, and repeating your preferences every single session. It works, but it’s painfully inefficient.
Every major AI platform now offers some form of a “Projects” or “Workspace” feature. ChatGPT calls them Projects. Claude has Projects as well. Open WebUI offers a Workspace environment with knowledge management and custom model configurations. Despite being available for months, these remain some of the most underused features in AI tools today. This post is here to change that.
What Exactly Are Projects and Workspaces?
At their core, these features let you create a self-contained environment for a specific task, client, subject, or workflow. Think of them as folders for your AI conversations, but smarter. They don’t just organise your chats; they actively shape how the AI responds to you.
Here’s what you can typically do inside a project or workspace:
- Upload reference documents (PDFs, spreadsheets, code files) that the AI can access across all conversations within that project.
- Set custom instructions that tell the AI your preferred tone, role, formatting rules, or domain-specific context.
- Keep a focused chat history so that conversations within one project don’t bleed into another.
- Collaborate with teammates by sharing a project’s knowledge base, instructions, and conversation threads.
How It Works Across Platforms
The feature goes by different names depending on where you use AI, but the concept is remarkably similar:
| Platform | Feature Name | Key Capabilities | Availability |
| ChatGPT | Projects | Upload files, custom instructions, shared chats, project-specific memory | Free, Plus, Pro, Go, Team, Enterprise |
| Claude | Projects | Knowledge base with RAG, custom instructions, role-based sharing, connectors | Free (5 projects), Pro, Max, Team, Enterprise (unlimited) |
| Open WebUI | Workspace | Unified model/doc/prompt management, custom RAG pipelines, Python functions, user groups | Self-hosted (free, open-source) |
If you’re self-hosting with Open WebUI (which I do on my home lab), the Workspace is where you manage models, prompts, documents, and tools from a single dashboard. It’s less about per-project isolation and more about building a unified AI operating environment with RAG pipelines, custom functions, and granular user permissions.
Why You Should Actually Be Using This Feature
Let me walk through the practical reasons this matters, especially if you use AI tools regularly for work.
1. You Stop Repeating Yourself
This is the most immediate benefit. Without a project, every new chat starts from zero. The AI doesn’t know your role, your company, your writing style, or the documents you’ve been working on. With a project, you upload your brand guidelines once, your product documentation once, your style preferences once, and every conversation inside that project automatically has that context. For someone running multiple ventures or managing outreach campaigns, this alone saves a significant amount of time each week.
2. Responses Become Dramatically More Relevant
When the AI has access to your actual files and understands the context of what you’re doing, hallucinations drop significantly. Instead of generating generic answers, the AI grounds its responses in your uploaded materials. If you’re working on a sales outreach campaign, for example, and you’ve loaded your ICP definitions, competitor research, and email templates into the project, the AI can write follow-up sequences that actually reference your specific product and target audience.
3. Separation of Concerns
One of the underappreciated benefits is isolation. If you manage a consulting business, a product company, and freelance work simultaneously, mixing all of those conversations into a single AI chat stream is a recipe for confusion. Projects let you create distinct environments: one for client A’s marketing strategy, another for your product’s cold outreach, a third for blog writing, and so on. The AI stays focused within each boundary.
4. It Turns the AI Into a Persistent Assistant
With memory features evolving across all platforms, projects take the AI from a “one-off question answerer” to something closer to a long-term collaborator. Claude’s Projects now support RAG (Retrieval Augmented Generation), which means even if your uploaded files exceed the context window, the system intelligently retrieves what’s relevant. ChatGPT’s Projects can reference previous chats within the same project. Open WebUI lets you build full RAG pipelines with custom vector databases. The AI remembers, references, and builds upon your earlier work.
5. Better for Teams and Handoffs
If you work with collaborators, shared projects mean that a team member can step into a project with full context: the uploaded documents, the custom instructions, and the conversation history. This is especially powerful for agencies, consulting teams, or any setup where multiple people contribute to the same client deliverables or research effort. ChatGPT supports shared projects for Team, Enterprise, and Education accounts. Claude offers role-based project sharing on Team and Enterprise plans. Open WebUI provides granular permissions and user groups for multi-user deployments.
6. Custom Instructions Per Project
This is a subtle but important distinction from global custom instructions. You might want the AI to be formal and data-driven in one project (say, investor reporting) and casual and creative in another (blog content). Project-level instructions let you define these boundaries cleanly, without toggling settings back and forth.
Practical Use Cases That Work Right Now
Here are some concrete examples of how you could set up projects today:
- Job Search Command Centre: Upload your CV, target company list, and job descriptions. Set instructions for the AI to tailor cover letters and interview prep to your background and target roles.
- Client Marketing Project: Upload the client’s brand guidelines, audience research, and past campaign data. Every chat within the project produces copy and strategy that’s aligned to their brand.
- Sales Outreach Workspace: Load your ICP definitions, email templates, and CRM export data. Use the project for writing personalised cold outreach sequences, follow-ups, and LinkedIn messages.
- Content Writing Hub: Store your blog style guide, SEO keyword research, and content calendar. The AI writes drafts that match your established voice and SEO strategy.
- Product Documentation: Upload technical docs, API references, and user guides. Use the project for drafting release notes, FAQ updates, and support articles.
- Personal Finance Tracker: Upload bank statements or budget spreadsheets. Ask the AI to analyse spending patterns and generate visual reports within the project context.
A Note for the Self-Hosted Crowd
If you’re running Open WebUI with Ollama or any OpenAI-compatible backend, the workspace concept is even more powerful because you control the entire stack. You can build custom RAG pipelines using your choice of vector databases (ChromaDB, PGVector, Qdrant, Milvus, and others), create custom Python functions that extend your AI’s capabilities, set up model-specific configurations for different tasks, and manage documents centrally with multiple extraction engines. The flexibility here is unmatched. The trade-off is that it requires more setup, but for anyone already running a home lab or VPS infrastructure, the workspace becomes the command centre of your self-hosted AI stack.
Getting Started: It Takes Five Minutes
Setting up your first project is straightforward regardless of which platform you use:
- Create a new project: Look for the “New Project” button in the sidebar (ChatGPT and Claude) or navigate to the Workspace section (Open WebUI).
- Give it a clear name: Be specific. “Screenshotbot Outreach Q1 2026” is better than “Sales Stuff.”
- Upload your reference files: Add the documents, data files, or code that the AI should reference when answering questions in this project.
- Write custom instructions: Tell the AI what role it should assume, what tone to use, what to prioritise, and what to avoid.
- Start chatting within the project: Every conversation now benefits from the full context you’ve provided.
The Bottom Line
AI chatbots have evolved far beyond simple question-and-answer tools. The Projects and Workspace features across ChatGPT, Claude, and Open WebUI represent a fundamental shift in how we should be interacting with AI: not as disposable conversations, but as persistent, context-rich working environments.
If you’re paying for a subscription to any of these tools and not using projects, you’re leaving a massive amount of value on the table. And if you’re on a free tier, the good news is that both ChatGPT and Claude now offer projects at no cost (with some limitations).
Set up your first project today. Upload the files that matter. Write instructions that reflect how you actually work. Once you experience the difference, you won’t go back to starting from scratch.
