What happens to your team's AI knowledge when the AI champion leaves?

Short answer

What you keep depends on where the knowledge lives and which workspace controls apply. Shared repository files can remain available, while private learning may be inaccessible to teammates. In managed ChatGPT workspaces, projects and GPTs are reassigned on removal; in Claude, project access depends on sharing settings and exports may preserve data.

Where an AI champion’s knowledge actually lives

Almost every team has one person who gets the most out of AI. They have the prompts that work, know which tool is good for what, and have explained to the AI again and again what your quotes, code or client emails should look like. The catch: most of that lives in their personal account, not in anything the team owns.

Where it lives Does it stay when they leave?
The memory Claude built up about them No. Memory is per person and Owners can’t view or edit it
Their Claude chats and shared chat snapshots No. Teammates lose access after removal; the Primary Owner may still export retained data
Their own, unshared projects Depends on the tool and workspace settings
Projects and GPTs in a managed ChatGPT workspace Yes, ownership transfers to a workspace owner; private conversations and files are not automatically exposed
Cursor User Rules No. They are global to that user and not shared with the team
CLAUDE.local.md in Claude Code No. That file is meant for private instructions
A shared Claude project (instructions and knowledge) Yes
CLAUDE.md, .cursor/rules or AGENTS.md in the repo Yes, they are in version control
Cursor Team Rules and a managed CLAUDE.md Yes, the organization manages them
What they knew but never wrote down No

Undocumented knowledge is another handover risk. A champion learns something every week: that the AI writes too formally for client X, that an integration chokes on large files, that the first draft of a quote never applies the right discount rule. They just correct it in their own chat. Nobody else sees it.

The misconception: “but we’re on a team plan”

A team plan means shared billing and admin, not shared memory. For Claude, Anthropic says plainly that memory is individual and not shared. A shared project shares the instructions and knowledge written into it, not what members learn in their conversations. Whoever leaves takes that learning with them, even if everyone was on the same plan.

What you can do now

  1. Put team knowledge in shared places. One shared project per process, with one or two people allowed to edit it. See how to share a Claude project.
  2. Keep coding instructions in the repo. A CLAUDE.md in the project folder or rules in .cursor/rules travel with the team through version control; the Claude Code docs describe this as the way to share instructions. See how to share Cursor context with a team.
  3. Have the champion write down, monthly, what they keep explaining to the AI. Those are exactly the rules that otherwise disappear.
  4. Agree that personal projects and GPTs used for work get shared once more than one person relies on them. See working agreements for AI in a team.
  5. Add an AI handover when someone leaves. Which instructions, projects, connections and standard prompts did they use, and where do they live now? Then let the successor run the most important tasks on their own before the last day, so gaps show up while the champion can still answer.

This helps, but it stays manual. It depends on someone remembering, and it works per tool: Cursor or ChatGPT doesn’t know what sits in Claude.

What this means for a team

The risk grows when important knowledge and access depend on one person. And because every tool has its own memory and its own rules, the knowledge is also scattered across places that don’t talk to each other. So the question is not only “how do we keep their knowledge” but “why does that knowledge live in one person”.

Where Bombos fits

Bombos is a plugin that gives every AI tool your team uses, Claude, Claude Code, ChatGPT, Copilot, Cursor and your own agents, one shared memory. What the champion teaches the AI becomes a team rule that the connected AI tools can also retrieve for their colleagues, and it stays when they leave, traceable, in the cloud or on your own server. The champion can try it first for their own agents.

Related questions

Can an admin take over a departing employee's Claude memory?

No. Anthropic says Owners can't view or edit a user's individual memories. They can turn memory on or off for the organization, nothing more.

What should they do before their last day?

Move their best instructions, prompts and personal projects into shared places, and write down the rules they kept explaining to the AI in a shared project or a file in the repo. A few weeks ahead beats the last afternoon.

Do their personal connections go away too?

In Claude, an Owner adds a custom connector for the organization, but each member then connects on their own. Whatever they connected personally, their successor has to connect again.

Sources

  1. Use Claude's chat search and memory to build on previous context (Claude Help Center)
  2. What happens to a user's data when they are removed from a Team or Enterprise organization (Claude Help Center)
  3. Data retention when a member is removed from a workspace (OpenAI Help Center)
  4. How Claude remembers your project (Claude Code docs)
  5. Rules (Cursor docs)
  6. Get started with custom connectors using remote MCP (Claude Help Center)
The first half hour · free

What does your AI forget?

  1. 01You tell us which AI you use and where it keeps starting over.
  2. 02We show what a shared memory changes there, on your own examples.
  3. 03You keep the plan, with or without Bombos.
Piet Baudoin
Piet BaudoinYou talk to Piet, from the Bombos teamDutch team from Utrecht and Nijmegen · data in Europe
We reply the same working day.