Hand a matter to a colleague — with its full state — and let their AI pick it up where yours left off.

Business situation

An associate at a London law firm is about to go on leave in the middle of a dispute for ABC Limited. A colleague will run the matter while they are away. The colleague needs to know the key dates, what has been agreed with the client, what is outstanding and what the other side last said.

Why existing tools alone are insufficient

Handover notes are written from memory in the last hour before leave. What each fee earner has worked out with their AI lives in that person’s chat history, which the colleague cannot see — and the colleague uses a different AI. Giving the colleague’s AI broad access to the document system to “catch up” ignores the matter’s team boundaries.

Systems involved

• Kroy State — the matter’s key dates, open tasks, decisions and next steps.
• SharePoint — the matter’s documents.
• Outlook — correspondence filed to the matter.

Kroy architecture

Both AIs connect to Kroy. The matter’s State is kept in Kroy and updated as work happens, by people and by their agents. The Prepare Matter Handover Skill may read the matter’s State, documents and filed correspondence and create a handover note. Access follows the matter team in KroyDB, which only a supervising partner can change.

Workflow

1. The associate asks Claude: “Prepare a handover note for the ABC Limited matter.”
2. Kroy checks the associate, the agent and the matter team, and returns the matter’s State with state.read, plus recent documents and filed correspondence.
3. Claude drafts the note — upcoming deadline, the client’s instructions on settlement, the outstanding disclosure request — and saves it with artifact.create.
4. The associate asks: “Add my colleague to the matter.” Claude requests matter.team.update. Kroy refuses: agents cannot change a matter team.
5. Before being added, the colleague asks ChatGPT about the matter. Kroy refuses: the colleague is not yet on the matter team.
6. The supervising partner adds the colleague to the matter team.
7. The colleague asks ChatGPT: “What do I need to do on ABC Limited this week?” Kroy returns the same State and the handover note. ChatGPT answers, and records the next step with state.update.

Agent permissions

Each AI acts under its own person’s delegation and sees only the matters that person is on. Both can read and update State. Neither can change who is on a matter.

Human permissions

Fee earners work on matters they are staffed on. The supervising partner controls the matter team.

State changes

• A handover note Artifact is linked to the matter.
• The matter team gains the colleague, recorded against the partner who made the change.
• The matter’s next step is updated by the colleague’s ChatGPT.

Audit outcome

Kroy records the handover note, the refused team change, the refused early access, the partner’s reassignment and every later read and update — with each person and each agent.

Security considerations

• Context moves between AIs through Kroy State, not through copied chat transcripts.
• Access follows the matter team, which agents cannot change.
• Microsoft 365 credentials stay in Kroy.

Setup requirements

• Matters and matter teams modelled in KroyDB.
• Microsoft 365 connected to Kroy.
• Claude and ChatGPT connected to Kroy.
• The Prepare Matter Handover Skill enabled for the litigation team.