Connecting your AI assistant
Last checked on
What you can do when you connect the AI assistant you already use to Trustbird, what it cannot do, what you can ask it, and where your data goes.
Why this protects your organisation
An assistant can save you time with questions about your organisation and with preparing work. It only helps if you know what it sees, what it may hand in and what stays with people. That is why this article says plainly where the data goes and that a person in Trustbird always decides.
Steps
- Ask an owner of your workspace whether the AI connection is on. It is off until an owner switches it on.
- Open the settings of your AI assistant and add a connection to an MCP server, sometimes called a connector.
- Enter the address of your workspace. The owner finds it under AI connection in the Workspace group.
- Sign in the way you do in Trustbird, including the second factor, and choose whether the assistant may only read or may also hand in concepts.
- Ask your first question, for example what needs your attention today.
The connection lets an AI assistant answer questions about your organisation and prepare work in your workspace. It acts for you: it works with your rights, in one workspace, and it stops as soon as the connection is ended.
Where your data goes
When you ask your assistant something, the answer from Trustbird goes to the AI service you chose for that assistant. That service can be outside the EU, and the agreement that Trustbird itself only uses Mistral in the EU does not apply to it. That is why an owner has to switch the connection on, and why it can be switched off at once.
What the connection can do
An assistant that may read can look at your attention list, tasks with the reason they exist, risks, measures, suppliers, assets, people, incidents, deviations, established documents, evidence without file contents, and the requirements of the standards you adopted with the measures that meet them. Answers come in your language and start with the action or the business risk.
With the right to hand in concepts it can also create a task, incident, risk, supplier or piece of evidence, suggest a change to an existing record and suggest an answer to a security questionnaire.
What it cannot do
An assistant never decides. It cannot mark a requirement as met, accept or close a risk, establish or publish a document or a Statement of Applicability, close an incident or record an audit conclusion. What it hands in is a concept or a suggestion, and a person in Trustbird takes it over or rejects it. Trustbird helps reduce cyber risks structurally; it does not guarantee that an assistant reads everything correctly, so check what matters.
What you can ask
- "What needs my attention this week?"
- "Which risks have no owner yet?"
- "Which measures have evidence that is about to expire?"
- "Make a task to review access rights of the finance team, due end of next month."
- "How do I add a supplier in Trustbird?" The assistant can look this up in this manual, in your language.
The manual for your assistant
The connection also offers this manual to your assistant, so it can explain how something works in Trustbird. It reads the articles in your language, and it can search them.
If it does not work
My assistant cannot connect
Check that the AI connection is on. While it is off, the address answers as if the workspace does not exist. Also check that you entered the address of your own workspace.
The assistant does not see something I see in Trustbird
It sees what you may see, and only what is established: drafts of documents are not shown. Ask the owner if you think your role is too narrow.
The assistant may not hand in anything
Advisers can only read through an assistant. Other members choose at sign-in whether the assistant may also hand in concepts. Sign in again to change that choice.