MenuStart a projectStart

AI & data

Staff handbook AI assistant for Microsoft Teams

A staff policy assistant answers everyday questions from your team using only the handbook and policies you have approved. Each answer quotes and links to the passage it came from, and questions on sensitive topics are passed to a manager instead of being answered.

An example of how the studio would build it. No live services are connected.

From a staff question to an answer with its source.

  1. Step 01

    Question asked

    A team member asks a question in Microsoft Teams, such as how to report sickness or what the uniform policy says.

  2. Step 02

    Topic screened

    Questions about safeguarding, a complaint, a grievance or someone’s health are routed to the right manager without an AI answer.

  3. Step 03

    Policies searched

    Claude calls a search tool on your MCP server, which looks only in the current, approved versions of your policies in Google Drive.

  4. Step 04

    Answer written

    Claude answers in plain words, quotes the relevant passage and links to the document and section it used.

  5. Step 05

    Gaps reported

    When the policies don’t cover a question, the assistant says so and logs it, so a manager can answer and update the document.

The result

Fewer interruptions, and answers that match the handbook.

The assistant can only read approved documents and can’t change them. If the policies don’t answer a question it says so instead of guessing, and sensitive topics always go to a person.

Before we connect it.

Where this is useful

Useful for nurseries, care providers, clinics and multi-site teams where the same policy questions reach managers every week, or where staff work shifts and can’t always ask someone. A sensible first version covers the staff handbook alone and runs with a small group, so wrong or missing answers are found early.

What we need to connect

  1. 01A Google Drive folder holding only the current, approved versions of your policies, with older versions moved elsewhere.
  2. 02Microsoft Teams access for the assistant, Claude API access and a small MCP server hosted where you are comfortable with your data being processed.
  3. 03The topics that must always go to a person, who should receive them and who keeps each policy up to date.

What needs careful handling

An assistant will confidently repeat an old policy if it can still see it, so only the approved folder is searched and each answer shows the document’s last review date. Some questions sound routine but aren’t, such as a sickness question that is really a wellbeing or safeguarding concern, so the screening rules lean towards passing things to a person. The assistant can still misread a policy, which is why every answer links to its source for anyone who wants to check.

Questions people ask.

Could the assistant give staff wrong information?

It can, as any AI model can, which is why each answer quotes and links the policy it relied on. Topics where a wrong answer could cause harm, such as safeguarding, are never answered by the assistant at all. The logged questions also show where the policies themselves are unclear.

Can staff use it on their phones?

Yes, wherever they already use Microsoft Teams, including the mobile app. Access follows your existing Microsoft 365 accounts, so leavers lose access when their account is closed.

Question asked → Answer given
Automation 116

Ask your business data in plain English with MCP

Hire recorded → Team told
Automation 91

New starter onboarding automation

All automation examples