Articles in this section

See more

MCP - Using a Softools AI Assistant in Microsoft Teams

Avatar
Mark Dyer
  • Updated
Follow

Overview

If your organisation has made a Softools-connected AI assistant available in Microsoft Teams, you can use it to ask questions about your Softools data directly from a Teams conversation.

Instead of opening individual Apps, filtering reports and reviewing Records one at a time, you can ask the assistant to retrieve, organise and analyse information from Softools using everyday language.

For example, you could ask:

  • Which Projects need my attention first?
  • What are the highest-priority Projects and when are they due?
  • What Risks are associated with a particular Project?
  • Can you summarise a document I've shared with you?

The AI assistant uses Model Context Protocol (MCP) to retrieve information from Softools. Your organisation must first configure and publish the assistant through Microsoft Copilot Studio.

This article explains how to access an existing assistant in Teams, connect your Softools account and start asking questions.

 

 

Before You Begin

To use a Softools-connected AI assistant in Microsoft Teams, you will need:

  • Access to Microsoft Teams.
  • A Copilot Studio agent that your organisation has made available to you.
  • A Softools user account with permission to access the relevant Apps and Records.
  • An approved MCP connection between the agent and your Softools workspace.

You do not need to configure the MCP server yourself. This is handled by your organisation's Copilot Studio administrators and Softools Advanced App Builders.

The assistant can retrieve information from published MCP-enabled Views, subject to your existing Softools permissions.

 

 

Opening the AI Assistant in Teams

Your organisation may make its AI assistant available through Microsoft Teams.

To open it:

  1. Open Microsoft Teams.
  2. Locate the AI assistant in your Chats list, or open it using the link provided by your organisation.
  3. Select the assistant to open its conversation.
  4. Type your question into the message box.

You can interact with the assistant in much the same way as a normal Teams chat.

For example, you could begin by asking:

What Softools tools and information do you have access to?

The assistant can explain which tools are available, including published Softools Project or Risk Views.

Note: The available tools depend on how your organisation has configured the agent. Different organisations may expose different Apps and Views.

 

 

Connecting Your Softools Account

The first time you ask a question that requires Softools data, the assistant may ask you to connect your Softools account.

This ensures that the assistant retrieves information using your own Softools permissions.

 

Step 1: Set Up the Connection

Ask a question that requires information from Softools.

For example:

Which Projects have the highest priority, and what order should they be reviewed in?

If you have not yet connected your account, Teams displays a Connection Required message.

Select Set up connection.

 

Step 2: Connect to Softools

A Copilot Studio connection window opens, showing the Softools MCP connection configured by your organisation.

Initially, the connection may display a Not connected status.

Select Connect.

If no existing personal connection is available, select Create new connection.

You can enter an optional display name, such as My Softools Connection.

Continue to create the connection.

 

Step 3: Sign In to Softools

A sign-in window opens.

Sign in using your normal Softools account and complete any required authentication steps.

 

Step 4: Allow Read Access

After signing in, Softools displays an authorisation request for the AI assistant.

This identifies the connection and the Softools workspace it will access.

Review the details carefully, then select Allow read access if you recognise and trust the connection.

The authorisation allows the AI assistant to read information from Apps, Views and Records that you are already permitted to access.

The Softools MCP connection is read-only. It does not provide tools for creating, updating or deleting Softools Records.

 

Step 5: Complete the Connection

Return to the Copilot Studio connection window.

Once your Softools account has been authorised, select the newly created connection and choose Submit.

The connection should now display a Connected status.

Return to the Teams conversation and select Retry on the original connection request, if it is still displayed.

You can now ask questions using the Softools information available to your account.

Note: Where your organisation uses End user credential authentication, each user connects with their own Softools account rather than sharing a common account.

 

 

Asking Questions About Softools Data

Once connected, you can ask the assistant to retrieve and organise information from your Softools Apps.

For example:

Which Projects have the highest priority, and what order should they be reviewed in?

The assistant can retrieve Project information and present a prioritised list using available Fields such as:

  • Project Name
  • Priority
  • RAG Status
  • Delivery Date
  • Baseline Cost
  • Forecast Cost

You can continue the conversation with more specific questions:

Which of these Projects are overdue?

Which Red Projects have the greatest forecast overspend?

Explain why the highest-priority Project should be reviewed first.

The assistant can combine information returned from Softools with calculations and explanations to help you understand where attention may be needed.

 

Improving Accuracy with Field Descriptions

The example above also highlights an important consideration for App Builders.

The Priority Field contains values from Level 1 to Level 5, but the assistant initially cannot determine whether Level 1 or Level 5 represents the highest priority.

Rather than silently assuming the meaning, it identifies the uncertainty and explains the assumption used for its ranking.

This is a useful example of why clear Field descriptions are important when exposing information through MCP.

An App Builder could improve the Field description to explain that Level 1 represents the highest priority and Level 5 the lowest, if that is how the App is configured.

Providing this context helps AI assistants interpret Field values more accurately and produce more reliable responses.

Users can also clarify the intended meaning directly in their questions.

Important: Always review the underlying Softools Records before making important decisions based on AI-generated recommendations.

 

Working Across Multiple Apps

An AI assistant can also combine information from different Softools Apps, provided suitable MCP-enabled Views have been published.

For example, an organisation might expose:

  • A Projects App containing Project status, priority, delivery dates and costs.
  • A Risks App containing Risks linked to Projects.

After identifying a Project that requires attention, you could ask:

For New Product Research & Development Facility, what are the biggest Risks and in which order should they be addressed?

The assistant can retrieve the linked Risk Records and use their recorded information to recommend an order of attention.

In this example, the assistant identifies two Risks associated with the Project:

  • Specialist equipment delivery delays, recorded as Red with a Risk Score of 20.
  • Specialist resource shortages, recorded as Orange with a Risk Score of 12.

Using the available Risk Scores and Impact values, the assistant recommends addressing equipment delivery first while beginning resource mitigation in parallel.

It also identifies where further action is needed, such as confirming delivery dates, assigning mitigation responsibilities and addressing missing approval information.

This demonstrates how an AI assistant can combine information from related Apps and turn recorded data into practical recommendations.

The assistant should distinguish between information retrieved from Softools and actions or conclusions it recommends.

For more examples, see Using AI Assistants with Softools Data.

 

 

Sharing Documents with the Assistant

Depending on how your organisation has configured its Copilot Studio agent, you may also be able to upload documents directly into the Teams conversation.

For example, you could upload a Word document containing a proposed Project and ask:

Summarise this Project proposal, identify the main Risks and suggest which Actions should be addressed first.

The assistant can analyse the document and present its contents in a structured format.

In this example, the uploaded document contains fictional Project Orion test data. The assistant recognises that it is a proposed Project and keeps it separate from the live Project information retrieved from Softools.

 

Uploaded Documents and Softools Records

There is an important difference between:

  • Information retrieved from Softools through MCP, which comes from published Views.
  • Documents uploaded to the Teams conversation, which are provided directly to the AI assistant.

Uploading a document does not automatically create a Record in Softools or add the document to a Softools App.

The assistant may be able to compare information from an uploaded document with live Softools data, but this does not mean that the uploaded information has been saved in Softools.

Important: Only upload documents that your organisation permits you to share with its AI services.

 

 

Shared Files and Links

Some AI assistants in Microsoft Teams include a Shared tab alongside the Chat tab.

This provides a convenient place to find content shared in the conversation, including:

  • Recent — recently shared content.
  • Files — documents shared in the conversation.
  • Links — shared website or document links.
  • Media — shared images and other media.

In this example, the previously uploaded Project Orion document appears in the Shared tab.

 

Asking Questions About Previously Shared Documents

Depending on the agent's configuration, you may be able to ask further questions about documents already shared in the conversation without uploading them again.

For example, after uploading the Project Orion proposal, the user asks:

From shared files, tell me when the Orion Project is targeting completion.

The assistant refers to the previously shared document and retrieves the relevant information.

In this example, it correctly identifies the proposed completion date of 30 November 2026 and the related decision deadline of 26 October 2026.

This can be useful when reviewing proposals, meeting notes, requirements or other supporting documents over the course of a conversation.

Important: Shared documents are separate from information retrieved through Softools MCP. Sharing a document in Teams does not add it to a Softools App or make it available to other Softools users.

The ability to retrieve a previously shared document may depend on the agent's configuration, the current conversation and the permissions available. It should not be assumed that the assistant can access the same document in every future conversation.

 

 

Saved Memories

Some Copilot Studio agents may also provide a Saved memories feature.

This allows the assistant to retain selected information about your preferences or working context for use in future conversations.

For example, an assistant might remember how you prefer reports to be presented or which types of information you regularly request.

Where this feature is available, you can use Manage memories to review and manage the information the assistant has retained.

In the example shown, the agent indicates that saved memories are deleted after 28 days of inactivity.

The availability and retention behaviour of memory may vary depending on the agent and its configuration.

Important: Removing a saved memory should not be treated as deleting the original Teams conversation, uploaded documents or information retained in other systems. These may have separate retention and deletion controls.

Saved memories are a feature of the AI assistant, not of the Softools MCP connection.

 

 

Managing Your Connection

You can review your Softools connection through the Copilot Studio connection management interface.

When your account has been successfully connected, the Softools MCP connection should display a Connected status.

If you need to reconnect, select the relevant connection and follow the available management options.

Your organisation's administrators can also revoke approved MCP connections in Softools.

Revoking a connection prevents further use of that authorisation, but does not necessarily remove information already retrieved by the AI assistant.

For more information, see Security Considerations When Connecting AI Assistants to Softools.

 

 

Troubleshooting

I cannot find the AI assistant in Teams

Your organisation must make the Copilot Studio agent available to you.

Contact your Teams or Copilot Studio administrator if you cannot locate it.

 

The assistant asks me to set up a connection

Select Set up connection and follow the Softools authentication process described above.

Each user may need to authorise their own account.

 

My connection shows Not connected

Open the connection management window, select Connect and complete the Softools sign-in and authorisation process.

 

The connection is successful but my original question has not been answered

Return to the Teams conversation and select Retry, or submit your question again.

 

The assistant cannot find an App or Record

The App may not have a published MCP-enabled View, the relevant Fields may not be included, or your Softools account may not have access to the information.

Contact your App Builder if you believe information should be available.

 

The assistant returns unexpected priorities or recommendations

Check the Field values and their intended meaning.

For example, confirm whether Level 1 or Level 5 represents the highest priority in your organisation.

Ask the assistant to explain which Records and Fields support its conclusions.

If a Field's meaning is unclear, ask your App Builder to review its description in the MCP-enabled View.

 

The assistant cannot analyse an uploaded document

Check that the document type is supported and that your organisation has enabled file uploads for the agent.

The ability to analyse files is separate from the Softools MCP connection.

 

A previously uploaded document cannot be found

The assistant may not have access to the file in the current conversation context.

Check the original Teams conversation and its Shared tab. If permitted, you may need to provide the document again.

 

Related Articles

Comments

0 comments

Article is closed for comments.