Trust & transparency

Microsoft data access

What Fronterio reads, writes, and deliberately never touches in your Microsoft tenant. One page for every integration. No marketing fluff. Suitable for security questionnaires.

Last updated against the live source code on the dev branch.

Request the full technical appendix

Every integration is opt-in and independently revocable

Every Microsoft integration Fronterio offers is opt-in. Nothing Microsoft-related happens until your tenant admin grants explicit consent, or you sign in with Microsoft. Every connection can be revoked from your tenant at any time. Fronterio deletes the related data on disconnect.

At a glance

IntegrationWhat it's forWho consentsTenant scope
Microsoft sign-inLogging in with MicrosoftThe signing-in userYour account only
Copilot adoption trackerCopilot usage by department, and by person for leadersTenant admin (delegated), in up to three stagesDirectory roster (required); Copilot usage reports and licence counts (each optional)
SharePoint intranet syncMirror governance data into your intranetTenant admin (app-only)One SharePoint site
Microsoft Teams agentFronterio bot inside TeamsAdmin sideloads, users installBot conversations only
Copilot Studio integrationDiscover & govern Copilot Studio agentsTenant admin (app-only)Power Platform environments
M365 Copilot ChatFronterio in Copilot ChatInherits Teams agentSame as Teams
Azure AI Foundry agent discoveryList the agents in an Azure AI Foundry projectCustomer Azure adminOne Foundry project, read-only
Microsoft Purview labelsBump EU AI Act risk on confidential dataTenant admin (app-only)Sensitivity-label metadata only
Free Copilot Chat activity trackerAdoption metrics for free/unlicensed Copilot ChatTenant admin (ActivityFeed.Read app permission)Single-tenant per connection
1. Microsoft sign-in

The “Continue with Microsoft” button on the login screen.

What we read

  • Your email address and display name from your Microsoft account.
  • The Microsoft account ID Supabase Auth uses to keep your sign-in linked.
  • We always show the Microsoft account picker (prompt: select_account). We never silently reuse a session.

What we write

  • A row in our profiles table with your email and full name. That's it.

What we never access

  • No Microsoft Graph API calls.
  • No mailbox, calendar, files, contacts, or directory data.
  • Nothing tenant-wide.

Why we need it

So you can sign in without a separate password. Optional: email/password sign-in is also supported.

Where it's stored

profiles table on Fronterio's Supabase database (Frankfurt, EU). Row-level security ensures you can only read your own profile.

2. Microsoft 365 Copilot adoption tracker

Powers the dashboard that shows which departments are getting value from Copilot and, for leaders, which named people in their function have started.

What we read

  • Three Microsoft Graph permissions, each granted on its own consent screen. Only the first is required. Directory (User.Read.All): /v1.0/users returns your directory roster. For each person we take the object ID, display name, UPN, mail, department, job title and whether the account is enabled. We do not read the manager link. With this permission we also read /v1.0/organization once, for your tenant's display name.
  • With the directory permission we also read each person's department and assigned licences. We keep only the licence count per department, never the person.
  • Copilot usage (Reports.Read.All, optional): /v1.0/copilot/reports/getMicrosoft365CopilotUsageUserDetail(period='D30') returns a CSV with one row per Copilot-licensed user. Each row holds the UPN, the display name, and the last date Copilot was used, overall and in each app, over the last 30 days. The apps are Teams, Word, Excel, PowerPoint, Outlook, OneNote, Loop and Copilot Chat. The report holds no action counts and no department. The department comes from your directory roster.
  • Licences (Organization.Read.All, optional): /v1.0/subscribedSkus for purchased and assigned licence counts per SKU.

What we write

  • Nothing. This integration is read-only against Microsoft Graph.

What we never access

  • No file content. No message content. No calendar events. No chat transcripts.
  • No Copilot prompts and no Copilot responses.
  • The usage report holds only the date of last use per app. It never says what was done.
  • The directory reads take only the fields named above. No manager link, no phone numbers, no addresses, no photos, no group memberships.

Why we need it

To show you “70% of Marketing used Copilot in Word in the last week; Engineering hasn't started”. It also lets a leader see which named people in their function have and have not started. You do not have to build the dashboard yourself.

Where it's stored

Everything is stored on the Fronterio side in Frankfurt (EU). Usage rows live in ai_tool_usage_data under a hash, never the UPN itself. The hash is a SHA-256 of your organisation's ID and the UPN, prefixed ms_. The directory roster lives in directory_users in plaintext: name, UPN, mail, department and job title, one row per directory user, so that a leader can see which people in a function use Copilot and so the Frontier Twin can link people to functions. The roster is joined to usage by recomputing the same hash. Licence counts live in licence_entitlements and licence_assignments_by_department. Neither holds a person. It is readable only by your organisation's owners, admins and AI leads, and by a department lead for their own function; it is never readable by a partner and never sent to an AI provider. Every named view is written to your data_access_log. Rows are deleted when you disconnect the integration, and a person who leaves your directory is purged 30 days after they disappear from it. Refresh tokens are stored encrypted at rest. Disconnecting via Settings or revoking admin consent invalidates the refresh token. Which permissions you granted is stored beside the connection, so the dashboard can show what is and is not connected. Removing a permission is done by disconnecting, or by revoking the app in Microsoft Entra. Microsoft offers no per-permission revoke.

3. SharePoint intranet sync

Mirrors Fronterio governance data into a SharePoint site so your existing intranet stays the source of truth.

What we read

  • The site you point us at, to verify it exists and to discover whether the lists and libraries we need already exist.
  • That's the entire read surface.

What we write

  • A fixed, known set of artefacts that Fronterio creates and owns:
  • Lists: “AI Use Cases”, “AI Agents”, “Compliance Status”, “Guide Completions”.
  • Document libraries: “AI Reports”, “AI Playbooks”.
  • Items inside those lists and libraries. These are rows for use cases, agents, EU AI Act obligations and guide completions, PDFs of assessment reports, HTML playbooks and weekly audit-digest PDFs.
  • One tracker item for an approved Transformation step. A person in your organisation approves the change first. Fronterio then creates or updates one item on the “AI Use Cases” list, with only the fields that step names. It reads the item back to confirm the change. Undoing it moves the item to the SharePoint recycle bin.
  • A FronterioKey column on the “AI Use Cases” list. It lets a retried write find its earlier item instead of creating a second one.

What we never access

  • No existing SharePoint files.
  • No site analytics. No SharePoint search. No user permissions.
  • We do not enumerate, read, or modify any list, library, or file we did not create.
  • The granted permission (Sites.ReadWrite.All) is broader than what we use (Microsoft does not offer a tighter scope today), and Fronterio's code only ever calls a fixed set of endpoints.

Why we need it

Customers want their intranet to be the canonical record of “what AI we use”. This is the clean way to keep SharePoint and Fronterio in sync.

Where it's stored

Connection metadata (tenant ID, site URL, site ID, encrypted client credentials, sync configuration) lives in intranet_connections. Every sync write is logged in intranet_sync_log so you can audit what we sent and when.

4. Microsoft Teams agent

The Fronterio bot inside Teams.

What we read

  • Only messages addressed to the Fronterio bot: direct messages, @Fronterio mentions in a channel, or replies in a Fronterio-initiated thread.
  • For each such message: text, sender Azure AD object ID, sender User Principal Name, tenant ID, channel ID, conversation ID.
  • Every inbound activity is JWT-validated against Microsoft's Bot Framework public keys before we even parse it.
  • The personal tabs (My work, Learn, Prompts, Human Intelligence) sign you in silently: Teams issues a short-lived token for the Fronterio app, we verify it against Microsoft's keys and read only your object ID, tenant ID and sign-in address from it. The token is exchanged once for a Fronterio session, never stored and never logged. The tabs make no Microsoft Graph call.

What we write

  • Replies and proactive notifications (text or Adaptive Cards) back to the same conversation.

What we never access

  • No messages that aren't directed at the bot.
  • No chat history. No team membership. No files in Teams. No user presence. No calendar.
  • We never crawl a channel; we only see what is sent to us. The Teams platform itself enforces this.

Why we need it

So employees can ask Fronterio questions without leaving Teams, and so Fronterio can deliver tip-of-the-day, prompts, and learning content where people work.

Where it's stored

teams_user_mappings caches the Azure AD object ID ↔ Fronterio user link. Only owners, admins, and AI Leads in your org can read it (RLS-enforced). teams_conversation_references stores the minimum needed to send proactive replies (service URL, conversation ID, channel ID). We do not store message bodies.

5. Microsoft 365 Copilot Studio integration

Discovers Copilot Studio agents in your tenant so they appear in Fronterio's governance registry.

What we read

  • Power Platform BAP API: to enumerate Power Platform environments your service principal can see.
  • Copilot Studio REST API: to list bots in those environments and read metadata (display name, description, environment ID, solution ID, publish status, last published date).
  • Microsoft Graph (Teams app catalog): only the /v1.0/appCatalogs/teamsApps endpoint, filtered to apps distributed inside your organisation.
  • Every four hours, we re-read the state of each discovered agent. A change made in Copilot Studio is then recorded in your Fronterio audit log.

What we write

  • Nothing. This integration is read-only against Copilot Studio.
  • Fronterio does not create, change, publish or delete Copilot Studio agents.

What we never access

  • No conversation transcripts. Microsoft exposes Copilot Studio transcripts via a Dataverse table; Fronterio's code does not query that surface.
  • No Power Platform data. No Dataverse rows beyond what's needed to identify a bot. No flows, no canvas apps.

Why we need it

“What Copilot Studio agents do we have, and which are governed?” is the most-asked governance question. This integration answers it without anyone manually maintaining a spreadsheet.

Where it's stored

Discovered bot metadata in copilot_studio_agents. State changes in copilot_studio_lifecycle_events and your audit log. Fronterio does not read Copilot Studio usage today. The copilot_studio_telemetry table holds one placeholder row per governed agent, with every count at zero. All RLS-scoped to your organisation.

6. Microsoft 365 Copilot Chat: custom engine agent

Fronterio appears inside Microsoft 365 Copilot Chat as a custom engine agent.

What we read

  • This rides on the Teams agent (§4): same Azure Bot registration, same scopes, same data flow.

What we write

  • Same as the Teams agent: replies and proactive messages.

What we never access

  • There is no separate Microsoft permission to grant. If you've reviewed the Teams agent above, you've reviewed this.

Why we need it

So Fronterio is reachable from the Microsoft 365 Copilot Chat surface as well as Teams, without doubling the bot footprint.

Where it's stored

Same backend as the Teams agent. Manifest version 1.21 declares copilotAgents.customEngineAgents pointing at the same bot ID.

7. Azure AI Foundry agent discovery

Lists the agents in one Azure AI Foundry project so they appear in Fronterio's governance registry.

What we read

  • A connection check. We list your subscriptions once to confirm the service principal works (GET https://management.azure.com/subscriptions).
  • The agent list of the one Foundry project you name, through that project's /assistants endpoint. A scan runs only when an owner, admin or AI lead starts it.
  • For each agent: its ID, name, model and description. Nothing else.

What we write

  • Nothing. This integration is read-only against Azure.
  • Fronterio does not create, change or delete Azure resources.

What we never access

  • No agent conversations, threads, files or run content.
  • No Azure RBAC policies.
  • No resource groups, networking or virtual networks.
  • No subscription-wide reads beyond the one connection check.
  • No Microsoft Entra ID directory data.

Why we need it

So agents built in Azure AI Foundry appear in your governance registry beside every other agent. Nobody has to keep a spreadsheet of them.

Where it's stored

Encrypted credentials in deployment_credentials. Discovered agents in external_estate_agents, readable only by members of your organisation (RLS-enforced). Each scan is recorded in your audit log.

8. Microsoft Purview sensitivity-label read

Reads sensitivity-label metadata from your M365 tenant so the EU AI Act risk classifier can elevate tiers when confidential data is in scope.

What we read

  • Microsoft Graph endpoint /v1.0/informationProtection/policy/labels: the list of sensitivity labels configured on your tenant.
  • Each label's id, name, and parsed confidentiality level (Public / Internal / Confidential / Highly Confidential / Restricted).
  • Per-document label assignments only when the agent register wizard or FRIA generator runs and the agent's data scope explicitly references SharePoint / OneDrive / Outlook surfaces.

What we write

  • Nothing. This integration is read-only against Microsoft Graph.

What we never access

  • No document, email, or message content. Ever.
  • No DLP policy rules, conditions, or exception details.
  • No audit log records. No retention policy details.
  • No user identity beyond what Graph already returns on label-assignment objects.

Why we need it

When a customer labels documents Confidential, Highly Confidential, or Restricted, an AI system processing those documents should be classified at a higher EU AI Act risk tier automatically. Reusing the customer's existing Purview taxonomy means we don't ask compliance teams to re-author their data sensitivity scheme inside Fronterio.

Where it's stored

Tenant ID and connection metadata in purview_connections (service-role write only). Sensitivity labels themselves are NOT cached. Every wizard run / FRIA generation calls Microsoft Graph live.

9. Free Copilot Chat activity tracker

Captures free / unlicensed Microsoft 365 Copilot Chat usage that the licensed Copilot report (§2) can't see.

What we read

  • Office 365 Management Activity API CopilotInteraction audit events (RecordType 261) from the Audit.General stream, over a trailing 2-day window.
  • From each event: the User Principal Name, the interaction surface (AppHost — e.g. Word / Teams / the free BizChat), and the event type. Nothing else.
  • App-only client_credentials auth with the ActivityFeed.Read application permission — a separate admin consent from the Graph apps.

What we write

  • Nothing into your tenant. We create an Audit.General subscription (a metadata pointer to poll available content); no data flows into your tenant.

What we never access

  • No Copilot prompts or responses. Ever.
  • No message or file content. No AccessedResources file references.
  • Only the interaction's user, surface, and count.

Why we need it

So 'is the free-tier majority of the workforce actually using AI?' gets a real answer instead of a blank the licensed report can't fill.

Where it's stored

ai_tool_usage_data with tool_name copilot_chat. The UPN is never stored in plaintext. It is hashed before it touches the database, as a SHA-256 of your organisation's ID and the UPN, prefixed mschat_. The tenant ID is stored encrypted at rest. Disconnecting (or revoking admin consent tenant-side) stops the daily sync on its next run.

Data residency & sub-processors

Microsoft services are your own tenant in every case above. They are not Fronterio sub-processors. Fronterio does not control the Microsoft data plane. It cannot transfer data out of your tenant beyond the scopes you explicitly granted. Fronterio's own infrastructure runs on Supabase (Frankfurt, EU) and Vercel (EU regions). AI features use an AI model provider in the US, under SCCs and the EU-US DPF. The sub-processors list names it. No customer Microsoft data is ever sent to that provider.

View the full sub-processors list

Your controls

  • Disconnect any integration

    Settings → Integrations on the Fronterio side, or revoke admin consent in Microsoft Entra at any time. Either action stops the next sync and clears the relevant credentials.

  • Right to access (DSAR)

    Export everything Fronterio knows about you as a JSON bundle from Settings → Privacy.

  • Right to deletion

    Delete your account or your entire organisation from Settings. Audit log entries are written before deletion so the record survives.

  • Audit log

    Every PII read against Microsoft-derived data is logged for Article 32 evidence.

  • DPA

    Data Processing Agreement available at /dpa with SCCs, EU-US DPF, and zero data retention addendum on AI calls.

Questions or a security review request?

Email hello@fronterio.com. Security reviewers can request a live walk-through of the source files referenced in the technical appendix.

hello@fronterio.com
Microsoft data access: what Fronterio reads, writes, and never touches | Fronterio