Yes. Firstweek reads Slack, Salesforce and your other tools through read-only connections, summarizes what it reads at ingest, and never stores raw messages or documents. It never writes to, changes or deletes anything in a connected tool. A hiring manager approves every artifact before the new hire sees it. Firstweek holds no third-party security certification and does not sign BAAs.

The rest of this post is the detail a security reviewer will ask for. Each answer matches the security page.

Can an AI onboarding tool be read-only?

Yes. Every Firstweek connection uses the provider's OAuth and is read-only: Slack, Confluence, Jira, Google Drive, OneDrive and SharePoint, Microsoft Teams, GitHub, HubSpot and Salesforce.

Where a provider offers read-only permissions, Firstweek requests only those. Two providers don't offer one: GitHub (for private repositories) and Salesforce. Their consent screens ask for broader access, and Firstweek's code never writes. In Salesforce you can enforce that yourself by connecting a Salesforce user that has a read-only profile.

Your admin chooses which channels, spaces, repositories, folders and records are in scope. Nothing outside that scope is read. The connections page lists what is read from each tool.

Does Firstweek store our Slack messages?

No. We read it, summarize it, and forget it. The raw messages never touch our database.

Raw messages and documents are processed in memory and never written to the database. Three things persist:

A citation can show a short excerpt of up to 200 characters. The excerpt is cut from Firstweek's summary of the source, not from the raw message. It is on by default, and a workspace admin can turn it off.

Can it change or delete anything in our tools?

No. Firstweek never writes to, changes or deletes anything in a connected tool. Any connection can be disconnected at any time from settings.

What happens to personal data?

You can disconnect any source, or delete any onboarding and its derived data, from the app. To delete a whole workspace, email legal@firstweek.app. It is deleted within 30 days of the request, and backups are purged within a further 30 days. A Data Processing Agreement is available on request.

Who sees the output before the new hire does?

The hiring manager. Nothing reaches a new hire until their hiring manager approves it. The manager can correct any statement before approving, and corrections are kept with the artifact.

Access inside the workspace is role-based: admins configure the workspace and connections, managers create and approve artifacts, and new hires see only their own. Sign-in is through Microsoft Entra ID or Google, and a workspace can require SSO for everyone. Sign-ins, connections, scope changes, approvals, deliveries and staff access are recorded in an audit log that workspace admins can review.

Which AI providers handle the data, and do they train on it?

Anthropic writes the artifact, with OpenAI as the fallback. OpenAI also narrates approved artifacts as audio. Questions a new hire asks about their own artifact go to the same providers.

Neither provider trains on data sent through its API, and Firstweek does not use customer data to train models. Every provider is listed on the subprocessors page before it processes customer data.

What does Firstweek not have?

Security reviews run on your own questionnaire. Send it to security@firstweek.app and it gets answered. For financial services, the 30-day pilot starts when your review clears, not before.

Ten questions to give your security reviewer

These apply to any AI tool that reads internal systems. Firstweek's answers are above.

  1. Is every connection read-only, and which permissions does each consent screen request?
  2. Where the provider has no read-only permission, what stops the tool from writing?
  3. Who chooses what is in scope, and can anything outside that scope be read?
  4. Are raw messages or documents ever written to storage?
  5. What exactly persists, and can citations expose source text?
  6. What is encrypted, at which layer, and with what?
  7. How is one customer's data kept separate from another's?
  8. Which AI providers receive content, and do they train on it?
  9. Who approves the output before anyone else sees it, and is that recorded in an audit log?
  10. Which certifications and agreements does the vendor not have?

FAQ

Is Firstweek safe to use at a financial services or healthcare company?

That is for your review to decide, and this is what it will find: read-only connections, no raw message storage, SSO, role-based access and an audit log. It will also find no third-party certification, no BAA and no EHR connections.

Can we limit what Firstweek reads?

Yes. Your admin selects the channels, spaces, repositories, folders and records in scope, and nothing outside that scope is read. Any connection can be disconnected at any time.

How do we start a security review?

Send your questionnaire to security@firstweek.app. For financial services, the pilot's 30 days start when the review clears.

Run it on your next hire. 30 days. Then we talk. Start a pilot or see a sample artifact. How it works for scaling teams and financial services.