blankit
Legal · Employee chatbot

What the chatbot does with your questions.

Plan members can ask about their group benefits coverage through a web chatbot their advisor links them to, or — where their employer has installed it — through a Slack or Microsoft Teams bot inside their own workplace. This is what happens to what they type, in each place.

On the web

The web chatbot is anonymous. It asks for no name, no email, and no sign-in — the link itself is the only key, and we do not know who is holding it.

  • Your questions are not saved. Each message is sent to an AI service to work out the answer, and nothing you type is kept afterwards — not by blankit (the software platform this chatbot runs on), not by your employer, and not by your benefits advisor.
  • We keep a record that a conversation happened — when it was, where you asked it, which plan booklet it was about, and a rough subject label like “dental” or “prescriptions” — so your advisor can see which topics come up. Never your name, and never what you typed.
  • Answers are generated by AI — Anthropic’s Claude, through Amazon Bedrock — from your plan booklet and can be wrong. Your question may be processed outside Canada, including in the United States; your booklet stays stored in Canada. Your official plan documents and your insurance carrier are the final word.
  • This is not medical or financial advice. The chatbot cannot see your claims, your medical history, or anyone else’s.
  • We set one cookie that lasts until you close your browser, so your advisor can see how often this chatbot is used. It holds a random ID — not your name, email, or IP address.
  • You can ask what information is held about you, or have it corrected or deleted. Contact your benefits advisor, who is responsible for this chatbot.

In Slack

What we retain does not change in Slack — still nothing a member types. What does change is who else can see it, and that is not up to us: the workspace belongs to the member's employer. We say so before their first question rather than leaving them to assume the two surfaces are equally private.

  • Your questions are not saved by us. Each message is sent to an AI service to work out the answer. Your last few messages are then held for about 30 minutes, so the chatbot can follow the conversation when you reply to it, and are deleted automatically after that — beyond those few minutes nothing you type is kept by blankit (the software platform this chatbot runs on) or by your benefits advisor, and none of it goes into your plan records. What you send in this chat app is a separate matter — see the next point.
  • You are using this chatbot inside your employer’s Slack workspace. Slack — not blankit, and not your benefits advisor — controls what is visible there, and on some Slack plans an administrator can export direct messages with apps. To ask privately instead, use the web link your advisor gave you.
  • We keep a record that a conversation happened — when it was, where you asked it, which plan booklet it was about, and a rough subject label like “dental” or “prescriptions” — so your advisor can see which topics come up. Never your name, and never what you typed.
  • So your advisor can see how many different people use the chatbot, we count each person’s questions as a monthly total under a scrambled code — never your name, and kept apart from the subjects above, so we hold no record of what any one person asked about.
  • Answers are generated by AI — Anthropic’s Claude, through Amazon Bedrock — from your plan booklet and can be wrong. Your question may be processed outside Canada, including in the United States; your booklet stays stored in Canada. Your official plan documents and your insurance carrier are the final word.
  • This is not medical or financial advice. The chatbot cannot see your claims, your medical history, or anyone else’s.
  • You can ask what information is held about you, or have it corrected or deleted. Contact your benefits advisor, who is responsible for this chatbot.

In Microsoft Teams

The same as Slack: nothing a member types is retained on our side, and the organisation the bot is installed in belongs to the member's employer, whose own Microsoft 365 controls govern what they can see. The notice is worded for Teams and shown before the first answer.

  • Your questions are not saved by us. Each message is sent to an AI service to work out the answer. Your last few messages are then held for about 30 minutes, so the chatbot can follow the conversation when you reply to it, and are deleted automatically after that — beyond those few minutes nothing you type is kept by blankit (the software platform this chatbot runs on) or by your benefits advisor, and none of it goes into your plan records. What you send in this chat app is a separate matter — see the next point.
  • You are using this chatbot inside your employer’s Microsoft Teams. Microsoft 365 — not blankit, and not your benefits advisor — controls what is visible there, and your organisation’s administrators may be able to review conversations with apps. To ask privately instead, use the web link your advisor gave you.
  • We keep a record that a conversation happened — when it was, where you asked it, which plan booklet it was about, and a rough subject label like “dental” or “prescriptions” — so your advisor can see which topics come up. Never your name, and never what you typed.
  • So your advisor can see how many different people use the chatbot, we count each person’s questions as a monthly total under a scrambled code — never your name, and kept apart from the subjects above, so we hold no record of what any one person asked about.
  • Answers are generated by AI — Anthropic’s Claude, through Amazon Bedrock — from your plan booklet and can be wrong. Your question may be processed outside Canada, including in the United States; your booklet stays stored in Canada. Your official plan documents and your insurance carrier are the final word.
  • This is not medical or financial advice. The chatbot cannot see your claims, your medical history, or anyone else’s.
  • You can ask what information is held about you, or have it corrected or deleted. Contact your benefits advisor, who is responsible for this chatbot.

On WhatsApp

Still nothing a member types is retained on our side. Unlike Slack and Teams, the WhatsApp account is the member's own rather than their employer's, so their employer cannot see these conversations. Two things are different here instead, and both are in the notice shown before the first answer: WhatsApp is run by Meta, outside Canada, so the message and the answer travel through Meta's servers — while the plan booklet stays stored in Canada, as on every other surface, and the AI that answers (Anthropic's Claude, through Amazon Bedrock) may process a question outside Canada, including in the United States — and because a message arrives from a phone number rather than a workspace account, we keep a scrambled code derived from that number so a returning member reaches their own plan. We do not keep the number, so nothing we hold can be used to contact anybody.

  • Your questions are not saved by us. Each message is sent to an AI service to work out the answer. Your last few messages are then held for about 30 minutes, so the chatbot can follow the conversation when you reply to it, and are deleted automatically after that — beyond those few minutes nothing you type is kept by blankit (the software platform this chatbot runs on) or by your benefits advisor, and none of it goes into your plan records. What you send in this chat app is a separate matter — see the next point.
  • You are using this chatbot on WhatsApp. WhatsApp is run by Meta, outside Canada, so your message and the answer travel through Meta’s servers — under an arrangement blankit has with them, not one your employer has. Your plan booklet stays stored in Canada, exactly as on every other version of this chatbot. The AI that reads it to answer you is Anthropic’s Claude, through Amazon Bedrock, which may process your question outside Canada, including in the United States.
  • So a returning message reaches your own plan, we keep a scrambled code made from your mobile number — never the number itself, so nothing we keep can be used to phone or message you. Send STOP at any time and this chatbot stops replying; send START to turn it back on.
  • We keep a record that a conversation happened — when it was, where you asked it, which plan booklet it was about, and a rough subject label like “dental” or “prescriptions” — so your advisor can see which topics come up. Never your name, and never what you typed.
  • So your advisor can see how many different people use the chatbot, we count each person’s questions as a monthly total under a scrambled code — never your name, and kept apart from the subjects above, so we hold no record of what any one person asked about.
  • Answers are generated by AI — Anthropic’s Claude, through Amazon Bedrock — from your plan booklet and can be wrong. Your question may be processed outside Canada, including in the United States; your booklet stays stored in Canada. Your official plan documents and your insurance carrier are the final word.
  • This is not medical or financial advice. The chatbot cannot see your claims, your medical history, or anyone else’s.
  • You can ask what information is held about you, or have it corrected or deleted. Contact your benefits advisor, who is responsible for this chatbot.

Who is responsible

The advisor firm that turned the chatbot on is the controller of plan-member data; blankit processes it on their documented instructions under our Data Processing Agreement. A plan member exercising a right — access, correction, deletion — goes to their firm, whose own copy of this notice and named privacy officer are on that firm's privacy page. The platform-level policy is at /legal/privacy.