Your company has five people who get AI. Make them the help desk.
Buoy for teams puts a colleague one level ahead on someone's screen in about a minute, keeps every session inside your walls, and writes down what the team learned. The goal is people getting better, not tickets getting closed.
Built for environments that answer to HIPAA, HITRUST, SOC 2 and ISO 27001. Where each attestation stands, in plain words, before you quote us to procurement.
Who it's for, in their own words
Different people want different things from the same ten-minute session.
IT
A help desk that never lands on your queue, with the controls, logs and paperwork an audit asks for. The list.
HR and L&D
A mentorship program that runs itself. Levels show who is growing, the log shows what the company is learning, nobody schedules a lunch-and-learn.
Team leads
See which question comes up every week, turn the best answer into a bot, and stop explaining it in standup.
The person who's stuck
Someone who says "it's just your build folder, hit y," instead of fourteen considerations.
The Buoy log: a team's memory of who helped whom.
Every session can leave a trace the whole team learns from. The person who asked decides whether their summary goes in, or switches on auto-share. Teammates add a note and give kudos.
- Who helped, who asked, outcomeevery entry
- The summaryonly if the asker shares it
- Who's carrying the teamin the stats
- What people get stuck onby tag, this month
- Where it livesencrypted per team
Your own topical helper bots
A bot that knows your way of doing things, built on the same API a human helper uses. It can talk, see the shared window (already masked), and point at things.
Trained on your policy
Feed it the handbook, the runbooks, the decision log. It answers the way your best people would, and says when it does not know.
Learns from what people do
Sessions become summaries. The bot team promotes the answers that worked and retires the ones that did not. Lessons flow from the people who figured it out to everyone else.
Every lesson audited
Who asked, who answered (person or bot), what was shown, what was said, who looked afterwards. Once a bot gives the approved answer, everyone gets the same one; change it in one place.
You host the bot, or we do. Keys stay with you. Capacity per bot is a setting, so one bot can help twenty people at once. Start at the developer portal.
Security and privacy, the way IT wants to read it
What the helper sees, what we keep, and who can get at it. Everything below is shipped unless marked planned.
| Area | What you get |
|---|---|
| Identity | LinkedIn-verified identities with a code-challenge sign-in today; SSO (SAML/OIDC) and SCIM planned. Second factor (TOTP) for privileged roles. Access tokens expire hourly, refresh tokens rotate and self-revoke if copied, API tokens expire by default. |
| What leaves the Mac | Only the window or screen the person picks, never automatically. On-device masking covers emails, phone and card numbers, SSNs, IBANs, API keys, tokens and your own words before a frame is encoded. Password managers, mail, chat and settings apps are excluded by default; team policy extends the list. |
| Recording | Every session is recorded and both people see it on screen the whole time. Transcription happens on the seeker's Mac. Recordings, transcripts and summaries belong to the person who asked; the helper sees the summary only; a seeker can delete a session. Designated staff can review a session for quality or abuse, and each review is in the audit trail. |
| Encryption | TLS in transit, SRTP/DTLS for media. At rest, each user has their own data key: recordings, transcripts, summaries, screenshots, profile data and MFA secrets are AES-256-GCM encrypted with it, and the key is destroyed when the account is deleted. |
| Audit trail | Sign-ins, tokens, bots, every read, upload and download of a record, exports, deletions, admin actions: actor, IP, user agent, time. People see their own trail; staff access needs MFA and is itself logged. |
| Team policy | Admins set forced masking, never-share apps, masked words, people-only (no bots), shorter retention and auto-share. Members' apps pick it up within seconds and lock the matching settings. Stricter wins when someone is on several teams. |
| Retention and deletion | Defaults: recordings 90 days, transcripts and summaries 1 year, sign-in history 1 year, audit 400 days; shorter per team. One-call export and account deletion with crypto-shredding. |
| Infrastructure | API, database and cache bound to loopback behind nginx; services run as an unprivileged user in a systemd sandbox. Per-IP rate limits everywhere. HSTS, CSP, frame denial and referrer policy on every page. Key-only SSH, unattended security patches. |
| Deployment | Shared cloud with team isolation today. Dedicated tenant and self-hosted planned. AI summaries on your own key or model gateway planned. |
| Supply chain | Locked dependencies, dependency audit and static analysis on every change, secret scanning, signed and notarized Mac builds, signed auto-updates. security.txt with a monitored contact; third-party penetration test planned before GA. |
Compliance status, honestly
Attestations take auditors and time; controls can ship today. So nobody is surprised in a security review:
| Framework | Status | What it means for you |
|---|---|---|
| SOC 2 Type II | audit window planned | We can complete your questionnaire against the Trust Services Criteria today and share the report when issued. |
| HIPAA | safeguards in place BAA on Enterprise | Masking, encryption and audit cover PHI that appears on a shared screen; the BAA sets the rest of the terms. |
| HITRUST CSF | e1 planned | The e1 control set maps to the controls above; we publish the certification when issued. |
| ISO 27001 | ISMS in progress | Policies, risk register and statement of applicability available under NDA. |
| FedRAMP | on request | Needs an agency sponsor and a dedicated US-government cloud deployment; we scope it with you. |
How a rollout goes
Small, then wider. The log is the artifact you keep either way.
1Week one
A pilot team of 20 to 50. Sign in, admin sets masking policy, never-share apps and retention. First helpers designated, first bot optional.
2Weeks two to four
Sessions happen. A weekly digest to the sponsor: questions asked, time to a helper, levels moving, top ten topics.
3Month two and on
First internal bots from the top ten topics. Quarterly: a learning report for HR, access review and audit export for IT, policy refresh for the bot team.
Questions procurement asks
Where is data stored?
In our hosted environment today; dedicated regions and self-hosting are on the roadmap. Subprocessor list on request.
Can we turn recording off?
Recording is how quality and abuse get reviewed, so it stays on; team policy sets shorter retention and who may review.
Do you train AI models on our sessions?
No. Summaries are generated per session and are yours. Nothing is used to train models, ours or anyone else's.
What do helpers see?
Only the masked window or screen the seeker chose, for the duration of the session, and the summary afterwards. Never files, never other windows.
Can we bring our own bots and keys?
Yes. Bots run on your infrastructure with credentials you can rotate or revoke at any time.
DPA, BAA, security questionnaire?
All available. Send yours or use ours.
Talk to us.
Tell us your team size and the frameworks you answer to. We reply with the trust packet and a pilot plan.
enterprise@buoy.be-n.comPilots run two weeks, one team, and you keep the Buoy log.