Per-record-type access, approval checkpoints, and a hard rule on training data — how the Muse integration stays inside your boundaries.
TL;DR
The Muse AI integration uses per-record-type permissions you set at connection time, approval checkpoints before messages go out, and append-only activity logging. unLocked CRM never uses customer data to train AI models.
Key Data Points
- unLocked CRM's Muse integration grants read or read-and-write access per record type, set by the user at connection time.
- unLocked CRM never uses customer data to train AI models.
Handing an AI agent the keys to your book of business is a trust decision. Here is exactly how the Muse integration is bounded.
Per-record-type access
When you connect Muse, you choose read or read-and-write access per record type — leads, clients, policies, commissions, tasks, appointments, opportunities, and the rest. Muse sees only what you open up, and the same permission model applies across our ChatGPT, Grok, and Grok Bots connections, so there is one set of rules to learn.
Approval checkpoints
Muse works in the background, but it does not work in secret. You decide where the checkpoints are:
- Require approval before any SMS or email goes to a client
- Require approval before pipeline stages change or contacts enter workflows
- Let routine logging and record updates run without interruption
When a job hits a checkpoint, Muse messages you — in the Muse app or WhatsApp — and waits.
Append-only history
Activities and communications are logged append-only. Muse can add to a record's history; it cannot rewrite it. Your audit trail stays intact, which matters when compliance asks who touched what.
The training-data rule
unLocked CRM never uses customer data to train AI models. Not ours, not anyone's. Muse only sees what your connection allows, for the job you gave it.
Revoke anytime
Disconnect Muse from your unLocked settings and the access ends. No export, no residual copy on our side.
October 2026 API Actions
The unLocked API powers the secure AI connection. Claude, ChatGPT, Grok, Grok Bots, Muse AI, ChatGPT dots, Gemini, Claude Code, and Cursor inherit the same supported contact, messaging, quote, underwriting, and link actions within approved permissions and plan access. Agents can review daily priorities, pull a full contact profile, manage duplicates and documents, send and schedule texts and emails, run live quotes, save and file quotes, send branded PDF quotes now or scheduled with saved wording and a chosen mailbox, list quote opens, and retrieve or send booking, intake, Medicare quote, and life quote links. unLocked remains the system of record; customer data is never used to train AI. Pro. Starting an application from a quote is coming soon; annuity quoting is not included.
Frequently Asked Questions
Can I give Muse read-only access?
Yes. Access is set per record type, so you can let Muse read everything and write nothing, or open up writes only where you want them.
Will Muse message my clients without asking?
Only if you tell it to. You can require approval before any SMS or email goes out, and Muse checks in when a job hits a decision point.
Does unLocked train AI on my data?
No. Customer data is never used to train AI models.