Is it safe to give AI your client database?
It is a fair question and it deserves more than a paragraph. Here is what we hold, who can reach it, what it is never used for, and what happens to it when you leave.
Your database stays in your CRM, and your CRM stays the system of record. We hold a working copy while your account is open, and it is deleted or returned when it closes.
Nothing you send us trains anybody’s model, and one brokerage’s data is never reachable from another’s.
Eight questions, and where each answer is written down.
None of this is a marketing claim you have to take on trust. Every line is in a document we are bound by, or in a registry you can read.
| The question | What is true | Where it says so |
|---|---|---|
| Does my data train a model? | No. Data sent for composition is not used to train them. | Privacy policy |
| Who writes the messages? | Anthropic’s models compose the drafts, on our instruction | Named as a sub-processor |
| Can another customer reach it? | No. Access is scoped per brokerage | Privacy policy |
| Is it sold, or used for ads? | No, and not shared for cross-context advertising | Privacy policy |
| Do you hold a copy? | Yes, a working copy, while your account is open | Retention section |
| What happens when I leave? | Deleted or returned on request | Retention section |
| Where do my CRM credentials live? | Encrypted with a key that is itself split across separate stores | Security section |
| Is there an audit trail? | Every decision is logged, including the rule that blocked a message | The guardrail registry |
The privacy policy is the binding document and it is written to be read, not to be survived. If a line on this page and a line in that policy ever disagree, the policy wins and we have a bug to fix.
A working copy, and nothing you did not ask for.
Reading a lead’s whole history means holding a copy of it. Any vendor claiming otherwise is describing a different product.
Contact and conversation data for the leads you point us at. When your account closes it is deleted or returned on request.
The brokerage keeps its own copy in its CRM whatever happens to us. That copy is theirs, not ours, and we never ask you to move off the system you already run.
The one thing we never delete. It is the only way to guarantee nobody is contacted again after they said stop.
Credentials for connected systems are encrypted with a key that is itself split across separate stores.
Everything, not just the credentials.
One customer’s data is not reachable from another customer’s session. On mobile, sessions live in the iOS Keychain and can be locked behind Face ID or Touch ID.
With the rules that fired and the one that failed, written onto your own records.
Every message clears twenty-five guardrails before it sends. Sixteen of them are described in full, and the page says plainly which ones are not and why.
Messages go out from your office’s own number, in your agent’s name. It never signs as the lead, and it never denies what it is.
What we are not going to promise you.
We hold your data. There is no version of this where we do not. Composing a message out of a lead’s two years of history means reading that history, and reading it means holding it. The questions that matter are how long, who else can reach it, and what it is never used for. Those are answered above.
A third party writes the drafts. Anthropic’s models do the composing. They are named in our privacy policy as a sub-processor rather than buried, and data sent for composition is not used to train them. If using any model provider at all is a hard no for your brokerage, this is the wrong product and we would rather you knew now.
No system is perfectly secure, and our policy says so. If a breach affects your information we will notify you and, where required, the relevant authority. A vendor who will not write that sentence down has not thought about it.
We cannot make your CRM vendor’s security our own. We run inside the system you already chose. What happens to your data inside that system is between you and them, and it was before we arrived.
AI and your database, answered.
The binding version is the privacy policy. What the engine will and will not say to a lead is on the security page.
Is it safe to give AI access to my CRM?
Does Substrait train AI on my client data?
Can another brokerage see my leads?
Do you store my contacts, or only read them?
What happens to my data if I cancel?
How are my CRM credentials stored?
Do my leads know they are talking to software?
Read the policy first. Then let it text you.
One flat monthly rate by database size. No setup fee, no per-message billing, no per-seat billing.