Log inExperience Cascand

Security and control

Cascand can decide what deserves attention. You decide what it is allowed to do.

Cascand keeps you in control. You choose what to connect, your data stays yours, access is enforced at the account boundary, and nothing is sent, booked or invited unless you've said it can.

Your systemsCRM · Mailbox · Calendar · LinkedIn
Cascand, on its ownResearch · Decide · Prepare · Verify
YouApprove anything that is sent, booked or invited
Then, and only thenIt is sent, booked or invited

Changes to your CRM records are proposed for your approval. If you connect a CRM, an approved send is also logged there as an activity.

Always available to you: Pause · Revoke · Disconnect · Export

In practice

Consequential software should be controlled. Cascand is.

  • Your data stays under your control

    Your connected systems remain yours. Cascand reads from them within the access you authorize. Changes to your CRM records are proposed for your approval, and if you connect a CRM, an approved send is also logged there as an activity.

  • Your account is isolated

    Each customer account is isolated from every other customer. Access is enforced at the account boundary, not only in the interface.

  • You decide what Cascand can do

    Permissions are explicit and per capability. Connect only what you want. Pause Cascand for a person, a team or a capability at any time. Disconnect a system and Cascand's access ends with it. Connecting a mailbox does not authorize everything in it: you name which addresses count as your correspondence, and revoking one takes effect on the next sync.

  • Cascand can prepare without you. It cannot send, book or invite without you.

    It can research, prepare and organize work on its own. Nothing is sent, booked or invited unless you've said it can. Before anything consequential runs, software, not a model, weighs how risky the action is and how hard it would be to undo, and requires your approval whenever either one is high. Anything it doesn't know about an action counts as the riskier case, never the safer one.

  • Cascand does not assume an action succeeded

    It labels each result as confirmed by the system that performed it, or declared by you, and tells you Uncertain when it cannot tell. Never a false Done.

  • Credentials are protected

    Connection credentials are encrypted at rest, can be rotated, and are never shown in the product or written to logs. Cascand does not hold your LinkedIn password.

  • Untrusted text stays text

    What arrives from email, the web or a spreadsheet is treated as information, never as instructions. It cannot make Cascand act.

  • Everything is on the record

    Every recommendation, approval and action is kept in a tamper-evident audit trail: even Cascand's own systems cannot quietly edit it. You can see what Cascand decided, what it did, and why.

  • You can leave

    Export the data in your account at any time. Ask for identifying details, names, emails, LinkedIn URLs, to be anonymized, and the record of what happened stays intact without the identity attached.

The record

What Cascand keeps about every decision.

Control is only real if you can check it afterward. Every decision Cascand makes is stored with what it saw, what it considered, which version of your model made the call, what you did, and what happened next.

It answers the questions a security review asks. Why did it recommend this? Was that a fact or a guess? Who approved it, and did it actually go out?

The record stays in your account, learning happens only from your own recorded outcomes, and nothing in it is shared with any other customer. Export it when you want, and if you leave, ask for identities to be anonymized while the record of what happened stays intact.

How to challenge any decision →

For your security review

The straight answers.

Where is data held
In Cascand’s own database, with account isolation enforced in the database itself, not only in the application.
Who can access it
People you invite, with roles you assign. Access checks fail closed.
Which outside services see it
A language model provider, for certain drafting and research tasks, and the managed database and sign-in service Cascand runs on. Your CRM, mailbox, calendar and LinkedIn connection are your own accounts, connected directly. The named subprocessor list is available to your security team on request.
Can Cascand act on its own
Three places, and nothing moves between them on its own. Cascand researches, drafts and organizes by itself. Anything sent, booked or invited waits for your approval. Anything it has no information about, or that touches a system you have not connected or a contact on hold, waits for a person.
What if something goes wrong
Pause everything, one team, one person or one capability. Right before anything runs, Cascand checks again that the action is still live, nothing is on hold and nothing is paused. A retry is never a second send.
Single sign-on and provisioning
Single sign-on and automatic user provisioning for enterprise accounts, set up together with your identity provider.
Security review
Your security team is welcome to review how the product works, how access is enforced and how data is handled. Ask us for the walkthrough.
Security questions: sales@cascand.io

Keep exploring

Where to go next.