Everything runs on your accounts.

Data

What we store and where.

• Your operational data lives in your systems: your CRM, your spreadsheets, your database.
• What Trascend stores is the minimum needed for the flow to run and be auditable: what ran, when, with what result.
• When a process has to retain content, we define up front how long and when it gets deleted.
• We don't use one client's data for another client, or to train models.

Access

Least privilege.

• We ask only for the permissions the process needs, not admin access "just in case".
• Every access is logged: who requested it, what for, and until when.
• When a piece of work ends, access that is no longer needed is revoked.
• Trascend's own administrative access uses two-factor authentication.

Credentials

Secure key handling.

• Credentials don't travel over chat or email: they're loaded into the system that will use them.
• They're never stored in code or in the frontend. They never reach anyone's browser.
• They are rotated when the team changes, when scope changes, or when you ask.
• Where the vendor allows it, we use OAuth instead of a copied key.

Audit

A record of what happened.

• Every flow execution is logged: when it ran, with what input, with what result.
• Errors are stored with their context, so what happened can be reconstructed.
• In the client portal you can see which automations are active and what was approved.

Continuity

What happens if the relationship ends.

• Because everything runs on your accounts, if Trascend stops working with you nothing switches off.
• We hand over documentation of what was built: what it does, how to configure it and how to disable it.
• You revoke our access whenever you want, without needing us to do anything.
• We don't hold your data hostage to hold on to you.

Support hours, what happens when something fails and if the technical lead is away: continuity and support.

AI

What data reaches a model.

• Only what the process needs to answer reaches the model. We don't send it your database.
• The model providers we use don't train on the content of commercial accounts.
• Where automated classification exists, rules define what it can and cannot decide on its own, and when a human steps in.
• A model can be wrong. Processes where that is unacceptable don't get solved with AI.