Trust

How Koltra handles your data and permissions.

Delivery, Intelligence and Improvement are live with pilot customers. This page sets out what they may do, what they record and where people take over: the rules all three are built and tested against.

Design requirements

Clear limits, at every stage.

Limit access

Approve actions. Keep exceptions with staff.

Access requirements

Limited access
Its own login, limited to the actions your business approved.
People decide
Exceptions and approvals stay with your staff.
Settings only narrow
Your settings can restrict it, never widen what Koltra approved.

Separate data

Keep customers separate. Set retention. No training on your data.

Data requirements

Only the data the request needs
Access is limited to the approved information needed for the request. Your business sets retention.
Customers kept apart
Each customer’s data, phone numbers, permissions and records are kept separate.
No training on your data
Calls, texts and emails are not used to train or fine-tune models.

Review changes

Log requests. Test and version changes so they can be rolled back.

Change requirements

Every action logged
Each system call, handoff and error, with the request that caused it.
Product updates can be rolled back
Tested and versioned, released by a named person, and able to be rolled back.

Deployment-specific findings stay with you. Deployment-specific findings stay with that deployment. Reusing them elsewhere requires an explicit, recorded decision.

The record

See what happened to each request.

Each request leaves a record of the actions, handoffs and failures, with timing and cost where measured, so staff can see what still needs attention.

Questions

What compliance teams ask.

Is our data used to train AI models?

No. Customer data, including calls, texts and emails, is not used to train or fine-tune models.

What data does a Koltra product access?

Only what the workflow needs, through an identity and permissions set for that workflow. Retention and deletion are set for each deployment.

What happens when the product cannot finish a request?

It is designed to stop and pass the conversation and actions taken to a person. The record distinguishes an accepted handoff from work still waiting, and shows where the request stopped and why.

Can we inspect what it did?

Yes. Each request has its own record of its outcome, the rules that applied, system actions and where it stopped. It includes timing and cost where measured.

Get started

Have a security or compliance question?

Tell us what you need to review: data flow, access or retention. Security questions and vulnerability reports can go to hello@koltra.ai.

We reply within one business day.