Your Data Choices

What we do record

When an AI employee does a job for you, we record how the job was done rather than what it was about. That record is called execution data. It holds the shape of the work: which tool was called and in what order, what went wrong, what was retried, how the agent recovered, where a person stepped in to approve or correct something, and how long each step took.

Values are removed as the record is written. Where a step handled a name, an email address, a record number, or an amount, the record keeps a marker saying a value of that kind was there. The value itself is never written down. This is not cleaning up afterwards. The value never enters the record, so there is no earlier copy that held it.

A record of a task might read as: look up a customer, get nothing back, search again by email address, find a match, update one field, ask a person to approve, get approval, finish. That teaches a model how the work is done. It says nothing about who the customer was or what the field contained.

The three settings

Every setting is held for the whole organization and changed by an account owner. They live in your workspace settings, under Data Controls. You can also change any of them by emailing info@humanik.io with the subject line “Data Controls” and naming the workspace. We will confirm in writing when the change is done.

What is never in any of them

• The contents of any file, record, table, email, call, chat, or browser session
• The text of a prompt as you wrote it, or of anything an agent produced for you
• Names, email addresses, phone numbers, postal addresses, account numbers, and amounts tied to a person or business we could identify
• Credentials, API keys, secrets, and anything held in a vault
• Personal data about you, your staff, or your own customers
• The name of your organization, or any field that identifies you as the source
• Anything from a model running on hardware you own
• Anything covered by a signed agreement that says otherwise

There is one more limit that does most of the work. A sequence of steps is only used if the same sequence has been seen across several unrelated customers. A way of working that only your business performs never reaches that count, so it is never included. It stays yours because it does not qualify, not because we chose to leave it out.

Questions we expect

Does turning any of this off cost me anything?

No. Shared templates and model improvement can be switched off with no change to your pricing, your access, or any feature. Dataset licensing works the other way: it is off already, and if you switch it on you receive a credit against your platform fees.

Are you selling my data?

We license de-identified datasets, and only if you turn that setting on. They contain no customer content and no personal data. We do not sell personal data, we do not share data with advertisers, and we do not give personal data to data brokers.

Can I change my mind?

Yes, at any time. A change applies to new data straight away, and data already held is removed from our working set within 30 days. Two things cannot be undone: a template already built, and a dataset already delivered to someone. Neither can be taken back, because no single customer's part of it can be picked out again.

We have a signed agreement with you. Which wins?

Your signed agreement. Where it says something different from this page or from the Privacy Policy, the agreement governs and these settings do not apply to you.