AI & DATA PRACTICES
How we use AI on your business, and where the lines are.
Last updated: July 22, 2026
Why this page exists
If you are going to let someone connect software to your email, calendar, files, or finances, you deserve to know exactly how they work. These are the standing rules for every Mote Ops engagement. The written scope for each client repeats them with the specifics filled in.
The five working rules
- Minimum access. A system gets only the access its agreed workflow requires. If the workflow is enrollment follow-up, the system does not touch payroll.
- Synthetic testing first. Every build proves itself on fictional records before it is connected to real business data.
- Visible approval rules. Consequential actions, such as messages to customers and anything involving money, wait for a person. The approval points are visible in the system, not buried in configuration.
- No invented results. We publish client outcomes only after they are measured. Until then the work is labeled as in progress.
- Documented model and data choices. Which models run, where data is stored, and how long it is kept are written into each engagement's scope.
Model choices
Mote Ops is tool agnostic. Depending on the task, the privacy requirements, and the budget, a system may use hosted models, locally run open models on client-controlled or Mote Ops-controlled hardware, conventional automation with no AI at all, or a mix. The choice is made after the workflow is understood, and it is documented so you always know what is running and where.
Data handling in client builds
- Access is granted per workflow and can be revoked by the client at any time.
- Retention is defined in the written scope, and data is removed when the scope says so or when the client asks.
- Client business data is not used to train models and is not shared between clients.
- The finished system belongs to the client. Mote Ops teaches your team to run it and does not hold it hostage behind a subscription.
What the public demonstrations are
Every demonstration on this website, including CC's Care Hub, uses fictional records. The interfaces and workflows are real; no production client data appears on this site, and the demonstrations make no live connection to any private system. Where a shown output was prerecorded, the demonstration says so.
The Email Organizer beta, specifically
The Email Organizer is the one Mote Ops system currently running against real mailboxes, in an invite-only beta. Participants join knowingly. It syncs mailbox content in order to organize it and prepare reply drafts. It is approval-first by design: it cannot send email on its own, because no send path exists in the code. During the beta, synced content is stored on Mote Ops-controlled hardware, and participants can have it deleted on request. Beta results are still being measured, and we say so wherever the organizer is mentioned.
Honest labeling
AI-generated images and mockups we publish are labeled as such. Numbers presented as live are live; numbers that are examples are labeled as examples. If you ever find something on a Mote Ops surface that looks functional but is not, tell us and we will fix it or remove it.
These practices are commitments about how we work, stated in plain language. Engagement-specific details always live in the written scope.