Data handling and deployment principles

Trust is a precondition for working on real business processes. These are the principles every engagement follows.

Draft principles. Final data-processing terms are agreed per engagement and are subject to legal review.

Start with safe examples

First conversations and Decision Sprints use anonymized or sample material and never connect to production systems.

Read-only by default

Hands-on work begins with read-only access to approved examples. Write access is granted only when needed, for a defined purpose.

Confidentiality first

Confidentiality and applicable data-processing terms are agreed before real data is shared.

No confidential samples in unapproved tools

Client data is never pasted into consumer AI tools. We agree which approved tools and environments are used.

Least privilege and audit trails

Access is limited to what the workflow needs and actions are logged so they can be reviewed.

Human approval for sensitive steps

Anything that sends something to a customer, moves money or changes a record of consequence has a named person approving it.

You own the process

Documentation, owners and runbooks are part of delivery, so you are never dependent on a black box.