How we use AI, and how we handle your data.
We build AI automation for clients and use AI tools in our own work. These are the rules we follow for both, written down so you can hold us to them.
When we build automation for you
You approve the tools first
Before we build, you get a list of every outside service that will see your data, what it will see and why. Nothing is added later without asking you.
We prefer tools that don't train on your data
Where a provider offers business or API terms that exclude your inputs from model training, we use those terms, and we show you where they are written.
Only the data the job needs
We send the minimum: the fields a step requires, not whole records or whole inboxes.
A person approves what matters
Anything customer-facing, financial or uncertain waits for a person's approval. The automation does the preparation; your team makes the call.
Every run is logged
You can see what the automation did, when, and with which inputs, so a wrong result can be traced and corrected.
There is an off switch
Each automation can be paused on its own without breaking your other systems, and it runs in accounts you own.
We measure before we claim
Time saved is measured on your actual work. We don't quote savings we haven't seen.
In our own work
We use AI tools in our day-to-day engineering and writing, including for drafting code, tests and documentation. An engineer reviews everything before it ships, including the pages on this website, and we fact-check anything we publish.
We don't paste client code, credentials or customer data into AI tools that a client hasn't approved for that use.
What we are not
We are engineers, not lawyers or compliance auditors. If your business is regulated, such as healthcare or finance, we will work with your compliance advisor and follow their requirements, but we don't give legal opinions on AI use.
Questions about a specific tool or dataset?
Ask before you commit to anything. We'll answer in writing.