Our AI principles

1. Humans stay in the loop on consequential actions

Anything that sends external messages, spends money, or changes records of record routes through review unless the client explicitly approves full automation with monitoring.

2. Transparency over magic

Every workflow we ship comes with a plain-language explanation of what runs, where, and why. No black boxes.

3. Data minimization

We process only the data needed for the task. We don't train foundation models on client data. We don't sell or share data with third parties beyond the processors required to run the system.

4. Honest measurement

We instrument outcomes — replies, qualified meetings, hours saved — not vanity metrics. If a workflow isn't moving the number, we say so and rebuild.

5. Bias and fairness

For workflows that touch hiring, lending, or other regulated decisions, we run bias checks against historical outputs and document mitigations.

6. We say no

We decline projects that deceive end users, evade compliance, or generate synthetic identities. There's no contract worth that.