1. Scope comes first
Every project starts from one workflow: input, output, owner, frequency, tools involved, data sensitivity, review points, and expected result. Vague requests like “automate my whole business” should start with a paid audit or narrower intake.
2. Human review is the default
AI systems can draft, classify, summarize, and assist. Important business decisions, customer-facing replies, financial/legal judgment, code merges, and destructive data changes should keep human approval unless a separate risk review says otherwise.
3. Client responsibilities
You are responsible for providing accurate context, safe sample data, access boundaries, timely feedback, and final approval for workflows used inside your business.
4. No secret sharing in intake
Do not submit passwords, tokens, API keys, full private datasets, customer records, or confidential documents through the quote form. Share anonymized examples first.
5. Payments and changes
Paid pilots and sprints use a written scope. Changes outside scope may require a separate estimate. Larger work normally starts with upfront payment before build work begins.
6. No guarantee of full automation
Some workflows should not use AI yet. A trustworthy answer may be “clean up the process first” or “keep this manual.”