What to confirm before sharing examples
Early AI conversations should be useful without inviting sensitive material too soon. Before any client examples are sent, confirm the workflow, the data owner, the sensitivity level, and whether anonymised or synthetic samples will be enough.
- Name the workflow and the business owner.
- Decide whether examples can be public, synthetic, anonymised, internal, confidential, personal, client-sensitive, or restricted.
- Confirm who is authorised to discuss the workflow and source material.
- Avoid sending live client, employee, financial, health, legal, credential, or regulated data before a route is agreed.
- Record whether the first step is fit check, paid diagnostic, governance, training, or stop.
Green, amber, and red signals
A simple traffic-light view helps the buyer avoid over-sharing. Green workflows can be discussed with safe examples. Amber workflows need clarification. Red workflows need a formal data, compliance, or tool route before materials are reviewed.
- Green: one named workflow, low-sensitivity or synthetic examples, and a named human reviewer.
- Amber: the workflow is useful but owner, tool route, reviewer, or retention rule is unclear.
- Red: sensitive live data is needed immediately, the data owner is unclear, or the buyer expects uncontrolled public AI upload.
- Red also includes autonomous regulated judgement, credentials, legally sensitive material, or high-consequence output without review.
- When in doubt, pause and define the boundary before receiving files.
How it changes the first call
The point is not to make the buyer fill out a heavy compliance form. It is to make the first conversation sharper, safer, and more commercially useful.
- Use descriptions of the workflow before sharing documents.
- Use anonymised or synthetic examples where possible.
- Name the approved tool route before testing anything live.
- Keep human review visible in the workflow design.
- Turn the agreed boundary into the proposal, kickoff note, and closeout evidence.