Start here
The free tool below runs entirely in this browser — no account, nothing stored or transmitted. Redact anything you would not want kept before you type it.
Open the companion tool →Why offloading the first step erodes judgement
The first ten minutes of a task is where you decide what matters. Handing that to a model is efficient until the model is wrong, unavailable, or confident about the wrong thing.
The goal is not less AI — it is keeping the professional judgement that makes your use of it safe.
The protocol
- Name three tasks where you now start with the model before thinking.
- For one of them, write the first version yourself for a week and note what changes.
- List the decisions you must never delegate: approvals, disclosures, numbers that go to clients, anything regulated.
- Set a verification habit for model output that applies every time, not when suspicious.
- Review after two weeks: what got better, what got slower, what you would keep.
What the tool does
Build a workplace AI boundary plan It runs locally in this browser and sends nothing.
Limit first: This is a planning and record-keeping aid, not therapy, medical advice or crisis support. If you are in crisis, contact a qualified human first.
Open the companion tool →Cost and free path
Cost: Free: one complete run in this browser, no account. Optional $4 one-time unlock allows unlimited runs of this tool on this browser. The free path and safety guidance never require payment.
Free path, in order (no cost):
- Your employer's compliance or HR policy on AI and client data
- NIST AI Risk Management Framework for a plain-language map of the risks
- A union or professional association, if you have one, for guidance you can rely on
What this cannot do: It cannot decide the outcome for you.
Sources checked 2026-10-03 · source bank reviewed against the live pages listed in metadata.
One human next step
If your employer has no policy on AI use and you handle client or regulated data, ask your manager or compliance team for written guidance and keep a copy of the answer.
Research log and safety checks
AI-assist path: not used for a verdict. This builder organises the reader's own documents and names an independent source; it does not infer authenticity, eligibility, or legal rights. §6.9/§6.10/§6.11 disclosed in the batch report; §6.12 status line carried in metadata as `status_line`.
§0.5 protection result: category is chatbot-dependency-check; crisis-adjacent: no; fear/urgency/scarcity toolkit used: no. The protective function is free and the paid feature is convenience only.