The free AI-use register: prove what your AI did, starting today.
Your firm uses AI every day. This is the simplest way to start proving it — no software required. Keep one row per AI-assisted client job, filled in as the work happens (not reconstructed later — that’s the whole point). A shared spreadsheet works to begin.
One row per AI-assisted job. Eleven fields.
| FIELD | WHAT TO RECORD |
|---|---|
| Date | When the AI-assisted task ran |
| Client / matter | Who it was for (use a code if sensitive) |
| Task / intent | What was actually asked of the AI |
| AI tool + model | Which tool and model were used |
| Inputs / sources | What information the AI was given |
| What the AI produced | The output, in one line |
| What changed in review | Edits or corrections a human made |
| Reviewer (who signed off) | The qualified person who stood behind it |
| Sign-off date | When it was reviewed and approved |
| Evidence location | Where the record or file lives |
| Client disclosed? | Was AI use disclosed to the client? |
Building inspection? Swap the middle fields for Property · Observation · AI-drafted finding · AS/NZS 4349.1 clause cited · Inspector sign-off · Photo/voice evidence ref.
A clause that does more than “AI may be used.”
“In providing these services we may use artificial-intelligence tools to assist with drafting, analysis and review. Where AI is used, a qualified member of our team reviews and approves the output before it forms part of your deliverable, and we maintain a record of how AI was used on your matter. That record is available to you on request. AI tools do not make final professional judgements; our practitioners remain responsible for the work.”
It commits to the three things a regulator or insurer actually wants to see — review, a record, and accountability. Have your own AU-qualified adviser confirm the wording before use.
Each field answers a regulator’s question.
| DRIVER | WHAT IT EXPECTS | COLUMNS THAT ANSWER IT |
|---|---|---|
| TPB Code of Professional Conduct | Demonstrate the quality-management system behind AI-assisted advice | Task · Reviewer · Sign-off · Evidence |
| APES 110 | Due care, objectivity, accountability | Reviewer · What changed in review |
| AUSTRAC (Tranche 2) | Evidence of who knew what, when | Date · Reviewer · Sign-off date |
| Privacy Act 1988 + ADM (from Dec 2026) | Transparency and control over automated processing of personal data | Inputs/sources · Client disclosed? |
| PI insurer (at renewal) | Documented AI-tool usage and controls | The whole register, kept current |
A manual register is a real start. Here’s where it breaks.
A register is far better than a Copilot log or a disclosure clause alone — but it has two failure modes: people forget to fill it in, and it gets reconstructed after the fact (which a sharp reviewer can tell). The fix is to capture the intent → evidence → sign-off loop automatically, at the point the work happens — so the proof is a by-product of doing the job, not extra admin. That’s loop engineering.
This template is provided free for AU practices to adopt as-is. It does not constitute legal or professional advice.
Three doors. Same destination: a firm whose AI work is provable.
Pick the one that fits where you are today — none commit you to anything.
Or simply email hello@xlooop.com — one human reads every message.