Let’s see if we’re a match! 🤝 Schedule a call with us today.

[wp_dark_mode style="2"]

Australian Firms Are Using AI Without Rules. Here Is the AI Use Policy to Write Before 10 December

AI writing tool suggesting a clearer sentence

Contents

Australian brokers, financial advisers and small firms have taken up AI faster than almost anyone planned for, and most of them have never written down how staff may use it. Two surveys reported in the past week put numbers on the gap. More than 80% of financial advice firms now use AI, yet fewer than half have a formal policy governing it, according to intelliflo’s Australian Advice Efficiency Survey. The broader finance and insurance sector looks the same. If your team is already pasting client details into chatbots and there is no rule about it, you have about six weeks before that gap stops being an internal risk and becomes a disclosure the law requires you to make. Digital One Agency closes that gap for Australian businesses with a free AI audit and a written policy that matches how your people actually work. This piece sets out what belongs in it.

What the surveys actually found

The intelliflo research, reported on 20 October, surveyed Australian advice firms and found more than 80% using AI in some form. Fewer than half had a formal policy governing that use. More than half of advisers said responsibility for errors made by AI tools rests with the advice business, not the software vendor. Nick Eatock, intelliflo’s chief executive and co-founder, said AI “has firmly entered the mainstream of financial advice” and that adoption “is outpacing the development of governance frameworks inside many advice businesses.”

Insurance broking tells the same story from a different angle. An Insurance Business report on 19 October pulled together several data points: NAB research found 64% of finance and insurance small businesses used AI, against 42% across all industries; the federal government’s AI adoption tracker put all-SME use at 43% in the December 2025 to February 2026 quarter. A 2025 University of Melbourne and KPMG study, cited in the same piece, found 44% of employees had used AI in ways that broke their workplace’s policy, and 57% relied on AI output without checking it for accuracy.

Broking had already been measured directly. Connective’s AI Readiness Report, released earlier in 2026 and drawn from more than 300 Australian brokers, found 97% had no formal AI policy, 65% had no documented AI strategy, and 40% had no governance framework at all. Put the numbers together and the most common situation in Australian professional services right now is simple to state: AI in daily use, and nothing written down about it.

An approved list is not a policy

Some firms believe they have this covered because they have told staff which tools are allowed. They have not. Cynthia Yap, managing director of Dynamic Insurance Services, made the point plainly in the Insurance Business piece: “An approved list alone does not establish that visibility.” Knowing that your team may use a given assistant says nothing about what information they feed it, whether a human checks the output, or who answers for a mistake. As Yap put it, “AI can assist a broker, but it cannot replace professional judgement.”

A policy worth having is short and enforced rather than long and ignored. One page covers most firms if it answers six questions:

  • Which tools are approved, and who signs off a new one before anyone uses it.
  • What data may go in, and what must never: client identity details, health and financial information, anything under confidentiality or legal privilege.
  • When a human must review output, which at a minimum means anything sent to a client or used in a decision that affects them.
  • What records you keep of significant AI-assisted work.
  • Who owns an error, and what happens when one surfaces.
  • What you disclose to clients and regulators, and where.

The value is in the enforcement, not the wording. A rule nobody checks is the state 44% of employees were already in when they broke the policy their employer thought they had.

Want this built or fixed for your business?

Digital One Agency builds AI systems, apps, platforms and AI-search visibility for Australian businesses. Start with a free AI audit or book a meeting and we will tell you what it would take.

Book a meeting → Or get the free AI audit

Why 10 December changes the stakes

Until now, an AI policy has been good practice. From 10 December 2026 part of it becomes law. New Australian Privacy Principles, APP 1.7 to 1.9, require organisations covered by the Privacy Act to describe in their privacy policy how they use personal information in computer-assisted decisions that could significantly affect a person’s rights or interests. The Office of the Australian Information Commissioner published its final guidance on 30 September, updating its APP 1 guidelines and releasing a fact sheet.

Two details catch people out. First, the obligation reaches well beyond chatbots. Law firm Allens notes the OAIC treats it as capturing a broad range of software, decision-support tools and embedded logic, not just the systems you think of as “AI”. A scoring spreadsheet that decides who gets a quote can be in scope. Second, it is a transparency measure: it does not give individuals a right to contest a decision, but it does require you to say, in writing and in public, what your systems do. We covered the mechanics of updating the policy itself in our guide to the automated decision-making privacy obligation.

The inventory you cannot skip

Here is the part most coverage misses. You cannot write the APP 1.7 disclosure honestly until you know which of your systems make, or substantially contribute to, decisions about people, and what personal information each one uses. That is the exact inventory a proper AI policy forces you to build first. The compliance deadline and the governance gap are not two problems. They are one problem wearing two hats.

That reframes the work. A firm with “an approved list” has catalogued logos, not decisions, and the law now asks about decisions. The sequence that actually gets you there is short:

  1. Map every AI and automated tool in use, including the ones nobody sanctioned. Shadow use is where the risk and the broken-policy statistics live.
  2. Classify the data each one touches, and mark anything that is personal, health or financial.
  3. Flag every use that feeds a decision affecting a person. These are your APP 1.7 cases.
  4. Write the one-page policy, then update the privacy policy to match what you found.
  5. Train the team on the two or three rules that matter, so the document changes behaviour.

Do step one and most firms find tools in daily use that leadership had never heard of. That is the finding, not the exception.

When the fix is the tool, not just the rule

For firms that handle client information at volume, a written rule is necessary but not sufficient. If the real risk is staff sending confidential data to a public tool, the durable answer is to give them one that keeps the data inside your walls. That is where private AI earns its place: models and assistants that run in your environment, so client information never leaves it. Where you are building AI into a product or a workflow, secure AI system development means logging every call, keeping a person in the loop on anything that affects a customer, and being able to show a regulator exactly how a decision was reached. For a regulated firm that is the difference between believing you are compliant and being able to prove it.

What this costs and how to start

Start with the audit, because you cannot price or policy what you have not mapped. Our free AI audit finds where AI and automated tools are being used across your business, what data flows through them, and where your 10 December exposure sits. From there we scope a costed plan: the policy, the staff training, and any private or secure build the audit shows you need. We do not quote a number before the audit, because an honest figure depends on what it turns up, and we would rather be right than fast on that one.

The reason to do this with a partner who builds the systems, rather than a template off the internet or a pure advisory firm, is that the policy then matches a real implementation. Our AI consulting services and build teams sit together, so the rules you adopt are ones your tools can actually enforce. Book a meeting and we will map your AI use and your 10 December obligations in one call, then send you a written AI use policy you can adopt that week.

Frequently asked questions

Does a small business need an AI use policy if staff only use ChatGPT occasionally?

Yes. Occasional use is still use, and the University of Melbourne and KPMG research found 44% of employees had broken a workplace AI policy and 57% used output without checking it. A one-page policy that names approved tools, bans certain data and requires human review of client-facing output covers most small firms. The effort is small; the risk of an unchecked tool handling client data is not.

What exactly has to change by 10 December 2026?

From 10 December 2026, new Australian Privacy Principles (APP 1.7 to 1.9) require organisations covered by the Privacy Act to describe in their privacy policy how they use personal information in computer-assisted decisions that could significantly affect a person. The OAIC released its final guidance on 30 September 2026. It is a transparency obligation, so you must disclose what your systems do, and it can apply to decision-support software, not only tools labelled AI.

Isn't an AI policy just a document? How do we make it stick?

A document that nobody enforces changes nothing, which is why we favour one page with two or three hard rules over a long manual. Make it stick by building the inventory first, so the rules reflect real tools, then training staff on the specific actions that matter and auditing a sample of AI-assisted work. The policy earns its keep when it is tied to how people actually do the job.

We are worried about client data going into AI tools. What is the safest setup?

The safest setup gives staff an approved tool that keeps data inside your environment instead of relying on a rule not to paste confidential information into a public one. Private AI runs models in your own environment so client data does not leave it, and secure AI system development adds logging, human review and an audit trail. For firms handling health or financial data, that combination is usually worth the build.

If an AI tool gives a client the wrong answer, who is responsible?

In practice, the business is. More than half of advisers in the intelliflo survey said responsibility for AI errors rests with the advice business rather than the vendor, and professional obligations do not transfer to a piece of software. That is the core reason to require human review of anything client-facing and to keep records: so that a person remains accountable for the output and you can show how a decision was made.

Need Custom AI & RAG Pipelines in Australia or the UK?

Digital One Agency is a senior, founder-led engineering team based in Australia (Gold Coast HQ) with active UK clients in London, Manchester and Edinburgh. We build production RAG pipelines and bespoke AI systems — fine-tuning, vector search, evaluation harnesses included.

  • Free 30-minute discovery call with a senior engineer
  • Fixed-scope quote with milestone billing — no hourly rates
  • AU & UK references on request
  • Bespoke code you own — not a SaaS reseller, not a ChatGPT wrapper

Book your free discovery call → Or send us your project details

Share

Kickstart a project today!

Fill out the form, and a member of our team will reach out to discuss your project and answer any questions you might have. 

You might also like...

Let’s see if we’re a match! 🤝 Schedule a call with us today.

[wp_dark_mode style="2"]
TOP
Part of our group: CISO Advisory Australia — independent cyber security & Virtual CISO services