If you run a financial advice business, a lending brokerage, an accounting practice, or any firm offering “financial services” under the Corporations Act, the way you use AI in 2026 is no longer just a tech question. It’s an ASIC question. And the regulator is paying attention.

This article is written for Australian business owners who are either already using AI tools in their finance workflows or are seriously considering it. I’m going to walk you through what ASIC has signalled for 2026, where the real risks are, what it costs in AUD to do this properly, and how I’d approach it if I were sitting in your chair.

I’m writing this from New Zealand where I run Enterprise DNA, but a meaningful chunk of my work sits across the Tasman. Many of the AU businesses we work with face the same AI governance questions as their NZ cousins, but with a different regulator, different cost base, and slightly different rules. The detail below is current as of mid-2026, though verify with your lawyer or advisor before acting on anything specific.

What ASIC Has Actually Said About AI

ASIC is not new to this. The regulator has been publishing on AI-enabled conduct since at least 2023, but 2025 and 2026 have been when the tone has shifted from “guidance” to “enforcement posture.”

The two pieces most relevant to you are:

• ASIC Information Sheet 271 (formerly RG 271) on internal dispute resolution still applies to AI-driven decisions, particularly credit and hardship handling.
• Regulatory Guide 265 on misleading conduct in digital environments has been updated to address AI-generated advice and chatbot outputs. If your AI produces advice, RG 265 is the lens ASIC will look through.
• APRA CPS 234 on information security applies if you’re an APRA-regulated entity (banks, insurers, super funds, some larger mutuals). Any AI vendor handling your data sits inside your CPS 234 perimeter.
• Privacy Act 1988 (Cth) and the proposed 2025 tranche reforms also matter, especially the clarified rules around automated decision-making.

The practical takeaway: ASIC is signalling that you cannot outsource accountability to a vendor. If an AI tool gives a client bad advice, your AFSL holder is on the hook. If a credit decision is wrong, your responsible manager answers for it.

What “Financial Services” Actually Covers in This Context

A lot of business owners I speak with don’t realise how broadly “financial services” is drawn under section 766A of the Corporations Act. It includes:

• Providing financial product advice (general or personal)
• Dealing in financial products
• Making a market
• Operating a registered managed investment scheme
• Providing custodial or depository services
• Various credit activities under the National Consumer Credit Protection Act

If your AI tool is producing any form of advice, recommendation, or decision about a financial product, you’re in scope. That includes a chatbot on your website that suggests a particular credit card, or an internal tool that screens loan applications.

One Sydney accounting firm I spoke with recently built an internal AI assistant to draft general advice memos for SMSF trustees. Useful, time-saving, and clearly caught by the financial services regime the moment the memo leaves the building. They had to retrofit governance, which I’ll cover below.

The Real Risks for AU Businesses in 2026

Let me be blunt about what we typically see when we audit AI usage inside Australian finance businesses. The risks cluster into four buckets.

Advice quality and hallucinations. AI models can produce confident, fluent, wrong answers. In an advice context this is catastrophic. A hallucinated comparison rate on a home loan could expose you to a client claim or an ASIC enforceable undertaking.

Disclosure failures. RG 265 plus the existing advice disclosure obligations mean your clients need to know when they’re getting AI-generated content. If your chatbot doesn’t disclose, you have a problem.

Data handling. APRA CPS 234, the Privacy Act, and your contractual obligations to clients all stack up here. Sending client financial data to a US-hosted LLM without proper controls is a real concern. Verify the specific cross-border data transfer rules with your advisor, as the 2025 privacy reforms tightened this further.

Bias and discrimination. ASIC has been explicit that AI-driven credit decisions need to be tested for adverse outcomes under the anti-discrimination and responsible lending laws. This isn’t optional.

What Does Doing This Properly Cost in AUD?

Let me give you a sense of the cost layers, based on what we see across Australian mid-market finance businesses. These are approximate figures, not quotes, and vary a lot with size and complexity.

Layer 1: Tooling. Most of our AU clients start with Xero or MYOB for the core accounting data, then layer AI on top. A typical stack looks like a foundation model subscription (roughly AUD $40 to $80 per user per month), a workflow automation tool (AUD $20 to $60 per user per month), and possibly a domain-specific AI for advice drafting (AUD $200 to $1,500 per month per practice). For a 10-person advice firm, plan on AUD $50,000 to $120,000 per year all-in for tooling, including integration and licenses.

Layer 2: Governance and policy. This is where most businesses underinvest. You need an AI usage policy, a model risk register, a data handling protocol, and disclosure language for clients. Expect to spend AUD $15,000 to $40,000 with a specialist consultant to get this right the first time. Cheaper than fixing it after an ASIC inquiry.

Layer 3: Ongoing monitoring. You cannot set and forget. Monthly model reviews, output sampling, and incident response drills cost roughly AUD $5,000 to $15,000 per month if you outsource, or about 0.5 of a full-time equivalent if you build internal capability.

Layer 4: Legal review. Budget AUD $10,000 to $25,000 annually for a law firm to keep your disclosures and contracts current with ASIC’s evolving position. This is one area where I refuse to be casual. Verify specifics with your lawyer.

For a small advice business with three advisers, a realistic total cost of ownership for AI done properly sits around AUD $80,000 to $150,000 in year one. That’s not trivial, but it’s also a fraction of an ASIC enforceable undertaking, which industry estimates suggest typically run into the seven figures.

Where Australian Businesses Are Getting Genuine Value

I don’t want to paint a picture of doom. The businesses in our network that have done this well are seeing real returns.

Compliance automation. AI is genuinely good at reading long product disclosure statements and comparing them to client circumstances. One Melbourne paraplanner we work with has cut statement-of-advice preparation time by roughly 40 percent using a domain-specific tool with human review at the end. ASIC hasn’t blinked because the human adviser remains the accountable party.

Client servicing at scale. Several AU advice firms are using AI to draft meeting prep notes from Xero data and CRM records. The adviser reviews and personalises. Clients get more relevant conversations. Time savings are real, usually 3 to 5 hours per adviser per week in our experience.

Internal knowledge management. A Brisbane credit assistance business built an internal Q&A assistant over their own policy library. New staff onboard faster. The risk surface is small because the AI only sees internal, vetted content.

Marketing and content. AI for website copy, social, and client education is low risk provided you have a human in the loop and you’re not making financial claims that need to be in your AFSL.

The Non-Negotiables If You’re Going to Use AI in AU Financial Services

Based on what ASIC has signalled and what we’ve seen working in practice, here is the floor.

1. Document everything. You need an AI register listing every tool, what data it touches, who approved it, and when it was last reviewed. ASIC has been asking for exactly this in its recent supervisory engagements.

2. Humans must remain accountable. AI can draft, suggest, or summarise. It cannot be the author of advice. The responsible manager on your AFSL signs off. No exceptions.

3. Disclose to clients. When AI is involved in client-facing output, say so. Plain English. “Parts of this response were generated using AI and reviewed by [adviser name].” Industry estimates suggest this level of disclosure will become standard in the next 12 months.

4. Lock down data. Use enterprise-grade AI deployments, not consumer chat tools, for any client data. Audit vendor data handling. Ensure contracts give you audit rights. CPS 234 compliance extends to your vendors.

5. Test for bias. If you’re making credit, insurance, or superannuation decisions, run quarterly bias testing on your models. Document the results.

6. Train your team. Everyone touching AI in your business needs to understand what it can and cannot do. We typically see a half-day workshop delivered internally every six months as the minimum.

The 2026 Watch List

A few things to keep an eye on over the next 6 to 12 months.

• ASIC’s updated guidance on AI and the advice licensee’s obligation, expected in the second half of 2026.
• The final form of the Privacy Act reforms, particularly the automated decision-making provisions.
• APRA’s stance on third-party AI risk, with cross-industry consultation likely in 2026.
• Possible industry codes from the Financial Services Council and the Customer-Owned Banking Association.

If any of these land in a way that affects you, you’ll want to act within 90 days. Build that buffer into your planning.

A Simple First Step

If you’re an Australian business owner in financial services and you haven’t started, here’s what I’d do this week. Pull together a list of every AI tool anyone in your business is using, including the free ones and the personal accounts. You’ll be surprised. Industry estimates suggest the average mid-market finance business has 8 to 15 AI tools in active use, often without IT visibility. That map is your starting point.

From there, the work is governance, disclosure, and discipline. Not exciting, but it’s what separates the businesses that will use AI to grow from the ones that will end up in an ASIC report.

Want the practical version of this? The free Working With Claude field guide covers the full Claude ecosystem, Claude Code, and how to roll it out across a real business. Download it here.