AI for Financial Advice Under DBFO
Australian financial advice practices are caught between two pressures at once. The cost of producing compliant advice keeps climbing, driven largely by the time it takes to document a single piece of personal advice. At the same time, the Delivering Better Financial Outcomes (DBFO) reform package is rewriting the rules that shape that documentation. For a licensee or advice group running dozens of advisers, the temptation to reach for AI to close the gap is obvious. So is the risk of getting it wrong under the eyes of the Australian Securities and Investments Commission.
This guide is written for principals, responsible managers, and compliance leads at Australian advice businesses that want to use AI seriously without loosening the professional and legal obligations that sit under every client file. It maps where AI genuinely reduces the documentation burden, where the DBFO reforms change the target you are aiming at, and where handing work to a model quietly manufactures licensee risk.
The problem DBFO is trying to solve
The advice affordability problem is well documented. Producing a personal advice document under the current framework is slow, and the Statement of Advice (SOA) has become the symbol of that slowness. The obligation to provide an SOA sits in the Corporations Act 2001 at section 946A, and years of defensive drafting have turned what was meant to be a communication tool into a long compliance artefact that few clients read in full.
The Quality of Advice Review, led by Michelle Levy and delivered to Government in late 2022, concluded that the system had drifted toward documents that protect licensees rather than inform clients. The Government's response became the DBFO program. Its stated aim is to make quality advice more accessible and affordable while keeping consumer protections intact.
DBFO has been delivered in stages. The first tranche, enacted through the Treasury Laws Amendment (Delivering Better Financial Outcomes and Other Measures) Act 2024, streamlined ongoing fee arrangements and consent, gave licensees more flexibility around the Financial Services Guide, and tidied several conflicted remuneration provisions. The second tranche, released in draft during 2025, contains the change that matters most for documentation: replacing the SOA with a more flexible client advice record, alongside clearer rules for how superannuation funds charge for and deliver limited advice.
The important point for planning purposes in 2026 is that tranche two is not settled. Public reporting through 2025 and 2026 indicates the timetable has slipped, with the fallout from high-profile investment collapses absorbing regulatory and political attention. ASIC has signalled it will keep issuing and updating guidance as the reforms land. That means practices should build AI capability that improves advice production today and can flex to a lighter documentation standard tomorrow, rather than betting the workflow on a specific future format.
Where AI Fits in Your Advice Practice
What AI can safely do inside an advice workflow
The single most valuable role for AI in an advice practice is drafting, not deciding. Personal advice in Australia carries a best interests duty, expressed in the Corporations Act at section 961B, and the related obligation to provide advice that is appropriate to the client. Those duties belong to a person, the relevant provider, and they cannot be delegated to a model. What can be delegated is the mechanical labour of turning a completed fact find, a set of file notes, and an adviser's reasoning into a clear, readable document.
Used this way, AI compresses the paraplanning bottleneck. An adviser records their reasoning during or immediately after the client meeting. A model then assembles a first draft of the advice document that reflects that reasoning, pulls the client's circumstances into a structured summary, and produces plain-English explanations of the strategy. The paraplanner or adviser then reviews, corrects, and takes ownership of the final content. The judgement stays human. The typing does not.
AI-Assisted Advice Production, Human in Control
There are four repeatable jobs where this pattern earns its place:
- Advice document drafting. Turning approved reasoning into a structured, readable advice document is the highest-value use. The model works from your inputs, not from the open internet, so the recommendation originates with the adviser.
- Fact find structuring. AI can convert a messy intake, including scanned statements and email threads, into a clean data summary for the adviser to confirm. It should surface what is missing rather than invent what is absent.
- Research synthesis. Summarising product disclosure statements, comparing features against a client's stated needs, and drafting the comparison narrative saves hours. The licensed adviser still makes the selection.
- Review and ongoing service preparation. Drafting review meeting agendas, flagging clients due for their annual review, and pre-populating updated circumstances keeps ongoing fee arrangements defensible.
Consider a typical advice group with 12 advisers each producing four to six pieces of personal advice a month. If paraplanning and drafting absorb several hours per document, an AI-assisted drafting layer that a human still reviews line by line can meaningfully lift capacity without adding headcount. The gain is real, but it only holds if the review is genuine rather than a rubber stamp.
Illustrative Impact: Advice Document Production
These figures are illustrative and will vary widely by advice complexity, licensee templates, and how disciplined your review process is. Treat them as a way to frame a business case, not as a promise. The point is the shape of the change, not a guaranteed number.
Where AI adds licensee risk, and how to contain it
The same capability that saves time can create exposure if you let the model do the reasoning. Several failure modes deserve explicit attention.
Fabricated facts and figures. Language models can produce fluent, confident text that includes numbers, product features, or projections that were never in the source material. In an advice document, a fabricated fee, an invented balance, or an incorrect projection is not a typo. It can breach the best interests and appropriate advice obligations and mislead the client. Every figure in an AI-drafted document must be traceable to a verified source before it goes out.
Advice that drifts from the adviser's intent. If the model is asked to recommend rather than to draft, it will generate a plausible-sounding recommendation that no licensed person actually formed. That is not personal advice under the law. It is a document dressed as advice. The control is a hard rule: the recommendation and its reasoning must originate with the relevant provider and be recorded before drafting begins.
Privacy and data sovereignty. Advice files are dense with sensitive personal and financial information governed by the Privacy Act 1988 and the Australian Privacy Principles, overseen by the Office of the Australian Information Commissioner. Feeding client data into consumer AI tools that may retain or train on inputs is a serious risk. Our guide on why the wrong AI tools leak business data into training sets covers this in detail, and it applies with extra force to financial files.
Tax advice boundaries. Advisers providing tax (financial) advice services operate under the Tax Practitioners Board regime and its Code of Professional Conduct. AI-drafted content that strays into tax positions the adviser is not authorised to give, or that misstates a tax outcome, creates a distinct compliance problem beyond the Corporations Act.
Safe Use Versus Risky Use of AI in Advice
| Metric | Risky | Defensible | Improvement |
|---|---|---|---|
| Who forms the recommendation | The model suggests it | The relevant provider | Duty stays human |
| Source of figures | Model-generated | Verified from client data | No fabrication |
| Client data handling | Consumer tool, unclear retention | Controlled, no training on inputs | Privacy protected |
| Final document | Sent largely as drafted | Reviewed and owned by adviser | Accountability clear |
The pattern across all of these is the same principle Solve8 applies to any regulated deployment. AI handles detection and drafting; a qualified person retains the decision and the accountability. We set this out more fully in our framework for deploying AI agents responsibly with data access controls and human override, which translates directly to an advice setting.
Building the governance layer before you scale
A single adviser experimenting with AI is a manageable risk. Twelve advisers each using their own tools, with no shared controls, is how a licensee ends up with client data scattered across consumer platforms and advice documents no one can fully vouch for. Before AI moves from experiment to standard practice, the licensee needs a governance layer.
A Staged Path to AI in Advice
The governance layer has four parts. First, a written AI usage policy that states plainly what advisers may and may not do, including the hard rule that recommendations originate with the adviser. Second, tool selection that guarantees client data is not retained or used to train external models, with Australian data handling where possible. Third, an audit trail that records what the AI drafted and what the human changed, so a file review can show that judgement was exercised. Fourth, training, because the biggest risk in practice is an adviser trusting a fluent draft without checking the figures.
This is the same discipline that well-run accounting firms are applying to AI under their own professional codes, and the parallels are close. Both professions carry personal duties, handle sensitive financial data, and answer to a regulator that expects evidence of process rather than good intentions.
How DBFO changes the calculation
The reforms do not change the core principle that a person owns the advice. What they change is the documentation standard you are drafting toward. If tranche two lands as drafted, the shift from a prescriptive SOA to a more flexible client advice record rewards practices that can produce clear, tailored, client-appropriate documents quickly. That is precisely the kind of output an AI drafting layer supports, provided the human review holds.
This is why the sensible move in 2026 is to build AI capability around your reasoning and data, not around a fixed template. A practice that has already separated the two, adviser reasoning captured cleanly, drafting handled by a controlled AI layer, human review owning the output, is well positioned whatever final shape the client advice record takes. A practice that has hard-wired AI to fill in a specific SOA template will have to rebuild when the standard moves.
There is also a broader consumer-law dimension. Every advice document is a set of representations to a client, and the Australian Consumer Law obligations around misleading conduct apply alongside the Corporations Act. Our analysis of how the ACCC and consumer law bite on AI implementations is relevant here: a fabricated or misleading figure in an AI-drafted document is a problem under more than one regime at once.
Practical starting point for a practice
The lowest-risk, highest-return way to begin is to target the drafting bottleneck with tight guardrails, and to leave the reasoning entirely with your advisers. Start with a small group, insist on line-by-line human review, log what the AI produced and what changed, and measure whether the time saved is real once genuine review is included. If it is, you have a repeatable process. If the review keeps catching material errors, you have learned something important before it reached a client file.
Financial advice sits inside a wider set of Australian compliance obligations that AI touches, from privacy and the Australian Privacy Principles to the broader APRA and ASIC expectations on financial services technology. The advice practices that will pull ahead are the ones that adopt AI with the review discipline, data controls, and human accountability their licence already demands, rather than the ones that adopt it fastest.
The reforms are still moving, and the regulator will keep issuing guidance. But the operating principle does not depend on the final legislation. Let AI carry the documentation load. Keep the advice, and the responsibility for it, firmly in human hands.
Related reading
- AI in Australian accounting firms: the regulator map
- Financial services AI compliance under APRA and ASIC
- Deploying AI agents responsibly: governance and human override
- Privacy Act compliance for AI in Australia
- How the wrong AI tools leak business data into training sets
- The ACCC, consumer law and AI implementations