What is reasonable care measured against?

The Code of Professional Conduct in the Tax Agent Services Act 2009 requires reasonable care in ascertaining a client's state of affairs (Code item 9) and in ensuring the taxation laws are applied correctly (Code item 10), but defines no fixed list of steps. The TPB's explanatory paper on the Code sets the benchmark as a competent and reasonable person possessing the skills, qualifications and experience required to become a registered tax agent, and for Code item 10 the standard expected of a reasonable person in the agent's position. It also states that reasonable care means what is reasonable in the circumstances.

Can a professional standard rise without anyone changing the rules?

It can, because the standard is written by reference to a competent and reasonable practitioner rather than to a closed list of steps. A checklist standard stays where it is until someone amends it. A standard defined by reference to competence tracks what capable practitioners in the field actually do, so it can move without any amendment at all. Whether it will move in this instance is uncertain, and the argument concerns years rather than months.

What is the argument against this view?

There are several. Peer practice shifts slowly and widely accepted is a high bar. Adopting a poor tool and relying on it without review is a worse failure than not adopting one at all. The quality management standards require a system, not a product, and a well-run manual review regime satisfies them. And there is currently no decision anywhere in Australia supporting the proposition. The argument concerns a direction of travel over years, not a present risk.

Opinion

Will not checking eventually be unreasonable?

Every article written about AI in Australian tax practice asks the same question: is using it risky? It is a fair question and we have written about it ourselves. But it is only half of one. Almost nobody has asked the version that points the other way.

Here is that version. Reasonable care is not a list. It is a comparison. And the thing you are being compared to is other practitioners.

If that comparison holds, then something follows that most firms have not considered: the standard you are held to is not entirely within your control, and it does not stay still.

Published 16 August 2026. This is opinion and general commentary, not legal, insurance or professional advice.

What the standard is actually attached to

Code items 9 and 10 of the Code of Professional Conduct require a registered tax practitioner to take reasonable care in ascertaining a client's state of affairs, and reasonable care to ensure the taxation laws are applied correctly. Neither says what those steps are. The Tax Agent Services Act 2009 does not contain a checklist, and it was plainly not drafted to.

So what is the yardstick? The TPB's explanatory paper on the Code answers it directly. For Code item 9, reasonable care is measured against "a competent and reasonable person, possessing the skills, qualifications and experience that are required to become a registered tax agent or BAS agent" - and where a practitioner holds themselves out as a specialist, against a competent and reasonable person professing to have those specialist skills. For Code item 10, it is the standard "that could be expected of a reasonable person in the agent's position".

The paper adds that reasonable care "means what is reasonable in the circumstances", varying with the scope of the service and the client's own knowledge and experience.

The yardstick is not a list of steps. It is a person - a competent, reasonable practitioner with your qualifications, in your position, facing your circumstances. And who that person is, is not fixed.

That is the whole argument in one paragraph. A checklist standard would stay where you left it. A standard defined by reference to a competent practitioner does not, because competence is a description of what capable people in the field actually do, and that description is rewritten quietly, by everyone, all the time.

Nobody announces a ratchet

The reason this is easy to miss is that standards of this kind never shift on a date. There is no transitional rule, no press release, no grandfathering. Something becomes common, then becomes ordinary, and only much later does its absence start to look like a gap.

Tax practice has a recent example. Client verification - actually confirming the person in front of you is who they say they are - was for years a matter of individual firm judgement. The TPB issued a practice note recommending minimum standards. It was guidance, not law. Firms adopted it. It was later reissued as a guidance statement, and the practices it describes are now simply what a careful practitioner is understood to do. At no point was there a day on which the obligation changed.

That is what a ratchet looks like from the inside: nothing happens, and then it has already happened.

So the question is whether systematic checking is on that path

Consider the direction of a few things that are not speculative.

Quality management obligations have tightened, and they have tightened inside the tax rules themselves. Section 40 of the Code Determination imposed a quality management obligation on registered tax practitioners from 2025 - larger practices first, then practices of 100 employees or fewer from 1 July 2025 - and the TPB has published guidance on supervision, competency and quality management under the Act. For members of Chartered Accountants ANZ, CPA Australia and the IPA in public practice, APES 320 has required a documented system of quality management since January 2023, including cyclical inspection of completed engagements by someone not involved in them.

None of that mentions technology. All of it pushes in the same direction: away from review as a matter of individual habit, toward review as a system the firm can describe, monitor and improve.

Meanwhile the volume of what has to be reconciled on a single file keeps growing, and the number of senior people available to reconcile it does not.

You can see how the argument assembles itself. If a growing share of firms come to run systematic checks across completed files, and that becomes the ordinary way careful firms work, then the competent and reasonable practitioner the Code measures you against starts to be someone who does it.

Why we might be wrong

We sell a review tool. That makes the argument above convenient for us, which is a reason to distrust it, so here is the case against - properly, not as a rhetorical gesture.

The ratchet is slow, and "widely accepted" is a high bar. A single practice does not become widely accepted because vendors say it should be. It takes years and a genuine majority. A firm reviewing files carefully by hand today is doing what the standards require, and will be for a long time.

The opposite failure is nearer and more likely. Adopting a tool and relying on its output without review is a live breach of reasonable care right now - not a hypothetical one in five years. The TPB's guidance on AI is explicit that using it does not transfer responsibility. On the current evidence, over-reliance will hurt far more firms than under-adoption.

The obligation is to a system, not a product. APES 320 and section 40 require a system of quality management. A manual review regime that is genuinely documented, genuinely monitored and genuinely remediated satisfies both, completely. Neither standard mentions software.

Bad tools make this worse, not better. A checker that is confidently wrong on Division 7A or trust distributions manufactures false comfort, which is more dangerous than no comfort. If a standard ever does shift, it will not shift toward "used a tool". It will shift toward "checked properly".

And there is no authority for any of this. No Australian court has held it. No tribunal, no regulator, no professional body. This is a reading of how open-textured standards behave over time. It is not a finding, and we are not going to dress it up as one.

What we actually think

Something narrower than the headline, and worth saying plainly.

We do not think a firm will be found negligent for not owning software. We think the more likely development is subtler: that as systematic checking spreads, the expectations around what a careful review demonstrably covers will quietly rise - that the range of things a reviewer is assumed to have reconciled will widen, because reconciling them will have become unremarkable.

That is a shift in expected thoroughness, not in expected tooling. A firm could meet it manually. It would just be more work than it used to be.

And it is worth noticing that this cuts against the usual framing of technology risk. The standard commercial pitch is that adopting AI is the risky, forward-leaning move and standing still is the safe one. If reasonable care is genuinely benchmarked against peers, then standing still is not a neutral act. It is a bet that the benchmark stays where it is.

It might. We are not certain it will.

A note on insurance, since it comes up

Practitioners often ask whether any of this affects professional indemnity cover, which tax and BAS agents must maintain to the Board's requirements as both a registration requirement and a Code obligation.

We are not going to claim that any tool lowers your premium. Nobody can substantiate that, and premiums turn on your claims history, practice profile and insurer appetite. But the recurring theme in what brokers and professional bodies publish is not about pricing at all - it is that documentation wins claims. Marsh's own published material on accountants' indemnity risk turns on a case where the accountant was well placed to defend the allegations because, in their words, the files were impeccable and the matters considered were documented.

Whatever happens to the standard of care, that part is already true. A file that shows what was checked is worth more at the moment a claim lands than one that shows only that someone signed it.

Where Taxpartna sits in this

Openly: this argument suits us. Taxpartna runs 45 or more automated checks across a completed Australian tax file, reconciles figures across source documents, and records what was tested, what it found, where it came from and who signed it off.

But the honest version of our own pitch is not that the law is about to require this. It is that a review which leaves a trace is better than one that does not, for reasons that hold whether or not the standard moves at all. The reviewer still makes every judgement. Taxpartna does not give advice, does not lodge, and does not sign off.

If the standard does rise, firms already working that way will not notice. That is usually how it goes.

See what an augmented review looks like

Watch the workflow demo Explore the features Talk to us

Frequently asked questions

No. There is no obligation on any Australian tax practitioner to use AI or any other particular technology, and no court, tribunal, regulator or professional body has held that failing to use one falls short of reasonable care. Anyone suggesting otherwise is overstating the position.