Dependency Is Not Advice: Why Tax Functions Are Taking Their Data Back

The market read the last decade as tax moving out of the building. The evidence moved the other way.

For most of that decade the logic was clean, and it was not wrong. Compliance is a cost. Costs get outsourced. So the corporate tax function contracts to a small core, a manager or two and a Head of Tax, and the work goes to a panel of advisors who do it at scale for many clients at once.

That structure made sense while the deliverable was a document. If what the business owed the world was a lodged return and a signed tax note, then whoever produced the document was doing the job, and it hardly mattered whose building they sat in.

It suited the firms too.

"When I worked in an advisory firm, compliance had a name. We called it Paid BD. The compliance work was priced to win the relationship. The bigger engagements that followed, audit defence, transaction work, structure questions, were where the margin lived."

It is a rational commercial model, and it produced good work for a long time. But it explains something important about where the compliance file ended up, and why nobody on either side of the relationship had much reason to move it.

Two things have since made that arrangement harder to defend. Neither was a change in cost. One was a change in what the regulator asks for. The other was a change in what boards will tolerate.

THE QUESTION THE ATO STARTED ASKING

Justified Trust altered the terms without anyone announcing a new era.

Under the old model, the examination was of the answer. Is the number right. Does the return reflect the law. That is a question an advisor can answer on your behalf, because it is a question about a document, and the document can be prepared by anyone competent.

Justified Trust asks something structurally different. It asks you to demonstrate that your organisation governs the production of the number. Not that the answer is correct, but that the process which produced it is designed, controlled, documented and repeatable, and that you can show this on request rather than assemble it on request.

You cannot outsource the demonstration that you own something. The moment the regulator asks how your organisation governs a process, the answer becomes a property of your organisation: your systems, your controls, your people, your records. An advisor can help you design it. An advisor can review it and tell you where it is thin. But an advisor cannot be the answer, because the question was about you.

This is why evidence has a location. Evidence that lives in your systems is a control being proved. Evidence that gets assembled from somebody else's files in response to a question is a control being performed. Both might produce a satisfactory review. Only one of them is governance.

THE RISK BOARDS STARTED PRICING

The second shift came from board tables, and it came fast.

Australian boards have now watched, more than once, a professional services relationship become untenable for reasons that had nothing to do with the quality of the tax advice. Confidential government information misused inside a firm. Conduct findings that made a name difficult to defend in a boardroom. Government work withdrawn, business units sold for nominal sums, senior partners departing inside a news cycle.

You know the matters. Everyone in this market does, and naming them adds nothing here, because the lesson does not depend on which firm it was or what exactly was done. The lesson is the shape of the consequence.

Relationships that were expected to run for decades ended in weeks, and the trigger was never a technical failure in the tax work.

And when one of those relationships ends, the tax function discovers what it actually outsourced. Not just the preparation. The workpapers. The mapping logic. The reasoning behind a treatment adopted four years ago by a manager who has since moved to another firm. The knowledge of why the prior-year true-up sat where it sat. All of it correct, all of it defensible, and none of it in the building. It is key person risk at the scale of a firm.

That is the exposure. It is not a quality risk, because the work was usually good. It is a continuity risk, and it turns a commercial decision that should take a quarter into a project that takes a year.

A Global Head of Tax at a fifty billion dollar multinational put it more precisely than I can:

"My concern was not the cost. It was that the workpapers and the working that produced the numbers were not owned. When the dispute landed, the absence of owned evidence stopped being a theoretical risk and became the actual problem."

Read the last clause again. Not a risk that materialised. A risk that stopped being theoretical and became the thing they were actually dealing with, at the worst possible moment to be dealing with it.

Boards have started asking the question that follows: if we had to change advisors next month, what would it cost us. Not in fees. In months.

DEPENDENCY IS NOT ADVICE

None of this says stop using the Big 4. The shrewdest tax functions in this country use them heavily, deliberately, and will continue to. They should.

A contentious position deserves an external view. When a treatment is genuinely arguable, the value of a major firm is not that they can prepare a workpaper. It is that they see hundreds of taxpayers, they know how the ATO has responded to similar positions elsewhere, and they can tell you where the market has landed. They are a barometer. On a difficult question, buying a considered second opinion from a firm with that visibility is competent risk management, and it is exactly what the board should want you to do.

Signing off on judgement is not the same as holding the file. The distinction is between buying an opinion and surrendering a record. One is a service you purchase when you need it, on a question you choose, and stop purchasing when you do not. The other is a structural position where your compliance history, your reasoning and your operating knowledge sit inside an organisation you do not control.

You should be able to change advisors without losing your memory.

Use the firms for judgement, for contentious positions, for the view across the market you cannot see from inside one taxpayer. Keep the ledger, the reasoning and the evidence in the building, where the regulator has now made clear the responsibility was always going to sit.

WHAT THIS DOES TO THE SHAPE OF A TEAM

The obvious objection is headcount. If the work comes back in-house, does the function need to double?

The evidence says no.

One ASX-listed group we work with runs tax for 126 legal entities across fifteen countries with three people. Not three people plus a panel doing the real work. Three people. Their data uploads went from five hours to ten minutes. The journals that took a fortnight take three days. Routine reporting no longer depends on an external firm at all.

What those numbers buy is the same three people with their weeks handed back, spending them on positions rather than production. The compliance did not move to more humans. It moved to a system, which is what freed the humans for the part only humans can do.

This is the shift the whole series has been circling. A function that exists to produce documents is measured on whether the documents went out. A function that protects value is measured on whether the business made better decisions because tax was in the room. The second is a different job, and you cannot get to it by asking a team that is already at capacity to try harder.

AI SHARPENS THE QUESTION

Every vendor in this market is now telling you that artificial intelligence changes the calculation. It sharpens the one already in front of you instead, and it does so in a way that should make the in-house question more urgent rather than less.

An AI is bounded by the structure of what it reads. Point a capable model at a tax ledger where every classification carries the reasoning that produced it, where the general ledger ties to the tax base by design and the workpapers are objects rather than files, and it is a genuine multiplier. It can answer the how and the why in seconds because the how and the why were captured when the transaction landed.

Point that same model at a decade of rolled-forward spreadsheets and it will produce something worse than nothing. It will produce confident answers that nobody can defend. The model cannot recover reasoning that was never written down, so it will infer, and its inference will be fluent, plausible and unsupported. In front of an auditor, fluent and unsupported is the worst possible combination.

So AI does not resolve the question of where your tax operation should live. It raises the price of getting it wrong. A function with its reasoning in structured form gets a multiplier. A function with its reasoning in people's heads and other firms' files gets a very articulate liability.

THE TELL IS IN YOUR OWN JOB ADS

If you want a quick read on which side of this your function sits, do not look at the strategy deck. Read the last job ad you posted.

Most tax job ads describe a treadmill: prepare the provision, reconcile the accounts, manage the lodgements, maintain the workpapers, repeat. Then a requirements line asking for years of experience in a particular piece of software, which quietly tells every applicant that operating the machinery is a qualification in its own right.

The best people read that correctly. They read it as a function where the work is production rather than judgement, and they do not apply. You end up hiring the best available operator of the machine, when what the role needed was the best available tax mind.

You can only write the other advertisement, the one describing advisory work and business partnering and a seat in the decisions, when the compliance genuinely runs itself. Until then the ad would not be true, and good candidates can tell.

WHERE THE REASONING LIVES

Strip away the regulatory language and the vendor noise and the question is not complicated.

The controller asks where the work should be performed, and reaches for whoever performs it most cheaply. That was the right question when the deliverable was a document.

The CFO asks where the reasoning should live, and reaches for whatever arrangement means the company can answer the regulator, change advisors, absorb the loss of a key person, and still know why every number is what it is. That is the right question now, and it produces a different answer.

The tax function is not coming back in-house because outsourcing failed. It is coming back because the thing being asked for changed, and the new thing cannot be held anywhere but here.

Keep buying judgement. Stop renting your memory.

Off the treadmill. On to impact.