The pile on Bianca's screen in late 2022 was not a queue. It was a stack of claims that had never been handed to anybody at all — veterans' liability claims lodged months earlier, sitting in a departmental holding pattern with no case officer's name against them, because there was no case officer to attach. She had come to the Department of Veterans' Affairs from eleven years in a bank's financial-hardship team, and she recognised the shape of the problem within a fortnight. It was not a hard job being done badly. It was a job with more work arriving each week than there were people to do it, and a decade of accumulated arrears behind that. Her induction cohort was one of several hired into claims processing across that year and the next, as the department's staffing was lifted in response to findings that the delays themselves were doing harm to veterans and their families. Two years later, the specific, humiliating category of a claim that nobody had even opened had been worked down to close to nothing (approx.) (Department of Veterans' Affairs, annual reporting and Senate estimates evidence, 2023–2024).
That is what a staffing decision looks like when it works, and this article will not pretend otherwise. But look at Bianca's employment from the other side of the ledger, the side the Commonwealth's balance sheet sees. She is an ongoing APS employee. Her salary rises each year under an APS-wide enterprise agreement. Her employer contributes 15.4 per cent of that salary to superannuation, the standard employer rate for the main accumulation scheme covering Commonwealth employees hired since 2005. She has accrued leave entitlements, a redundancy framework, and a job that exists until a future government makes the politically expensive decision to abolish it. None of that is unreasonable, and none of it is a criticism of her. It is simply the nature of the commitment. A consultant's invoice ends when the engagement ends. Bianca's cost recurs, indexed, every year, and it grows with every wage settlement the Commonwealth signs.
There are roughly 39,000 more people in that position than there were in June 2022. The Australian Public Service employed 159,469 people at 30 June 2022 and 198,529 at 30 June 2025 — a rise of about 24.5 per cent in three years (APS Employment Database; Australian Public Service Commission, State of the Service reports, 2022–23 to 2024–25). The government that oversaw that increase campaigned explicitly against the cost of consultants and committed to rebuilding public-sector capability in-house. This article asks the question that matters and that the political argument mostly skips: not whether Bianca should have been hired, because on the evidence she plainly should have been, but whether the whole 39,000 produced capability gains proportionate to what has been permanently added to the budget's base — and what it means that the thing being cut was flexible and the thing replacing it is not.
What the government did
Three separate decisions sit behind the headcount number, and they need to be untangled before any of them can be judged, because they were taken for different reasons and carry different weight.
The first was structural and came almost immediately. Since 2015–16, Commonwealth agencies had operated under an average staffing level cap — a ceiling on the number of full-time-equivalent public servants each agency could employ, set centrally and held broadly flat. The cap did not reduce the amount of work government did. It changed who did it: agencies that needed more capacity and could not hire bought it instead through consultants, contractors and labour-hire firms, whose costs sat in supplier expenses rather than employee expenses and therefore did not count against the cap. The October 2022 Budget abolished that cap (Australian Government, Budget October 2022–23, Budget Paper No. 4: Agency Resourcing). Whatever else follows, this was an honest accounting decision. A cap that suppresses one category of spending by pushing the same work into a more expensive category is not a saving; it is a disguise. Removing it made the true size of the Commonwealth workforce visible again in a single number, which is a precondition for arguing about it at all.
The second decision was fiscal. The same budget booked savings from reducing spending on external labour, advertising, travel and legal services — a package in the order of several billion dollars over the forward estimates (approx.) (Budget October 2022–23, Budget Paper No. 2). Subsequent budgets extended and enlarged that measure. The logic offered publicly was substitution: the Commonwealth was paying consultancy day rates for work that permanent staff could do more cheaply and, over time, better, because the knowledge would stay in the building rather than leaving with the contractor. On its own terms this is a coherent argument, and it is the argument a private firm makes when it decides to stop renting a function and start owning it.
The third decision was institutional, and it followed the Royal Commission into the Robodebt Scheme, whose report was delivered on 7 July 2023. The Commission examined an automated income-averaging debt-recovery scheme that raised debts against welfare recipients on a basis later found to be unlawful, and it documented real harm — wrongly raised debts, aggressive recovery against people with no capacity to pay, and, in evidence put before the Commission, contribution to serious distress. Its findings ranged well beyond staffing, but they included an account of an administrative culture in which internal capability, institutional memory and the willingness to give frank advice had been degraded, and in which external providers had been engaged in ways that blurred responsibility for decisions affecting citizens' legal entitlements (Royal Commission into the Robodebt Scheme, Report, July 2023). Running alongside it was the collapse of confidence in the external-advice model itself: the tax-advice confidentiality breach at PwC, which became public in early 2023 and was examined at length by a Senate committee across 2023 and 2024, made "just hire a consultant" a politically untenable default for a period (Senate Finance and Public Administration References Committee, inquiry into the management and assurance of integrity by consulting services, 2023–2024).
The government's institutional response was a set of measures aimed squarely at insourcing. The APS Strategic Commissioning Framework, announced in 2023 and applying from 2024–25, required agencies to identify core work then being performed by external providers and set targets to bring it back in-house (Australian Public Service Commission, APS Strategic Commissioning Framework). An in-house consulting unit, Australian Government Consulting, was established within the Department of the Prime Minister and Cabinet to provide the kind of strategic advice departments had been buying from firms. And the Public Service Amendment Act 2024 legislated changes to the framework governing the service itself, including an explicit stewardship value — the obligation to maintain the institution's capability for future governments, not merely to deliver for the current one — and a system of capability reviews (Public Service Amendment Act 2024 (Cth)).
One clarification of terms matters here, because the numbers get misquoted in both directions. "APS headcount" as published by the Australian Public Service Commission counts people employed under the Public Service Act 1999 at a point in time. It is not the whole Commonwealth workforce: it excludes the Australian Defence Force, and it excludes staff of a range of Commonwealth entities employed under their own legislation. Average staffing level, or ASL, is a different measure again — a budget concept, averaged across a year and counted in full-time-equivalent terms, which is the number Budget Paper No. 4 uses and the number appropriations are built around. Headcount is always the larger figure. When this article says 198,529 it means APS headcount at 30 June 2025; when it says +4,199 it means budgeted ASL added in a single budget. They are not the same currency and should not be added together.
The strongest case for the policy
A Royal Commission found the state had hollowed out its own capacity. Rebuilding it is what a state is for.
Start with the harm, because it is not hypothetical. The Robodebt scheme raised unlawful debts against Australians who could least afford to contest them, and a Royal Commission spent a year documenting how a modern administrative state came to do that to its own citizens. Part of what it found was an institution that had lost the internal expertise, the institutional memory and the confidence to say no — the capacity to give a minister frank advice about legality and consequence, and the capacity to test its own systems before turning them on real people. That capacity does not come from a procurement panel. It comes from experienced officials with tenure, statutory obligations and careers that outlast a contract. A government that responds to a Royal Commission's findings by rebuilding the exact capability whose absence the Commission described is not being profligate; it is doing the thing an inquiry is for. Second, the alternative was not free and was not working. Consultancy and external-labour spending had grown to levels that Auditor-General reports and Senate committees criticised on value-for-money and probity grounds well before this government took office, and the PwC episode demonstrated that outsourcing sensitive government work carries risks that never appear in a day rate. Paying a firm a premium for a person who then leaves, taking the knowledge with them, and who owes their primary duty to a partnership rather than to the Commonwealth, is not obviously the cheaper option once the full cost is counted. Third, some functions should never be outsourced at all as a matter of principle. Deciding whether a veteran is entitled to compensation, whether a family keeps its payment, whether a debt is lawfully owed — these are exercises of statutory power over a citizen's rights. The person making them should be a public official, accountable through the public-service framework and the courts, not a contractor whose engagement can be terminated when the decisions become inconvenient. Fourth, and most concretely, it worked where it was tried hardest. The veterans' claims backlog that had accumulated over years fell substantially once staffing was lifted, and the improvement showed up in the department's own reporting, not in a press release. That is a measurable service-delivery gain delivered to people the state had demonstrably failed. And finally, the honest framing of the ASL cap: it did not save money, it hid cost. Reversing an accounting fiction and paying the true price of the work openly is better public administration than pretending the work is cheaper than it is. A permanent function properly staffed by permanent public servants is not waste. It is what a state is for, and the burden of proof should sit with those who want the state to rent its core capabilities rather than own them.
The evidence
The shape of the increase
The primary record on headcount is unusually clean, which is a mercy in a policy debate that generates more heat than most. The Australian Public Service Commission maintains the APS Employment Database and publishes point-in-time headcount every year in its State of the Service report. At 30 June 2022 the APS employed 159,469 people. At 30 June 2024 it employed 184,858. At 30 June 2025 it employed 198,529, an increase of about 7.4 per cent on the previous year alone (Australian Public Service Commission, State of the Service 2022–23, 2023–24 and 2024–25; APS Employment Database). The three-year increase is approximately 39,060 people, or about 24.5 per cent.
Set that against a benchmark to feel its size. Roughly 39,000 people is more than the entire population of a regional city like Warrnambool or Gladstone. It is an increase of about a quarter in three years in an organisation that had been broadly flat for most of the preceding decade. Australia's total employment grew over the same period, but not at anything like that rate — the APS grew several times faster than the labour market it recruits from (approx.; ABS Labour Force, 2022–2025). Whatever else it was, this was not incremental drift. It was a deliberate, rapid expansion, and it should be assessed as one.
What the increase costs, and why the cost is different in kind
The Commonwealth does not publish a single tidy line called "the cost of the extra 39,000". But Budget Paper No. 4 reports both average staffing levels and departmental employee expenses for every agency, and dividing one into the other gives an average total employment cost — salary plus superannuation plus payroll on-costs — in the order of $130,000 to $150,000 a year per employee across the general government sector (approx.; derived from Budget Paper No. 4: Agency Resourcing, recent years). Apply the low end of that range to 39,000 additional people and the recurring annual cost is on the order of $5 billion a year. Apply the high end and it approaches $6 billion. This is arithmetic, not a budget line, and the true figure depends on the classification mix of who was actually hired; a service delivery officer costs materially less than an Executive Level policy adviser. But the order of magnitude is the point, and the order of magnitude is billions of dollars a year, recurring, indefinitely.
Three features make that cost different in kind from the consultancy spending it was meant to replace, and they are the analytical heart of this article.
The first is permanence. An ongoing APS employee is not a contract that lapses. Reducing the number requires either a hiring freeze that takes years to bite through natural attrition, or redundancies, which cost money in the year they are made and cost political capital immediately. A consultancy engagement, by contrast, ends by default. It has to be actively renewed to continue. That asymmetry is the whole difference between a discretionary and a structural commitment, and it is why treasurers of both persuasions have historically reached for the consulting line first when they need savings: it is the line that can actually move.
The second is indexation. The wage bill does not merely persist; it grows. The APS-wide enterprise agreement concluded in 2024 delivered pay increases in the order of 11 per cent across three years (approx.; Australian Public Service Commission, APS bargaining outcomes, 2024). On a base of roughly $5 billion, a 3 to 4 per cent annual settlement adds $150 million to $200 million a year to that base, before a single additional person is hired. Compounding runs in the same direction as the headcount, not against it.
The third is superannuation. The 15.4 per cent employer contribution to the main scheme covering post-2005 Commonwealth hires is roughly a third higher than the superannuation guarantee that applies across the private economy. On a $5 billion wage base that difference alone is worth a few hundred million dollars a year relative to an equivalent private-sector payroll (approx.). More significantly, superannuation is a claim on the future: today's contribution funds a liability that will be paid out over decades. Employment decisions taken in 2023 and 2024 create obligations that outlast several parliaments.
A consultancy contract ends by default and has to be renewed to continue. A public-service job continues by default and has to be abolished to end. That single asymmetry is the entire fiscal argument.
Did the offsetting saving actually arrive?
The substitution case stands or falls on a testable proposition: if permanent staff were hired to replace external providers, external spending should have fallen by an amount of similar scale. This is where the public record becomes harder to read, and where honesty requires stating what can and cannot be established.
What can be established is direction. The Commonwealth publishes contract notices on AusTender, and departments report consultancy expenditure in their annual reports. Both show a reduction from peak years in the reported value of new consultancy engagements, and agencies reported progress against Strategic Commissioning Framework targets to convert outsourced core work into APS roles (approx.; AusTender contract notices and agency annual reports, 2022–23 to 2024–25; Australian Public Service Commission reporting on the Strategic Commissioning Framework). Something real happened; the direction of travel is not in dispute.
What cannot be established from the published record — at least not by this analysis — is an offsetting reduction of comparable magnitude. Two problems get in the way. The first is definitional: "consultancy" is a narrow reporting category, and the broader external-labour envelope also includes contractors, labour hire, professional services and outsourced service delivery, which are reported differently and are not always captured when a headline consultancy figure falls. A reduction in consultancies that coincides with continued or rising spending on contractors and labour hire is a reclassification, not a saving. Senate estimates evidence across 2023 to 2025 records agencies continuing to engage external labour at scale while simultaneously reporting insourcing progress (Senate Finance and Public Administration Legislation Committee, Budget Estimates, 2023–2025). The second problem is the counterfactual: some of the added staff replaced external providers, but some were added to do new or expanded work — new programs, new regulatory functions, expanded service delivery — where there was no external spending to save in the first place. Only the first group generates an offset.
This matters for how the policy should be judged. The government's case is a substitution case, and the burden of proof for a substitution case sits with the party asserting it. Where a saving is claimed, it should be demonstrable in a named line item, across the full external-labour envelope rather than the narrow consultancy category, and net of the staff added for new work. That evidence, published clearly and audited, would settle the argument. Its absence does not prove the saving did not occur. It does mean the claim remains unproven at the scale required to offset a recurring multi-billion-dollar wage bill, and a claim of that size ought to be provable.
What measurable output improved?
The second test is capability. Thirty-nine thousand additional people should produce something an ordinary citizen can see. On this, the evidence is genuinely mixed, and the mix is instructive.
The clearest positive result is veterans' claims. The Royal Commission into Defence and Veteran Suicide, in its interim report of August 2022, identified the backlog of unallocated compensation claims as an urgent problem causing harm and recommended it be eliminated. Staffing at the Department of Veterans' Affairs was increased substantially from that point. The department subsequently reported that the backlog of claims awaiting allocation to a delegate had been reduced to near zero (approx.; Department of Veterans' Affairs annual reporting; Royal Commission into Defence and Veteran Suicide, interim report Aug-2022 and final report Sep-2024). This is the strongest single data point in the government's favour anywhere in this analysis. It is a specific, measurable, independently identified problem that was fixed by adding people, and it was fixed reasonably quickly. It should be said plainly and without qualification: that worked.
Elsewhere the picture is less clean. Service-delivery performance at Services Australia — call wait times, claims processing times for payments including age pension and disability support — remained a subject of adverse commentary from the Commonwealth Ombudsman and in Senate estimates across the same period, with processing backlogs and telephone wait times persisting well after significant additional staffing had been added (approx.; Commonwealth Ombudsman reporting and Senate Finance and Public Administration Legislation Committee estimates, 2023–2025). Some of that reflects demand growth and system complexity rather than staffing. But it demonstrates that headcount alone is not a mechanism. Where the constraint was people — as in the veterans' backlog — people fixed it. Where the constraint is legacy technology, legislated process complexity or workflow design, adding staff produces slower improvement per dollar, and sometimes very little.
On the policy-advice side, the claim is harder to test in either direction, and this article will not pretend to have resolved it. "Better advice" is not a measurable output on a quarterly release. The proxies available — whether major programs are delivered on time and on budget, whether the Auditor-General's performance audits find fewer failures of administration, whether legislation requires less remedial amendment — move slowly and are influenced by far more than staffing. The Australian National Audit Office continues to publish performance audits identifying significant shortcomings in Commonwealth procurement, grants administration and program delivery across this period (Australian National Audit Office, performance audit reports, 2022–2025). That is not evidence the rebuild failed; audit offices exist to find problems and would find them under any staffing model. But it is not evidence the rebuild succeeded either, and three years in, the absence of a clear, published measurement framework for what the capability investment was supposed to deliver is itself a finding. If you spend billions a year permanently to buy capability, you should be able to say what capability you bought.
The critique
The criticism that follows is not that the public service is too large in some abstract sense, nor that public servants are unproductive, nor that the work does not need doing. It is narrower and, in the view of this analysis, harder to answer: the government solved a genuine problem using the least reversible instrument available, at speed, without establishing in advance how it would know whether the instrument had worked.
Begin with what was inherited and what was chosen, because the distinction does real work here. The government inherited the ASL cap, the outsourcing culture it produced, the hollowing of internal capability the Robodebt Commission described, and the veterans' claims backlog. It did not create any of those. It also inherited a fiscal position in which the pandemic-era spending surge had already lifted Commonwealth payments as a share of the economy well above pre-COVID levels. What it chose was the remedy: a rapid, permanent, across-the-board expansion of ongoing employment rather than a targeted one. Those are different decisions and only the second belongs to this government.
The first problem with the chosen remedy is that it converts a flexible cost into a fixed one at precisely the moment the budget most needs flexibility. Commonwealth spending is already dominated by categories that cannot be reduced quickly: pensions and payments set in legislation, the National Disability Insurance Scheme, health and aged care driven by demography, defence commitments running decades ahead, and public debt interest, which is now among the fastest-growing items in the budget. Each of those is either legislated, contracted or demographically locked. The discretionary residue — the part a treasurer can actually adjust in a given year — is a small and shrinking share of the whole. External labour was inside that residue. Converting it to permanent employment moves it out. The service the money buys may be identical; the budget's capacity to respond to a shock is not.
The second problem is compounding. A structural commitment does not stay the size it was when it was made. Wage settlements index it upward. Classification creep — the steady drift of positions toward higher pay bands as roles are reclassified and staff progress — pushes average cost up independently of numbers. Superannuation accrues against future budgets. A commitment of roughly $5 billion a year entered into in 2024 is not a $5 billion commitment in 2034; on plausible wage growth it is materially larger, and none of that growth requires a further decision by anyone.
The third problem is the missing measurement. If the rationale was substitution, the test is external-labour spending across the full envelope, published and audited. If the rationale was capability, the test is service and program outcomes, specified in advance. Neither test was framed clearly at the outset in a way that would allow a citizen to check the result three years later. Governments routinely require this discipline of the entities they fund — grant recipients, states under national agreements, service providers under contract, all must state what an appropriation will buy. A capability investment of this scale, funded permanently, should have been held to the standard it imposes on others.
The fourth problem is one of sequence. Robodebt's failures were not solely, or even mainly, failures of headcount. The Royal Commission's account described failures of legality, of governance, of frank advice, of record-keeping and of the willingness to escalate a known problem. Hiring more people does not by itself supply legal rigour or the institutional courage to tell a minister something unwelcome; those come from culture, structure, protections for advice and consequences for failure. The Public Service Amendment Act 2024 addresses part of that, and deserves credit for doing so. But headcount was the fastest and most visible response, and there is a persistent risk in public administration that the visible response substitutes for the harder one. If the diagnosis was cultural and the treatment was numerical, the treatment may be sound and still not be the cure.
If the diagnosis was cultural and the treatment was numerical, the treatment may be entirely reasonable and still not be the cure.
The fifth problem is asymmetric reversibility, and it is the one that will outlast this government. Suppose the substitution turns out to have been only partly effective — that some meaningful share of the 39,000 replaced work that had already stopped being necessary, or that could have been done by fewer people with better systems. There is no clean mechanism to unwind it. The consultancy line was cut in a budget. The staffing line can only be cut through redundancies, which cost cash upfront and land on individuals, or through attrition freezes, which fall on whichever functions happen to lose people first rather than on the functions the government would choose. Neither is a policy instrument any treasurer reaches for willingly. The decision was made in months. Reversing it, if reversal were ever warranted, would take years.
What they got right, and the honest caveats
A fair audit has to say clearly where this policy earns credit, and there is more of it here than the critique above might suggest.
The veterans' claims result is the strongest case, and it should not be buried in qualifications. An independent Royal Commission identified a specific harm, the government added specific staff, and the harm was substantially reduced within about two years. That is a policy chain that ran from evidence to action to measurable outcome, and it is precisely what people say they want government to do. Bianca's team is real in the sense that matters: the improvement it produced is visible in departmental reporting, not just in ministerial claims.
The abolition of the ASL cap deserves credit as a matter of accounting honesty, independent of what followed. A cap that pushes work into a more expensive category while reporting a lower staffing number is a fiction, and removing a fiction is a public good even when the truth it reveals is expensive. The Commonwealth's workforce cost is now more visible in a single set of numbers than it was, which is why an article like this one can be written at all.
The principle that statutory decisions over citizens' entitlements should be made by accountable public officials is correct and should not be controversial. Whatever the fiscal argument, there is a category of government work — deciding a debt, a payment, a compensation claim, a visa — where the identity and accountability of the decision-maker is part of the legitimacy of the decision. Reversing the drift of that work toward contractors is defensible on grounds that have nothing to do with cost, and the Robodebt record supplies the reason.
And the growth is decelerating, which is the caveat this analysis is most obliged to state plainly because it cuts against its own thesis. The 2025–26 Budget added 4,199 ASL (Budget 2025–26, Budget Paper No. 4: Agency Resourcing). That is a materially smaller increase than the sharp rises of the preceding years, and it indicates the expansion phase is being brought to an end rather than continuing indefinitely. Anyone extrapolating the 2022 to 2025 growth rate forward is misreading the record. The correct characterisation is a large one-off step change in the size of the service, now flattening — not a runaway trajectory.
Two further caveats belong here. First, some portion of the headcount increase is conversion rather than addition: people already doing Commonwealth work as labour hire or contractors, moved onto the APS payroll. Those people are new to the headcount but were not new to the payroll in any economic sense, and the true net increase in people working for the Commonwealth is smaller than 39,000 (approx.). This cuts both ways — it moderates the cost criticism, and it strengthens the demand for evidence that external spending fell correspondingly, since conversion is exactly the case where the offset should be visible. Second, part of the increase funded genuinely new functions created by Parliament, including new regulatory and integrity bodies. Staffing a body Parliament has legislated into existence is not discretionary in any meaningful sense, and criticising the resulting headcount without acknowledging the legislation that required it would be unfair.
The risk to the future
A permanently larger wage base narrows the room to move in the next downturn
Every recession arrives with the same demand: find savings quickly, without cutting the things that protect people at the worst possible moment. What a treasurer can reach for in that moment is not the whole budget. It is the discretionary residue — programs that can be paused, contracts that can lapse, engagements that can end when they end. External labour sat squarely inside that residue, which is exactly why it was cut first, and exactly what made it useful. Converting a multi-billion-dollar annual flow from that residue into permanent, indexed, superannuation-bearing employment does not make the spending wrong. It makes it immovable. Combine that with the categories already beyond quick adjustment — legislated payments, the NDIS, health and aged care rising with an ageing population, defence commitments running to the 2040s, and a public debt interest bill growing faster than almost anything else in the budget — and the share of Commonwealth spending a government can actually adjust within a single year keeps shrinking. The risk is not that the public service is unaffordable today; on current projections it plainly is affordable. The risk is compositional and it compounds: when the next shock comes, and one always does, the flexible spending that could have absorbed it will already have been spent on commitments that cannot be. At that point the options narrow to the ones no government wants — higher taxes, deeper cuts to services that are harder to cut, or more borrowing at an interest rate the next generation pays. None of that is a reason not to have staffed the veterans' claims queue. It is a reason to state the true price of the decision, to prove the offsetting saving was real, and to be honest that the flexibility being traded away is not recoverable at the moment it is most needed.