1. Introduction
Australia is working harder and producing less for the effort. Labour productivity fell by 0.6 per cent in the March quarter of 2026 and grew by only 0.3 per cent over the year to March, while hours worked rose 2.2 per cent over the same period. The Productivity Commission’s reading of that arithmetic is blunt: output is rising, but not as fast as the hours poured into it. The Reserve Bank downgraded its 2026 labour productivity expectation from 0.2 per cent growth to a fall of 0.5 per cent. Productivity in the non-market sector which is where most government service delivery sits, has been consistently weaker than in the market sector.
This is not a story about effort. It is a story about how work is organised.
Set against that flat productivity line is an ICT cost base that has grown without interruption. Gartner’s most recent forecast puts worldwide IT spending at 6.37 trillion US dollars in 2026, up 14.2 per cent on 2025 and the forecast has been revised upward three times inside twelve months. Australian organisations, public and private, are participating in that increase.
The question that matters for a chief executive or an agency head is what share of that spending is buying capability and what share is buying the right to keep operating.
The evidence suggests a large share is the latter.
McKinsey’s research with chief information officers found that technical debt represents between 20 and 40 per cent of the value of an entire technology estate before depreciation, that between 10 and 20 per cent of the budget nominally allocated to new products is diverted to resolving debt-related issues and that around 30 per cent of surveyed CIOs believe more than 20 per cent of that new-product budget is consumed this way.
The Australian public sector has already paid a well-documented price for this. GovERP was intended to consolidate the corporate systems of around 100 entities onto a single platform. The Department of Finance reported its cost at AUD344.7 million over four years. The program was halted in late 2023 and the independent review that followed found no components that could be used as they stood by departments without further fitting work.
The minister’s own characterisation of the underlying assumption was that it had been naive to think complex corporate systems could be standardised across that many entities.
That failure is usually read as a procurement failure or a program management failure. It is more accurately a category failure. The program attempted to impose a single System of Record on organisations whose operating logic, the actual sequence of decisions, approvals, evidence and accountabilities by which each agency does its work, was never the same and was never going to be.
Customisation is what happens when operating logic is forced into a system that was not built to hold it. Every customisation becomes an integration. Every integration becomes a dependency. Every dependency becomes a cost that recurs annually and delivers nothing to output.
There is a measurable gap here that the technology budget does not address. A survey of 35,000 US manufacturing plants across two waves with the US Census Bureau, found that structured management practices account for more than 20 per cent of the variation in productivity between plants, a share similar to or greater than that accounted for by research and development, ICT or human capital. They also found that 40 per cent of the variation in management practice occurred between plants inside the same firm.
The same company, the same technology, the same capital and materially different productivity, explained by how the work was structured.
That finding reframes the spending problem. If how work is defined, executed and governed carries productivity weight comparable to the technology itself, then an organisation that invests continuously in the System of Record while leaving its operating logic undocumented, inconsistent and resident in individual heads is funding one factor of production and neglecting an equally significant one.
The discipline that already knows how to close that gap is quality. It has known since the 1940s, when the proposition was tested at national scale under the most severe constraint an industrial economy can face and it has the measurement apparatus to prove the result. It has also been applied to manufacturing far more rigorously than it has ever been applied to the administrative and decision-making work that consumes most of the cost base of a modern agency or enterprise and its own institutional history explains why that application stalled.
2. Quality as a Scaling Mechanism: 1942 and 1950
Quality is remembered as a manufacturing refinement. It entered the modern economy as a mechanism for scaling national industrial capacity under emergency conditions.
When the United States converted its civilian economy to war production, the binding constraint was not factory floor space or raw material. It was the problem of getting an enormous number of newly recruited and largely untrained workers, in firms that had never produced military equipment, to make output that conformed to specification at volume, so that ammunition functioned, radios worked and components from one supplier fitted assemblies from another. Inspection could not solve that problem. Inspecting a defective national output stream would have identified the shortfall more precisely and produced nothing additional.
The response was to build the capability into the process. At the request of the War Department, the American Standards Association issued American War Standards Z1.1 and Z1.2 in 1941 and Z1.3 in 1942, defining American control chart practice and drawing on the statistical methods Walter Shewhart had developed at Bell Telephone Laboratories from the 1920s. Those standards became the teaching material for a national training program. The first intensive course ran at Stanford in July 1942 for representatives of the war industries and the procurement agencies of the armed services, developed by Holbrook Working, Eugene Grant and W. Edwards Deming. The Office of Production Research and Development, together with the Engineering, Science and Management War Training program in the Office of Education, then commissioned Working to disseminate the courses nationally and eight-day versions ran in many localities through the mid-1940s. Joseph Juran, working within the Lend-Lease Administration, was among those training civilian suppliers. Trainees at these sites formed local societies that later consolidated into the American Society for Quality Control.
That is what scaling an industrial base under constraint actually required. A defined method, taught to the people performing the work, applied at the point of production, with variation controlled rather than sorted out afterward.
Then it lapsed. With postwar demand for American manufactured goods exceeding supply, most statistical quality control programs were terminated once wartime contracts ended. Deming’s documented frustration with American management dates from that period. The lesson is worth stating plainly because it recurs: when there is no competitive pressure and no external scrutiny, process discipline is the first capability an organisation stops funding and it stops funding it precisely because the discipline was never structurally embedded in how the work ran.
Japan had no such luxury. Its prewar manufacturing reputation was for cheap goods of poor quality and its industrial base had been destroyed.
The Union of Japanese Scientists and Engineers was founded in 1946 with the objective of rebuilding the economy through the systematic application of science and engineering. JUSE invited Deming, who arrived in mid-1950 under the auspices of the Supreme Commander of the Allied Powers and delivered an eight-day course on quality control in Tokyo, followed by a one-day course on quality control for top management at Hakone. Across sixty-eight days in Japan that year he gave ten lectures, three training courses and the top management course and he returned during 1951 and 1952. Juran followed at JUSE’s invitation, lecturing to senior executives and middle managers in 1954 on managing for quality. Deming donated the royalties from the published lecture transcripts to JUSE, which used them to establish the Deming Prize in 1951.
Two features of the Japanese application matter more than the statistical technique. The first is that it was taught to top management directly. Deming’s Hakone course existed because a method taught only to engineers does not survive contact with the way a company is actually run. The second is that the Japanese approach moved past inspection of finished product toward improvement of every organisational process through the people performing it, which is the distinction between quality control and quality management set out in Section 5 below.
The economic result is a matter of record. Japanese manufacturers took share from US industry in automobiles and electronics through the 1970s on the combination of higher quality and lower price and the initial American response was to attribute the gap to price and to seek relief through domestic cost reduction and import restriction rather than through method. Nissan won the Deming Prize in 1960. Toyota adopted Total Quality Control in 1961 and won the prize in 1965.
Two nations, one scaling under emergency and one rebuilding from destruction, both achieved the outcome by making quality a property of how work was performed rather than a check applied to what the work produced. In both cases the method was owned by management. In neither case was it certified by anybody.
3. How the Discipline was Hollowed Out
The trajectory from that position to the present one is instructive, because the mechanism of decline is the same mechanism this paper is arguing against.
ISO 9000 was published in 1987, drawing on British Standard 5750 and diffused rapidly. The diffusion was driven substantially by procurement. Government buyers and large corporates began requiring suppliers to hold certification, which converted the standard from a method an organisation might adopt into an entry ticket an organisation had to purchase. Once that happened, the commercial ecosystem reorganised around the ticket. Consultants sold preparation and documentation.
Certification bodies sold assessment and recurring surveillance. Organisations bought a certificate. Underneath all of it, in a great many cases, the work continued to run exactly as it had run before.
The research literature documents this directly. Heras-Saizarbitoria and Boiral, working from sixty-five in-depth interviews with general managers, middle managers and employees across eight Spanish firms that had held the standard over long periods, found systematic decoupling between the rhetoric of ISO 9000 and its practice and concluded that adoption frequently functioned as the acquisition of an organisational credential whose optimistic language masked wide variation in what was actually done. Christmann and Taylor examined the same decoupling between implementation quality and certification status in the related environmental standard. Naveh and Marcus found that performance improvement occurred where the standard was assimilated into daily practice and built upon and did not occur where it was merely installed. Swift, Humphrey and Gor examined the financial legacy of quality audits and found the returns considerably weaker than the promise. John Seddon’s critique, published as ten arguments against ISO 9000 in 1997 and expanded into a book, made the operational case: the documentation requirement creates two jobs where there was one, the work itself and the writing about the work performed so that an assessor can verify it; the fear of being found non-conforming during assessment encourages concealment rather than correction; and quality becomes defined by an external assessor rather than by the customer.
On the question of who benefited, the honest formulation is structural rather than personal. I cannot substantiate a general claim about the motives or integrity of individual auditors and asserting one in a paper of this kind would weaken it in front of exactly the audience it is written for. The structural version is stronger and harder to argue with. The certification body is paid by the organisation it certifies. The revenue recurs through surveillance audits and is contingent on the relationship continuing.
The consultant who prepares the organisation and the assessor who evaluates it operate in the same commercial market and the standard governing certification bodies, ISO/IEC 17021-1, exists in part to prohibit the same body from doing both, which is itself evidence that the conflict was recognised as material. Accreditation case reporting continues to identify instances of auditors assessing clients they had previously advised. Where the assessed party pays the assessor and where the assessed party’s objective is to pass rather than to improve, the equilibrium is documentation. That equilibrium does not require anyone to behave badly. It only requires everyone to behave rationally.
The consequence for the organisation is the one that matters here. A certificate is inspected-in assurance applied at the level of the whole enterprise. It is an opinion, formed by an external party, about a sample of evidence, at a point in time, concerning a process that may or may not be the process by which the work is actually done. It carries all of the weaknesses Deming identified in mass inspection of product and it carries them at a higher level of aggregation, which makes them harder to detect.
The discipline did not fail. Its delivery mechanism was commoditised and the delivery mechanism was inspection.
4. What Quality Costs When it is Absent
Quality is generally treated as an expense line. The literature treats it as a recovery. The prevention-appraisal-failure model divides the total cost of quality into four categories. Prevention costs are incurred to stop defects occurring: process design, training, standard work, control planning. Appraisal costs are incurred to find defects that have already occurred: inspection, testing, review, audit. Internal failure costs are the cost of defects caught before the customer sees them: rework, scrap, delay, escalation. External failure costs are the cost of defects the customer does see: remediation, penalty, complaint handling, reputational loss, regulatory intervention.
The American Society for Quality has long estimated that total quality-related costs run between 10 and 20 per cent of revenue in typical organisations, rising toward 40 per cent in poorly performing ones, with mature operations holding total cost of quality below 5 per cent. Crosby estimated the price of nonconformance at around 20 per cent of sales. Juran observed that appraisal and failure costs in most organisations substantially exceed prevention costs. In organisations without a developed quality capability, the historical distribution of quality spending has been roughly 60 to 70 per cent on failure, 20 to 25 per cent on appraisal and 5 to 10 per cent on prevention.
Those percentages carry a caveat worth stating directly. They are estimates drawn largely from manufacturing, the underlying measurement varies between studies and much of the cost is hidden in cost centres that are reviewed separately and never aggregated. I cannot confirm a precise cost-of-quality figure for any particular Australian agency or enterprise and no one should represent these ranges as an audited number. What the ranges establish is the order of magnitude and, more importantly, the shape of the distribution. Most organisations spend the large majority of their quality budget on finding and fixing failure and a small minority on preventing it.
That distribution is the productivity problem stated in different language. Rework is work that produces no additional output. Inspection is work that produces no additional output. Escalation, reconciliation, exception handling, the meeting convened to work out why two teams did the same thing differently, the report reconstructed because the evidence trail was never captured at the time: all of this consumes hours and adds nothing to the numerator. An organisation with a 15 per cent cost of poor quality is running its productivity denominator 15 per cent higher than it needs to be.
The economics of the shift are asymmetric in a way that matters. Money moved from failure into prevention does not trade one cost for an equivalent cost. A defect prevented costs the price of the control that prevented it. The same defect caught at inspection costs the control plus the detection plus the rework. The same defect reaching the customer costs all of that plus the remediation, the relationship and, in a regulated or scrutinised environment, the exposure. The ratio is commonly expressed as one to ten to one hundred. That ratio is a heuristic rather than a measured constant and its precise value differs by industry, but the direction is not in dispute across the quality literature: the cost of a defect escalates sharply with the distance it travels before it is caught.
For organisations under public scrutiny the escalation is steeper still, because the external failure cost is not bounded by the transaction. It is bounded by the inquiry.
5. Quality Control and Quality Management are not the Same Discipline
The distinction is frequently collapsed and collapsing it is expensive.
Quality control is detection. It operates on output. It samples, inspects, tests and reviews work that has already been performed and it separates what conforms from what does not. It is reactive by design. Its cost sits in the appraisal category and the defects it finds convert into internal failure cost.
Quality management is prevention. It operates on the process rather than the output. It defines how work is to be performed, establishes the conditions under which conforming output is the expected result, controls variation in those conditions and treats a defect as evidence that the process requires correction rather than as evidence that a person requires correction.
The economic consequence follows directly. Quality control can improve what reaches the customer, but it cannot reduce what the organisation spends producing it. Every defect it finds has already consumed the labour, the materials and the time that went into producing it. Increasing inspection increases appraisal cost while leaving failure cost substantially intact. Improving the process reduces both.
This is the argument behind the third of Deming’s fourteen points. His instruction was to “cease dependence on inspection to achieve quality,” and to eliminate the need for mass inspection by building quality into the product from the outset (Deming, Out of the Crisis, 1982). The point is regularly misread as an argument against inspection. It is an argument against dependence on inspection. Deming’s position was that by the time the work is inspected, the quality is already in it, good or bad and the inspection changes nothing except which of two costs the organisation pays.
The operational expression of that principle is quality at the source. Toyota’s jidoka concept gives the process the ability to halt itself when an abnormal condition arises, so that a defect is contained where it is created rather than propagated downstream. Shingo’s poka-yoke, mistake-proofing, designs the process so that the error is difficult or impossible to make, which is prevention embedded in the mechanism rather than imposed by supervision.
In both cases the assurance is a property of how the work is performed. It is not a separate activity performed afterward by someone else.
Assurance built in and assurance inspected in produce different cost structures, different cycle times and different evidence. Inspected-in assurance produces an opinion about a sample. Built-in assurance produces a record of every instance.
6. Repeatability, Consistency and the Cost of Variation
Quality is often defined as conformance to requirement. The operational definition that matters more for productivity is the reduction of variation. A process that produces an acceptable average while varying widely around it is not a controlled process. It is a process that generates a stream of exceptions and exceptions are where cost accumulates.
Variation in administrative and service work is systematically under-measured because it is rarely visible as a defect. Two officers assess comparable cases and reach different conclusions. Two teams run the same approval with different evidence thresholds. The same report is produced three times in three formats because no definition of the report exists. A capable individual departs and the process departs with them. None of these appear in a defect register. All of them consume hours and all of them generate downstream rework, reconciliation and, eventually, audit findings.
Repeatability is the property that eliminates this class of cost. When a process is defined once and executed the same way each time, the organisation stops paying for the redesign of work that has already been designed. Consistency of product and service is the external face of the same property. It is what allows a client, a citizen or a regulator to rely on the outcome and reliance is what removes the need for verification at every interaction.
The evidence that structured, consistent process produces measurable outcome improvement extends well outside manufacturing. Haynes and colleagues, reporting in the New England Journal of Medicine in 2009, studied the introduction of a nineteen-item surgical safety checklist across eight hospitals in eight countries. Thirty-day mortality fell from 1.5 per cent to 0.8 per cent and the complication rate fell from 11 per cent to 7 per cent. The intervention added no equipment, no staff and no technology. It defined the sequence of the work and made the definition binding.
Two qualifications should accompany that result and stating them strengthens rather than weakens the argument. A later study of checklist introduction across Ontario, published in the same journal in 2014, found no statistically significant reduction in operative mortality or complications. Subsequent work, including a Norwegian study, identified a dose effect in which the improvement scaled with how completely the checklist was actually followed. The conclusion supported by the whole body of evidence is not that documenting a process improves outcomes. It is that a process which is genuinely and consistently executed improves outcomes and one that is documented but inconsistently executed does not.
The same qualification applies to the research on formal quality standards. Studies of ISO 9001 certification and firm performance produce mixed results. Some longitudinal work finds improvements in productive efficiency and financial performance among certified firms and some finds the effect weak, heterogeneous or attributable to selection rather than to certification. The most defensible reading of that literature is that certification is a signal, not a mechanism. Organisations that use the standard to change how work is actually performed obtain benefit. Organisations that treat it as a document produced for an auditor obtain a document.
This is the central practical insight for anyone considering how to lift productivity in a complex organisation. The benefit does not come from writing the process down. It comes from the process being the way the work actually runs, every time, with the evidence generated as a by-product of doing the work rather than assembled afterward to satisfy someone asking about it.
That is a systems problem and it is not one that a System of Record was ever designed to solve.
7. A System Rebuilt with DOLIUM – but Better
DOLIUM is a System of Work. It holds an organisation’s operating logic, the institutional intellectual property describing how work is defined, executed and governed, in a layer that sits above the System of Record rather than inside it.
The following describes what that architecture is designed to deliver. These are design propositions grounded in the quality principles set out above and they should be read as such rather than as measured outcomes from a deployed baseline.
Operating logic is owned by the organisation, not by the vendor of the record system
The GovERP experience demonstrated the cost of the alternative. When the way an agency works is expressible only as customisation of a packaged platform, the agency’s process knowledge becomes a liability held in someone else’s product and it is re-purchased at every upgrade, every migration and every vendor negotiation.
Separating operating logic from the System of Record allows a common record backbone to deliver its economies of scale while each entity retains and controls its own way of working. The standardisation debate that GovERP lost is one that does not need to be fought at the record layer.
Assurance is generated by execution rather than reconstructed for inspection
Because work is defined and executed within the same structure, the evidence of what was done, by whom, under what authority and against what standard exists as a consequence of the work having been performed. This is the built-in position rather than the inspected-in position. Its immediate effect is on appraisal cost, which in most organisations is a permanent overhead carried by internal audit, assurance teams, compliance functions and the executive time consumed by preparing for external review. Its second effect is on external failure cost, because controls that operate at the point of execution catch conditions before they leave the organisation.
The evidence is continuous rather than sampled
The failure mode described in Section 3 arises because a certificate is an opinion about a sample, formed by a party with a commercial interest in the relationship continuing, concerning a documented process that may differ from the operating one. When the defined process is the executing process, that gap closes. Conformance is observable continuously and directly by the organisation’s own executives, by its regulator and by its auditor and the question shifts from whether the documentation is adequate to whether the work is conforming. An assessment performed against a live record of execution is a different exercise from an assessment performed against a manual.
Repeatability is structural
A process defined once in DOLIUM executes the same way each time it runs and it executes the same way across teams, sites and business units. The variation that generates rework, exception handling and inconsistent service is addressed at the point where it originates. The Bloom finding that 40 per cent of management-practice variation occurs between plants inside the same firm describes exactly the condition this addresses.
Process knowledge survives its authors
Operating logic held in a structured layer does not resign, retire or move agencies. This removes a specific and expensive dependency in the Australian public sector, where continuity of process understanding is frequently maintained through external consultants who are re-engaged to rediscover how an organisation’s own work is performed.
Reuse across entities without re-implementation
Operating logic defined for one process, one agency or one regulatory obligation can be adapted rather than rebuilt. The cost curve for the second implementation is materially different from the first, which is the property that whole-of-government programs have historically sought and failed to obtain by standardising the record layer instead of the work layer.
Accountability structure for AI at scale
As organisations deploy AI agents in operational volume, the governing question becomes what each agent is doing, under whose authority and to what standard. This is the same question quality management has always asked about human work and it requires the same answer: a defined process, a defined standard, a defined authority and evidence generated at execution. An organisation that cannot answer that question about its people will not be able to answer it about its agents. The accountability framework that makes responsible AI adoption possible at scale is the same framework that makes conforming work possible at scale.
8. Conclusion
The productivity data and the technology spending data point in opposite directions and they have done so for several years.
Organisations are increasing investment in the System of Record while output per hour stalls. The missing factor is not more technology. It is the structure of the work that the technology is supposed to support and the evidence that this structure carries productivity weight comparable to ICT investment itself is now well established.
Quality management supplies the mechanism and the measurement. The cost of poor quality is real, large and largely invisible because it is distributed across cost centres that are never aggregated. It is dominated by failure and appraisal, which means it is dominated by the cost of finding and fixing work that was performed incorrectly. Shifting that expenditure into prevention reduces total cost rather than relocating it, because prevention removes the failure rather than detecting it.
Doing that requires the operating logic of the organisation to be explicit, executable, consistent and evidence-generating. Written procedures do not achieve this, as the checklist and certification literature demonstrates clearly. The process has to be the way the work runs.
The historical record settles the question of whether this is achievable. It was achieved across an entire national industrial base in 1942 under conditions considerably more constrained than any Australian organisation faces today and it was achieved again in the rebuilding of Japanese industry from 1950.
In both cases quality was owned by management, taught to the people doing the work and executed at the point of production. What was lost between then and now was not the knowledge. It was the delivery mechanism, which was commoditised into a certificate and thereby converted back into the inspection model the original discipline had been designed to replace.
A System of Work is the layer that makes that possible. DOLIUM is built to hold it, to keep it sovereign to the organisation that owns it and to produce assurance as a property of execution rather than as the output of an inspection performed after the fact.