The Migration of Value: Why the complements, not the technology, decide who wins

John H. Howard[1], 4 August 2026

This working paper is one of a series prepared by the Acton Institute for Policy Research and Innovation to inform the UTS research project Turning AI into Productivity: The Role of Innovation Ecosystems, supported by the Google Foundation. The project's final report will be launched on 22 September 2026 by Senator the Hon Tim Ayres, Minister for Industry and Innovation.

Technological progress rarely eliminates human value. It moves it. Each wave of innovation makes some capabilities abundant and cheap while making others scarce and valuable.

The value of a general-purpose technology does not sit in the technology itself. It sits in the complements that give the raw capability productive effect: data, skills, redesigned workflows, governance, judgement, relationships and business models.

Because the technology can be bought by anyone, it earns commodity returns. The complements are firm-specific, built over time and hard to copy, and that is where value gathers. The corporate record is emphatic. EMI invented the CT scanner and lost the market to GE. Xerox invented the personal computer interface and watched Apple and Microsoft capture it. Kodak invented the digital camera and went bankrupt. Apple, Virgin, Toyota and Walmart built their positions on complements rather than on technology they invented or owned.

AI repeats the pattern. It makes competent expression and routine analysis abundant, lifting the value of judgement, problem framing and accountability. Firms that buy the model but neglect the complements may see little return. Migration is not redistribution: the people who lose value when a capability becomes abundant are rarely the people who capture it elsewhere. For policy, the implication is to give the complements the same attention as the technology, through education, workforce development, organisational capability and institutional adaptation.

---

Introduction: an old pattern, newly visible

Artificial intelligence has prompted a wave of anxiety about the future of human work. This Insight treats that anxiety as the latest expression of a recurring process rather than a break with the past. The pattern can be stated simply.

Technological progress rarely eliminates human value. It changes where that value resides. Each wave of innovation reduces the scarcity of some capabilities and raises the scarcity, and therefore the value, of others.

Earlier transitions displayed this movement, but its logic was easy to overlook at the time. AI makes it visible because the change is so fast and so broad. Within a few years, competent written analysis, once the mark of an educated professional, has become something a machine produces in seconds, across almost every field at once.

That speed draws a sharp line between what the machines now supply in abundance and what remains scarce. This Insight sets out the pattern as a framework, shows why the value created by a new technology flows to its complements rather than to the technology itself, and examines which firms have understood this and which have not.

The substitution story and what it misses

Most discussion of technological change is framed around substitution. Machines replace workers, automation replaces labour, and AI replaces knowledge workers. Every major transition has produced this fear. Mechanisation threatened artisans, industrial machinery threatened manual workers, computers threatened clerks, and the internet threatened intermediaries of every kind.

In each episode the fear was real and the disruption was genuine. The historical record is nevertheless more nuanced. The calculator did not remove accountants, the spreadsheet did not remove financial analysts, and the internet did not remove the need for experts. What each technology did was change the location of value.

Some capabilities became less valuable because they became abundant. Others became more valuable because they remained scarce. Something similar may now be occurring within organisations: the capacity to produce strategy documents is becoming abundant, while the capacity to implement strategy remains stubbornly scarce. The substitution story is not wrong, only incomplete. It records what disappears and stops there, before the value has finished moving.

The core proposition: value migrates

Economists have long connected value to scarcity, and the same principle applies to human capability. When a capability becomes abundant, the price attached to it falls. When a capability stays scarce, its value rises. Technological change continually rearranges which capabilities are abundant and which are scarce.

Before calculators were widespread, mental arithmetic was a prized skill. It remains useful, but its scarcity has largely gone, and value migrated towards activities calculators could not perform: modelling, interpretation and problem formulation. The internet produced a similar movement. Information became cheap, and the judgement needed to evaluate it became costly.

The cycle repeats with each wave of technology, and it can be drawn as a loop (Figure 1). A new technology makes an established capability abundant, its price falls, value migrates towards what remains scarce, and complementary investment determines who captures it. Institutions then adjust, or lag, until the next wave begins.

Figure 1: The value migration cycle

The framework in five propositions

The pattern can be condensed into five linked propositions, summarised in Table 1. Together they form the analytical core of the framework, and each is illustrated by both the historical record and the current AI transition.

Table 1: The migration of value in five propositions

	Proposition	Historical pattern	The AI instance
1	Technology makes established capabilities abundant	Printing made written knowledge abundant; mechanisation, physical effort; computing, calculation; the internet, information	AI is making competent expression and routine analysis abundant
2	Value migrates towards what remains scarce	The calculator lowered the value of arithmetic and raised the value of mathematical reasoning	The value of polished prose falls; the value of judgement, synthesis and ethical reasoning rises
3	Complementary investment decides who captures the value	Electricity required factory redesign; computers required new workflows; the internet, new business models	AI requires investment in data quality, skills, workflow redesign, governance and leadership
4	Institutions lag because they are organised around earlier scarcities	Education and accreditation systems continue to reward skills whose scarcity has declined	Organisations optimise or automate established processes rather than redesign them
5	Innovation is the early recognition of emerging scarcity	Entrepreneurs spot unmet needs; researchers, unanswered questions; investors, undervalued assets	Innovation ecosystems act as systems for detecting where value is moving before others do

Why the value sits in the complements

Proposition three carries the weight of the argument and deserves close attention, because it is the proposition that businesses most often miss. A general-purpose technology, on its own, creates almost no value for the firm that buys it. The reason is straightforward: anyone can buy it.

A capability available to everyone at market price cannot be a source of advantage for anyone. The value lies in the complements that surround it, because the complements are firm-specific and hard to copy.

Paul David (1990) made the point with electricity. The dynamo delivered no productivity dividend for nearly four decades, because the gains required factories to be rebuilt around the new power source: single-storey layouts, unit drive, reorganised workflows and retrained workers. Firms that simply swapped a steam engine for an electric motor gained almost nothing.

David Teece (1986) generalised the insight. Innovators frequently fail to profit from their own inventions because value flows to whoever holds the complementary assets: manufacturing capacity, distribution channels, service networks, brand and customer relationships. The invention can be imitated or licensed. The complements cannot, at least not quickly.

Figure 2 applies the logic to AI. The model at the centre is available to every competitor at a subscription price, and it earns commodity returns accordingly. The ring of complements around it, proprietary data, tacit skills, redesigned workflows, governance and trust, human judgement, and customer relationships, is where value accrues and where advantage is built.

Figure 2: Value accrues in the complements, not the technology

This may explain why aggregate productivity gains from AI have so far been modest. Brynjolfsson, Rock and Syverson (2021) describe a productivity J-curve in which measured output flattens early while firms invest in intangible complements that take years to pay off. Two firms running the same model can obtain very different results, because they differ in everything that surrounds it.

The firms that understood, and the firms that did not

Few businesses fully grasp the principle of complementarity, and the corporate record shows both sides of the ledger with unusual clarity. The pattern is consistent enough to be stated as a rule of thumb: the inventor of a technology and the captor of its value are often different firms, and what separates them is the complements.

Inventors who lost the value

EMI invented the CT scanner, an achievement that earned Godfrey Hounsfield a Nobel Prize. Within a decade EMI had exited the medical imaging business. GE and Siemens captured the market because they held the complements hospitals actually buy: sales forces, service networks, manufacturing scale and clinical reputation. Teece built his theory on this case.

Xerox invented much of modern computing at its Palo Alto Research Center, including the graphical interface, the mouse and Ethernet. It captured the value of only one invention, the laser printer, because that was the one that fitted its existing complements: a copier sales force and a document-centred customer base. Apple and Microsoft captured the rest.

Kodak invented the digital camera in 1975 and understood the technology better than anyone. Its complements, film chemistry, processing networks and retail relationships, were assets of the old scarcity and liabilities in the new one. The firm that owned the invention filed for bankruptcy; the value migrated to smartphone makers.

Nokia offers a more recent case. It led the world in handset technology and manufacturing, but the value of the smartphone migrated to the developer ecosystems and app marketplaces built by Apple and Google. Superior hardware could not compensate for missing complements.

Complement builders who captured it

Apple entered the MP3 player market three years after the pioneers. The iPod won because Apple built the complements the technology needed to become useful: iTunes software, licensing agreements with the music labels, a 99-cent store, industrial design and retail presence. The iPhone repeated the play at greater scale. The handset could be imitated; the App Store ecosystem, developer community and customer relationships could not.

Virgin is perhaps the purest case, because it owned almost no technology at all. Its assets were complements: brand, customer experience design and marketing capability. It carried them across records, airlines, telecommunications and finance, entering industries where incumbents held the technology and capturing value they had left on the table. Virgin Atlantic flew the same Boeing aircraft as its competitors. The value sat in everything around the aircraft.

Toyota used the same machine tools available to General Motors. Its advantage was the Toyota Production System, an organisational complement built from supplier relationships, problem-solving routines and accumulated tacit knowledge. Even after the system was documented and taught in business schools, competitors took decades to approach it, because tacit and organisational complements resist copying.

Walmart adopted barcode scanning and electronic data interchange when every retailer could buy them. It captured the value because it built the complements around the technology: cross-docking logistics, supplier integration and a data-driven replenishment culture that competitors could observe but not reproduce.

The Australian evidence

The pattern holds in Australia. Cochlear commercialised Graeme Clark's implant research, but its durable advantage lies in the complements: surgeon training networks, clinical evidence, precision manufacturing and lifetime customer relationships. CSL similarly built its position on a global plasma collection network and regulatory capability that rivals cannot quickly assemble. In both cases, the science could travel; the complements could not.

Table 2: Inventors and captors: the complements decide

Technology	Inventor or pioneer	Captor of value	Decisive complements
CT scanner	EMI	GE, Siemens	Sales, service, clinical trust
Graphical computing	Xerox PARC	Apple, Microsoft	Product design, distribution, developers
Digital photography	Kodak	Smartphone makers	Platforms, software ecosystems
Digital music players	Rio, Creative	Apple	iTunes, licensing, design, retail
Retail IT (barcodes, EDI)	Available to all	Walmart	Logistics, supplier integration, data culture
General-purpose AI	Model developers	Undecided	Data, skills, workflows, governance, judgement

The final row of Table 2 states the live question. The developers of general-purpose AI models hold a position closer to EMI than is commonly assumed: the capability they sell is available to every competitor, and its price is falling. Who captures the value of AI remains undecided, and the answer will turn on who builds the complements.

The AI era and the new scarcity

AI offers a clear illustration of the wider framework. Its most visible capability is the production of language. Reports, briefing notes, presentations and business cases can now be generated at a standard that would have seemed remarkable a few years ago. AI is reducing the scarcity of competent expression.

This may explain why some experienced writers feel uneasy. The difficulty is not that AI produces poor prose. The difficulty is that good prose is becoming abundant, and when a capability becomes abundant, value migrates elsewhere. The question worth asking is where it goes.

The answer lies in the distinction between expression and judgement. Professional life often conflates the two: a well-written report appears authoritative and a polished presentation appears persuasive. Expertise has always involved more than communication. The public servant who anticipates an implementation failure, the entrepreneur who sees a market opening and the researcher who frames a significant question are all exercising judgement, and judgement is becoming the scarcer capability.

Figure 3 places the AI transition in its historical sequence. Each technology pushed one capability into abundance and lifted another into scarcity. The pattern that took decades to become visible in earlier waves is unfolding in years this time.

Figure 3: Five waves of abundance and scarcity

What remains scarce: three distinctions

The claim that value migrates towards scarce capabilities invites an obvious question. Which capabilities remain scarce, and why? Three distinctions give the answer analytical content.

Codified and tacit knowledge. AI excels at knowledge that has been recorded, structured and made available in text, code or data. Tacit knowledge, acquired through practice and reflection, resists this treatment (Polanyi, 1966). It tends to remain with experienced people and may gain value as codified knowledge becomes abundant.

Convergent and divergent tasks. Current AI systems are strong at convergent tasks, which narrow towards a known answer by identifying established patterns at scale. Divergent work, framing new questions, combining ideas across fields and challenging accepted assumptions, may remain a largely human domain for some time, though the boundary is contested and moving.

Analysis and judgement under uncertainty. Analysis processes available information. Judgement decides what to do when information is incomplete, objectives conflict and consequences are uncertain (Simon, 1947). AI can strengthen analysis, but choosing under uncertainty, and accepting responsibility for the choice, remains a human function.

None of these boundaries is fixed. Each new model shifts the frontier, and capabilities once thought distinctively human have repeatedly been absorbed by machines. The framework does not claim a permanent human reservation. It claims that, at any given moment, a scarce margin exists, and that value gathers there until the next advance moves it again.

From general-purpose capability to applied value

A sharper account separates two forms of AI. General-purpose AI is the raw capability in the abstract, indifferent to any use. Applied AI is that same capability once an economy has built the complements that give it productive character. The capability can be bought. The application has to be made, and value migrates through the work of conversion, along several paths at once.

Industrial AI acts on matter and physical systems across factories, supply chains, energy and logistics. Its complements include sensors, operational data, robotics, simulation and process redesign. Value migrates towards those able to integrate AI into physical operations.

Cognitive or professional AI acts on documents, analysis and decisions. Its complements include workflow redesign, data quality, professional judgement and clear accountability. Value migrates from competent expression towards the judgement that decides what to do with it.

Scientific or discovery AI acts on hypotheses and experiments. Its complements include instruments, datasets and research design. Value migrates towards the framing of significant questions and the interpretation of results.

Civic or service AI acts on public administration and service delivery. Its complements include institutional capability, data governance and public trust. Value migrates towards the judgement, legitimacy and care that public decisions require.

Consumer AI sits apart as a vector of diffusion rather than a domain of production. Its main effect on the migration of value may be indirect, accelerating the pace at which the four productive registers are obliged to adapt.

Because the conversion involves four distinct tasks rather than one, each with its own complements, a national approach may need to be plural rather than a single program, and the regions and sectors least able to supply the complements risk being left behind.

Migration is not redistribution

A framework built on the movement of value can be read too comfortably. The fact that value migrates does not mean it migrates to the same people who lose it. The clerk displaced by the spreadsheet does not automatically become a financial analyst. The displaced worker and the eventual beneficiary are often different people, in different places, with different prospects.

Timing compounds the problem. New sources of value can take years to develop, while displacement can arrive quickly. The gap between the two is where transition costs fall, and it is rarely shared evenly. Complementary investment is therefore as much a question of equity as of efficiency. Without access to retraining, capable institutions and quality local innovation systems, value may migrate while opportunity does not.

Where the framework might fail

A general framework is strengthened by acknowledging its limits, and three objections deserve consideration. The most serious holds that AI differs in kind from earlier technologies: previous tools automated specific tasks, while general-purpose AI may automate the general capability of analysis itself. If that proves correct, the scarce margin could narrow faster than new sources of value emerge.

A related risk is that judgement is also codified. As systems are trained on records of expert decisions, some judgement may itself become reproducible. The boundary between analysis and judgement may move further than expected, and the framework should be held as a working hypothesis rather than a settled conclusion.

A third objection concerns the assumption of orderly adjustment. History includes episodes, among them the early decades of industrialisation, in which displacement ran well ahead of compensating gains for a generation or more. The framework survives these objections, but with its claims trimmed. It is a guide to where attention and investment should go, not an assurance that adjustment will be painless. History suggests it often is not.

Implications

For business

The corporate cases carry a direct lesson. Buying the technology confers no advantage, because every competitor can buy it too. The strategic questions are about complements: what proprietary data, skills, workflow redesign, governance and customer relationships would make the capability productive here, and which of them can rivals not quickly copy? Firms that ask what the model can do are asking the wrong question. The better question is what the firm can do that the model makes more valuable.

For universities

Universities have long transmitted scarce knowledge. As knowledge itself becomes abundant, that role narrows. If competent prose is becoming abundant, the ability to write remains useful but no longer distinguishes a graduate; the ability to think becomes the decisive capability. Curriculum, assessment and accreditation that still reward the reproduction of codified knowledge may be rewarding a capability whose scarcity is in decline.

For public policy

Public policy rarely fails for want of information. More often it fails in interpretation, sequencing, prioritisation and implementation, which are matters of judgement. Much innovation policy focuses on supporting technologies. The migration of value suggests equal attention to the complementary capabilities that decide who captures the gains: education, workforce development, organisational capability and institutional adaptation.

Because institutions organise around earlier scarcities, a standing task of policy is to identify that lag and act on it.

For innovation ecosystems

Innovation, on this reading, is the early recognition of where value is moving. The most effective innovation ecosystems may be those that detect and respond to value migration earlier than competing locations. This connects the framework to the study of innovation districts, where complementary assets, institutions and relationships determine which places capture emerging value.

Conclusion: the new scarcities

The debate about AI tends to focus on what machines can do. History suggests a different question: what becomes more valuable because the machines exist? That question shifts attention from substitution to complementarity, from technology to capability, and from displacement to adaptation.

The corporate record adds a warning. The firms that owned the technology, EMI, Xerox, Kodak and Nokia, did not capture its value. The firms that built the complements, Apple, Virgin, Toyota and Walmart, did. General-purpose AI holds no productive identity until the complements supply one, and that moment of conferral is the moment value begins to migrate.

The useful question is not whether AI replaces people. It is where value moves when technology changes what is scarce, and which individuals, organisations and places are equipped to follow it.

References

Bresnahan, T. F., & Trajtenberg, M. (1995). General purpose technologies: 'Engines of growth'? Journal of Econometrics, 65(1), 83–108.

Brynjolfsson, E., Rock, D., & Syverson, C. (2021). The productivity J-curve: How intangibles complement general purpose technologies. American Economic Journal: Macroeconomics, 13(1), 333–372.

David, P. A. (1990). The dynamo and the computer: An historical perspective on the modern productivity paradox. American Economic Review, 80(2), 355–361.

Drucker, P. F. (1969). The age of discontinuity: Guidelines to our changing society. Harper & Row.

Howard, J. H. (2026). Making sense of AI in 2026. Acton Institute Publishing.

Polanyi, M. (1966). The tacit dimension. Doubleday.

Schumpeter, J. A. (1942). Capitalism, socialism and democracy. Harper & Brothers.

Simon, H. A. (1947). Administrative behavior: A study of decision-making processes in administrative organization. Macmillan.

Smith, A. (1776). An inquiry into the nature and causes of the wealth of nations. W. Strahan and T. Cadell.

Teece, D. J. (1986). Profiting from technological innovation: Implications for integration, collaboration, licensing and public policy. Research Policy, 15(6), 285–305.