Chapter 6 - Adjacent Professions: Programming and Law
6.1 What the adjacent professions show
This chapter examines two adjacent professions - software development and law - where AI mediation has been operating at scale for longer than in management consulting, and where the mechanism I name has visibly operated in ways that the analytical conversation has begun to register. The chapter draws on this comparative evidence to demonstrate that the proletarianisation mechanism operates beyond European infrastructure advisory, and to develop four findings I then deploy analytically in Chapter 7’s stage-by-stage analysis.
The chapter is interpretive of published evidence on adjacent professions, not grounded in lived material from my own practice. The deliberate choice was made because the comparative work the chapter does requires engagement with the empirical and theoretical literatures on AI in software and law, where the published evidence is richer than the comparable evidence on AI in consulting at the time of writing. The first-person voice is sparing here; the chapter operates in interpretive register.
The chapter is also where the framework engages its most analytical foil: the Susskinds’ The Future of the Professions. The Susskinds advance a substitutionist account of how AI mediation transforms the professions - predicting that professional services will increasingly be delivered through AI-mediated systems, with human practitioners reduced to narrower roles. The framework I deploy is formationist rather than substitutionist: it treats AI mediation as operating on the formation pathway through which professional expertise is reproduced, with consequences that operate at the level of formation rather than at the level of substitution. The distinction is important and the chapter develops it explicitly.
The chapter also engages the AI-washing critique that has become in the labour-market literature on the 2024-2026 tech sector layoffs. The critique is that some firms have framed cost-cutting as AI-driven displacement when the cost-cutting decisions were made for other reasons. The chapter takes this critique seriously and demonstrates that the framework’s argument addresses formation rather than headcount, with the consequence that AI-washing complicates labour-market data without undermining the framework’s claims.
The chapter closes with four findings synthesised at the chapter’s end: selective displacement, pipeline rupture regardless of headcount, uneven upward value flow, and the requirement of explicit narration of professional legitimacy. These four findings are then carried into Chapter 7’s stage-by-stage analysis, where they operate as analytical lenses applied within each ChMC stage.
6.2 The software case
The software development profession has been the canary for AI mediation in knowledge work, and the evidence from this case is the richest comparative material available to the framework. Three components of the evidence carry the chapter’s argument.
The first is empirical evidence of selective displacement at the level of individual practice. The published literature on AI mediation in software development demonstrates that AI tools - particularly the family of code-generation systems that became widely deployed from 2022 onward - have absorbed certain components of the developer’s work. Routine code generation, boilerplate production, basic test writing, code review at the level of style and convention, and what was historically the most repetitive component of software development have been mediated by AI tools in current practice. The displacement is selective: it operates on certain components of the work and leaves others unaffected.
The second is empirical evidence of pipeline disruption at the level of formation. The evidence from the software case demonstrates that the formation pathway through which junior developers have historically been formed - the engagement with routine code, the failed attempts and corrections, the gradual development of contributory expertise through sustained engagement with the work - is being structurally altered by AI mediation. Junior developers in current practice produce competent code faster, with the cost that the formative engagement with the failures and corrections is structurally compressed. The published evidence on this dimension is present in the literature on developer education, in the discourse of senior software engineers writing about their concerns for the next generation, and in the academic literature on the mechanism of expertise development in the software profession.
The third component is labour-market data with the AI-washing complication. The 2024-2026 tech sector layoffs have been analysed in the labour-market literature, with disagreement about how much of the displacement is attributable to AI mediation and how much is attributable to broader cost-cutting decisions framed in AI terms. The analytical conversation has registered both components: that AI mediation is producing real displacement in some segments of the software profession, and that some firms have framed cost-cutting as AI-driven for institutional reasons. The framework treats this complication as important: it means the labour-market data alone cannot carry the framework’s claims, and the framework’s argument must operate at the level of formation rather than headcount.
The consequence of the AI-washing complication is that the framework’s claim about the proletarianisation mechanism in software is not a claim about job losses. It is a claim about formation - about what is happening to the developer’s work and the formation pathway through which contributory expertise has historically been reproduced. The framework’s claim survives the AI-washing critique because the evidence on formation operates independently of the labour-market data: even if all the 2024-2026 layoffs were entirely cost-cutting and unrelated to AI displacement, the evidence on formation pathway disruption would remain operative.
6.3 The law case
The legal profession has been the second case where AI mediation has been operating at scale for substantially longer than in management consulting, and the evidence from this case complements the software case in ways I treat as important. The legal case is institutionally distinctive: law is a heavily regulated profession with chartered-status architecture (bar admission, seniority frameworks, ethical codes), with long-standing engagement with the question of how professional expertise is reproduced across generations.
The evidence from the legal case operates through three components.
The first is theoretical articulation of the formation problem. The legal profession has articulated the formation problem in terms that anticipate this thesis’s central concern. What has been called mediated evolution in the legal-AI literature [@watkin2025trajectories] - the structural alteration of the formation pathway through which junior lawyers develop contributory expertise - has been engaged in the published literature on AI in law. The articulation is richer than the equivalent articulation in software, partly because law has institutional architectures (bar admission requirements, mandatory continuing legal education, formal apprenticeship structures) that make formation visible institutionally through mechanisms less regulated professions lack.
The second component is empirical evidence of similar patterns to the software case. The empirical evidence from law demonstrates that AI mediation has been operating on certain components of legal work - research, document drafting, contract review, due diligence - paralleling the displacement visible in software. The selective character of the displacement is similar: it operates on certain components of the work and leaves others unaffected. The pipeline-disruption mechanism is similar: junior lawyers in current practice are formed differently than junior lawyers in earlier generations, and what they are being formed in has been altered by AI mediation.
The third component is institutional-response evidence. The legal profession’s regulated status means that institutional responses to AI mediation have been visible. Bar associations, law schools, and regulatory bodies in multiple jurisdictions have begun to develop explicit governance of AI mediation in legal practice - disclosure requirements, training standards, ethical guidance, formation pathway protections. The institutional responses are important for the framework because they demonstrate that a regulated profession can engage the formation question through institutional architectures, and the analytical chapters of the thesis draw on this evidence in ways the recommendations of Chapter 9 then develop.
6.4 The AI-washing complication
The AI-washing complication is important for the chapter’s argument and requires direct engagement. The complication is this: some firms have framed cost-cutting decisions as AI-driven displacement when the underlying decisions were made for other reasons (cost discipline, restructuring, market repositioning, response to interest-rate pressure on technology investment). The consequence is that labour-market data alone cannot demonstrate that AI mediation is producing the displacement attributed to it.
The framework’s response is twofold.
First, I treat the AI-washing complication as real and analytically important. The analytical chapters of the thesis do not depend on labour-market data alone. The framework’s claims about contemporary advisory transformation operate at the level of formation, sovereignty, extraction, and political economy - analytical levels that operate independently of headcount changes in the consulting profession. The chapter develops this point explicitly because Chapter 7’s analysis depends on the framework operating at the formation level, not headcount.
Second, the framework demonstrates that the evidence on formation operates independently of the labour-market data. Even where AI-washing has complicated the labour-market evidence, the evidence on formation pathway disruption remains operative. What junior practitioners are being formed in has been altered by AI mediation, regardless of whether the labour-market data is partially explained by cost-cutting framed in AI terms. The framework’s analytical contribution is partly to make this distinction visible: the question of what is happening to professional formation is independent of the question of how much labour-market displacement is attributable to AI mediation specifically.
The AI-washing complication has a third consequence I treat as analytically important. The complication demonstrates that the political-economic conditions under which contemporary AI mediation operates - including the incentives firms have to frame cost-cutting decisions in AI terms - are part of what the framework has to engage. The narrative of AI-driven transformation has institutional weight regardless of how much of the transformation is actually AI-driven. Chapter 5’s political-economic argument develops this consequence.
6.5 The Susskinds as foil
Richard and Daniel Susskind’s The Future of the Professions is the analytical foil against which the framework’s formationist account is developed. The Susskinds advance a substitutionist account of how AI mediation transforms the professions: they predict that professional services will increasingly be delivered through AI-mediated systems, with human practitioners displaced from the work and reduced to narrower roles where their contribution is institutionally recognised but quantitatively diminished. The substitutionist account has been influential in the analytical conversation about AI in professional work.
The framework’s engagement with the Susskinds operates as disagreement at the level of analytical apparatus, not the empirical level. Three points of differentiation carry the framework’s response.
The first is the unit of analysis. The Susskinds address professional services delivered to clients; the framework addresses professional formation across generations. The two units of analysis ask different questions about the same situation. The Susskinds’ question is: how will professional services be delivered as AI mediation matures? The framework’s question is: what is happening to the formation pathway through which professional expertise is reproduced as AI mediation operates on professional work? The two questions are not in opposition; they are at different analytical levels. The framework’s contribution is that the formation question cannot be reached from the service-delivery analysis the Susskinds undertake, and that the concerns of the framework operate at the formation level the Susskinds do not engage.
The second point of differentiation is the content of expertise. The Susskinds treat professional expertise as the institutional architecture through which services have historically been delivered, with the implication that AI mediation can replicate the expertise once the institutional architecture is sufficiently mature. The framework treats professional expertise as contributory expertise in Collins’s sense - the ability to contribute to the practice through embodied engagement, reproduced through formative engagement with consequence over the arc of a career. The framework’s claim is that contributory expertise is constitutively different from the institutional architecture through which it has historically been delivered, and that AI mediation cannot replicate contributory expertise even where it can absorb the institutional architecture through which it has been delivered.
The third point of differentiation is the normative content. The Susskinds advance a neutral account of the transformation they predict: their position is that AI-mediated service delivery will operate more efficiently than human-mediated service delivery, and that the normative question is whether the institutional architectures of the professions adapt to this efficiency gain. The framework I deploy is normative: what is at stake in European infrastructure advisory is the answerability of the work to the third box, the European public-interest tradition, and the integrity of the formation pathway through which contributory expertise is reproduced. The framework’s position is that these contents are at structural risk under contemporary AI mediation, and that the question is not whether the institutional architectures adapt to the efficiency gain but whether the content survives the structural transformation.
The framework’s engagement with the Susskinds is disagreement at the analytical level. The Susskinds are not refuted; they are answered with an alternative analytical apparatus that works at a different level and engages concerns the substitutionist account does not engage.
6.6 Four findings the thesis carries forward
The chapter’s contribution is four findings synthesised from the comparative evidence. The four findings are then carried into Chapter 7’s stage-by-stage analysis, where they operate as analytical lenses applied within each ChMC stage.
First finding: selective displacement. AI mediation operates as selective displacement of savoir-faire into technical systems. What is displaced is suited to what AI mediation can absorb - routine analytical components, expressive layers of professional output, formative tasks at the early stages of professional progression. What is not displaced is suited to what AI mediation cannot absorb - situated judgment, relational intelligence, engagement with consequence, the contributory expertise exercised at the senior tiers of professional progression. The selectivity is the proletarianisation mechanism I name: it does not abolish the practitioner; it displaces specific components of professional work.
Second finding: pipeline rupture regardless of headcount. The consequence of selective displacement is structural disruption to the formation pathway through which professional expertise is reproduced. The disruption acts on formation, not employment, with the consequence that headcount stability does not protect against pipeline rupture. The evidence from software and law demonstrates that the formation pathway is being structurally altered even where employment levels are stable, and that what junior practitioners are being formed in has been transformed. I treat this finding as the most analytically robust of the four because it survives the AI-washing critique that complicates the labour-market evidence.
Third finding: uneven upward value flow. AI mediation produces distributional consequences within the professions where it operates. What is being absorbed at the formative tiers (the routine analytical work, the expressive output, the formation pathway) flows upward in value terms to the senior tiers (where situated judgment remains valuable) and to the institutional architectures that own the AI mediation systems (the platform vendors, the firms that own integration capability, the institutional architectures that capture displaced value). The consequence is that AI mediation is not value-neutral; it is redistributive, with the direction of redistribution operating upward through the institutional hierarchy of the professions and outward into the institutional architectures of platform ownership.
Fourth finding: explicit narration of professional legitimacy. The consequence of the first three findings is that the institutional architectures through which professional legitimacy has historically been recognised are under pressure. The implicit legitimacy that has historically operated through institutional architectures of expertise, formation, and senior practice is insufficient under AI mediation. The consequence is that legitimacy now requires explicit narration: explicit articulation of what is being done, who is doing it, what AI mediation operates, and what content the practitioner is contributing. The evidence from law’s institutional response (bar associations developing explicit AI governance, disclosure requirements, formation pathway protections) demonstrates that this finding operates and that institutional responses are available.
These four findings are the chapter’s contribution to the analytical apparatus the thesis deploys. Chapter 7 carries them forward into the stage-by-stage analysis of European infrastructure advisory, where they operate at each ChMC stage. Chapter 8 carries them into the futures analysis. Chapter 9 carries them into the recommendations to constituencies.
The cross-professional picture assembled from software and law acquires additional empirical depth from the medical literature. Natali et al.’s mixed-method review of AI-induced deskilling in medicine [@natali2025deskilling] synthesises the available evidence across healthcare and identifies specific competency domains at risk from AI-driven decision support: physical examination, differential diagnosis, clinical judgment, and physician-patient communication - precisely the formation-through-engagement capacities the tacit-as-formation account predicts will be most vulnerable. The authors introduce a distinction of direct relevance to the pipeline-rupture argument: upskilling inhibition (the foreclosure of skill acquisition) is analytically distinct from deskilling (the loss of already-acquired skills), and it is upskilling inhibition rather than deskilling that captures the framework’s primary concern - the failure of junior practitioners to acquire the capabilities they would have acquired through formative engagement, not the loss of capacities already formed. The figure is striking: 55% of surgical residents in 2023 reported reduced skill development attributable to AI-assisted decision support [@natali2025deskilling] - a high proportion and a recent datum that suggests the mechanism is not a theoretical projection but a live empirical process. The paper’s explicit cross-domain extension licenses its use here: the authors frame their findings as applicable wherever AI decision support displaces practitioner judgment in professional domains. The PACES-MRCPUK clinical competency framework Natali et al. deploy - which structures medical professional progression through stages, each dependent on formation-through-practice - serves structurally the same function that the ChMC framework serves in consulting, and the parallel itself is analytically significant: the formation-pipeline-rupture concern appears wherever a structured competency framework governs professional progression.
The four findings synthesised here from comparative empirical evidence have, since the thesis was drafted, acquired theoretical support from a different methodological direction. Acemoglu, Kong and Ozdaglar’s recent formal model of agentic AI under collective learning [@acemoglu2026collapse] derives, from quite different premises and through quite different analytical apparatus, a structurally similar conclusion: that AI mediation operating as substitute for human effort can produce long-run depletion of the collective knowledge that sustains future expert action, even as it improves individual decision quality. Their model formally derives what the cross-professional empirical evidence here registers descriptively: that displacement is selective, targeting the production of expertise, not its use, that pipeline rupture operates on knowledge reproduction independently of headcount, and that the value flows the framework predicts are robust features of how AI mediation interacts with knowledge-intensive work. The convergence between empirical evidence from adjacent professions, formal economic theory, and the philosophical framework I deploy is itself analytically significant. The proletarianisation mechanism the thesis treats as the framework’s central analytical commitment is now visible in three independent registers: in the empirical patterns of software and law, in formal economic modelling of collective learning, and in the philosophical reading of professional formation that the thesis develops. None of the three is sufficient on its own; together they constitute mutually reinforcing evidence that the structural concern is real.
6.7 Four findings, one picture
Software and law together produce a picture of how AI mediation operates in knowledge-intensive professions that the framework carries forward: selective displacement at the component level, pipeline rupture at the formation level regardless of headcount stability, uneven value redistribution toward senior tiers and platform owners, and a new requirement that professional legitimacy be narrated explicitly rather than assumed institutionally. The AI-washing complication does not undermine this picture - the formation evidence is independent of the labour-market data complications, and it is at the formation level that the framework’s claims are load-bearing.
The Susskinds’ substitutionist account is the analytical foil because it reaches the same empirical domain from a different unit of analysis. The difference is not disagreement about what is happening; it is disagreement about what question to ask. The framework asks about formation; the Susskinds ask about service delivery. The formation question cannot be derived from the service-delivery analysis, and the concerns the framework names - pipeline rupture, sovereignty risk, third-box answerability - are invisible from the Susskinds’ vantage point.
Chapter 7 carries the four findings into the stage-by-stage core of the analytical work: how selective displacement, pipeline rupture, upward value flow, and legitimacy narration manifest at each of the three ChMC stages of European infrastructure advisory specifically.