Chapter 8 - Futures: Scenarios for European Infrastructure Advisory

8.1 Constructing the four scenarios

This chapter takes the cross-stage pattern Chapter 7 developed into futures. It constructs four scenarios for professional advisory under AI mediation: not predictions 💡 , but analytical exercises that stress-test the framework against alternative paths. European infrastructure advisory is the primary illustration: the actors and the third-box stakes are drawn from this research site. But the two axes the scenarios derive from are general: where expertise lives (in human practitioners or in AI models) and under whose sovereignty AI models operate (extractive Anglo-American commercial or alternative). The same four quadrants apply to professional advisory broadly. Commodity Advisory is open to any advisory firm where AI mediation absorbs the formative base; Artisanal Advisory to any firm that recognises the formation question and builds its institutional choices around it.

Between the fourth scenario and the cross-scenario synthesis, the chapter includes a supplementary thought-experiment (the Knowledge Commons) that sits deliberately outside the four-quadrant structure. It is there to show that sovereignty and formation may well be independent problems.

The chapter then identifies seven cross-scenario findings: concerns that hold whichever future unfolds, set out in full in §8.8 and carried into Chapter 9’s recommendations to constituencies.

8.2 Method: scenarios as philosophical thought-experiments

The scenarios in this chapter are philosophical thought-experiments.

The corporate-strategic scenario tradition (developed by Pierre Wack at Royal Dutch Shell in the 1970s 1 and elaborated by Kees van der Heijden 2 , Peter Schwartz 3 , and adjacent figures in the strategic-foresight literature) constructs scenarios as narratives that stress-test institutional decisions against alternative futures. Corporate-strategic scenarios serve the decisions they are built around: they help an organisation learn to operate under conditions structurally different from the present. The tradition has shaped strategic-foresight practice, and I draw on its apparatus here. 💡

The philosophical thought-experiment tradition (developed in twentieth-century analytic philosophy, with contributions from Williamson on the epistemology of thought-experiments 4 , Brown and Fehige on the philosophical apparatus of thought-experimentation 5 , and adjacent figures in philosophy of science) constructs counterfactual scenarios as exercises that stress-test a framework against alternative conditions. Philosophical thought-experiments serve the frameworks they test: they show a tradition what its claims commit it to under conditions unlike the present.

We’ll try and bridge the two. The scenarios work as corporate-strategic exercises (they bear on what happens to European infrastructure advisory) and as philosophical thought-experiments (they stress-test the framework’s claims). The chapter keeps both in view: what European infrastructure advisory becomes, and what the framework’s apparatus commits to under alternative conditions.

The scenarios are counterfactual rather than forecast. I make no claim that any one is more likely than the others, and the analytical work they do does not depend on probability. They lay out paths whose conditions the framework can analyse, with consequences for the four constituencies Chapter 9’s recommendations then engage.

The two structural axes come out of the analytical chapters. The expertise location axis follows from the framework’s distinction between contributory expertise (which lives in human practitioners, reproduced through the integrated three-scale system) and the content AI mediation can absorb (which lives in the technical systems the mediation runs through); its poles are expertise concentrated in practitioners and expertise concentrated in systems. The sovereignty conditions axis follows from Chapter 5’s political-economic argument: where AI mediation systems are produced and who owns them. Its poles are extractive Anglo-American commercial sovereignty (the current dominant condition) and alternative sovereignty (European public-interest or firm-proprietary).

Crossing the two axes gives the four scenarios.

8.3 Scenario one: Commodity Advisory

The project team is three people: a partner, a senior consultant, and a recent graduate running generation pipelines and approving outputs. Five years ago this engagement would have required eight people over four months, two of them at the Applied stage. The work passes quality checks, the client does not ask who did what, and nobody has said out loud what the graduate is not learning, or where the next senior consultant will come from. 💡

Commodity Advisory is the quadrant where expertise concentrates in technical systems under extractive Anglo-American commercial sovereignty. European infrastructure advisory is absorbed into the generic capabilities of AI mediation systems built to be sold as services to consulting firms, infrastructure clients, and the institutions around them.

Generic AI mediation matures across the components of advisory work that Chapter 7’s analysis identified: the formative tasks of Foundation stage, the deliverable production of Applied stage, increasingly the analytical components of Chartered-stage work. The cumulative effect is commoditisation: advisory work becomes replicable across firms, with distinctiveness compressed into a narrow band where senior contributory expertise still operates. The consulting profession narrows rather than disappears. Value flows upward to the platform vendors that own the AI mediation systems and inward to the senior practitioners whose contributory expertise has not been absorbed.

The Commodity Advisory scenario corresponds closely to the path Acemoglu, Kong and Ozdaglar’s model (Chapter 5, §5.6) follows when agentic AI accuracy is sufficiently high and the stock of general knowledge is allowed to deplete unchecked: the knowledge-collapse steady state 6 . My reading of Commodity Advisory as the path of least resistance under current conditions is the philosophical and political-economic analogue of that result: under unchecked extractive conditions, the system depletes even as individual outputs continue to look competent.

The three ChMC stages each take a distinct hit. Foundation-stage work is absorbed into AI mediation and its formation pathways collapse, with consequences for what Applied-stage practitioners can be formed into and what Chartered-stage practitioners can become two decades from now. Applied-stage work is commoditised and the deliverable-as-performance moat erodes. Chartered-stage work shrinks to the content that AI mediation cannot replicate, which is less than senior advisory covers in current practice.

The third box loses its footing. Advisory answers to the publics whose lives infrastructure decisions shape only through what we treat as constitutive: contributory expertise reproduced through the integrated three-scale system, engagement with consequence, answerability across long time horizons. Under Commodity Advisory that is absorbed into generic AI mediation, and whatever answerability remains runs through systems tuned to extractive Anglo-American commercial sovereignty rather than to the European public-interest tradition.

Milan, 2024. An Italian infrastructure advisory firm has lost three consecutive bids for energy regulatory strategy work to competitors who have structured their offerings around AI-mediated analytical production: faster turnaround, lower price point, outputs that cover the required regulatory categories in the required format. The firm’s partners are confident the outputs are analytically thinner: less specific to the Italian regulatory context, less attentive to ARERA’s particular institutional character, less capable of the strategic framing that has historically been the firm’s differentiator. The clients appear either not to notice the difference or not to weight it against the price gap. Asked by one client what the higher fee would have bought, the partner who led the last bid could describe the firm’s process but not, in terms the client could check, the judgment it produces. One partner prepares a presentation for the firm’s annual strategy day titled: The market is pricing our distinctiveness at zero. The conversation that follows is long. Nobody disputes the diagnosis. Nobody has a confident answer to what it implies. 💡 💡

The market-level evidence for this trajectory is already measurable, though it rests here on a single industry source and not a peer-reviewed body of work: the vendor convergence documented in Chapter 5 (§5.3). The same analysis observes that the firms will keep delivering the same four workstreams to every client, and that for a client buying its positioning this way the result is “a slow, expensive way to disappear into your category” 7 . From inside the Commodity Advisory reading, the analyst’s observation is exact: the same four workstreams, the same vendor shortlist, the same generic mediation apparatus, marketed as differentiation and operating as its opposite. The scenario this chapter treats as a thought-experiment is, at the market level, an ongoing process. A controlled experiment shows the same mechanism in another domain: in Doshi and Hauser’s online study of short-story writing, access to story ideas from a large language model led evaluators to rate individual stories as more creative, especially those by less creative writers, yet made the stories more similar to each other, an increase in individual creativity “at the risk of losing collective novelty” 8 . The study concerns eight-sentence fiction rather than advisory, so it supports the convergence argument by analogy only: each output improves while the range of outputs narrows, which is the Commodity Advisory pattern in miniature.

The market’s proposed response to this trajectory (the exit from Commodity Advisory through proprietary knowledge management) is Fuller’s rebuilt learning loop, described in Chapter 4 (§4.3). 💡 For large well-capitalised firms it has the structure of the Digital Twin Economy scenario and is a real alternative. But it is silent on formation. It assumes practitioners who can design the proprietary loop, evaluate its outputs, and sustain the judgment required to govern a system trained on the firm’s own accumulated knowledge. Where those practitioners come from, once the apprenticeship blow has landed at Foundation stage and disrupted the formation pipeline, it does not say. The market’s most sophisticated proposed exit from Commodity Advisory describes where knowledge should flow without accounting for how the practitioners capable of making it flow well are formed.

Sovereignty gives way at both scales. National-cultural sovereignty erodes as European public-interest practice is absorbed into generic AI mediation whose content is dominated by Anglo-American patterns. Organisational sovereignty erodes as firm-distinctive ways of seeing are compressed under a mediation that works the same way in every firm.

Extraction is at work in this scenario. Firm-externalised knowledge flows outward into vendor systems whose content is then recirculated through generic capabilities to competitors and clients. The dark mirror of the SECI model operates at commercial scale: what firms have historically reproduced as competitive distinctiveness becomes part of the generic distribution that AI mediation draws on.

Commodity Advisory is the scenario that worries me most, and the one current conditions tilt toward unless the institutions involved act on the concerns named here. Chapter 9 is about that action.

8.4 Scenario two: The Digital Twin Economy

The procurement brief has a forty-page annex: technical specifications for sovereign AI deployment, including EU data residency requirements, national jurisdiction for training data, and regulatory inspection rights for model query logs. It is an advisory engagement, not an IT tender. The regulatory body issuing it has spent three years building the institutional architecture this annex requires. Nothing in the annex asks whose way of reading the regulation the model will carry. The firm reading it takes three days to decide whether to bid. 💡

In the Digital Twin Economy, expertise again concentrates in technical systems, but under alternative sovereignty conditions. European infrastructure advisory is still absorbed into AI mediation; the systems doing the mediating are European-public, firm-proprietary, or otherwise outside extractive Anglo-American commercial sovereignty.

The absorption of advisory work into AI mediation runs much as in Commodity Advisory; what differs is who owns and governs the systems. European public-interest AI initiatives (sovereign AI capability, European model-governance rules, public-interest curation of training data) supply the content the mediation runs on. Firm-proprietary capabilities (built in-house, trained on the firm’s own material, governed by the firm) give firm distinctiveness another way to reproduce itself.

Western France, 2024. A port authority has commissioned a digital twin of its logistics infrastructure under an EU-funded port digitalisation programme. The twin is built on a platform developed by a French technology consortium under a national AI investment programme, trained on French port operations data, hosted within the consortium’s infrastructure, governed by an agreement that keeps training data and model weights within French jurisdiction. An infrastructure advisory firm is brought in not to advise on port operations but on the governance architecture for the twin: who owns the model’s outputs, who can query it on what questions, what happens when the model’s recommendations conflict with the port authority’s own operational judgment, and how the twin’s advice should be documented for regulatory purposes and explained to the towns that live with the port’s decisions. The engagement concerns epistemic infrastructure for governing port infrastructure (a distinction the firm’s engagement scope had not previously needed to make), not port infrastructure itself. 💡

Here both scales of sovereignty hold: the national-cultural through European public-interest AI capability, the organisational through firm-proprietary capabilities that carry firm-distinctive content.

Extraction runs differently here. The knowledge firms externalise flows into systems that record the firm’s ownership; the knowledge public bodies externalise, into systems that record public ownership. The dark mirror of the SECI model is dimmed by ownership terms that keep track of who put what in.

Measured against Commodity Advisory, each stage fares differently. Foundation-stage work is absorbed, but what is absorbed is European public-interest practice rather than generic Anglo-American; formation pathways are altered, yet formation stays oriented to the European tradition. Applied-stage work is commoditised in technical terms, inside systems built for European public-interest practice. Chartered-stage work becomes supervisory judgment over the mediated system, exercised under the alternative sovereignty conditions.

Here the third box is harder to call. Answerability to the publics runs through AI-mediated systems that know the third box only as far as the European public-interest tradition has been written into them. Whether that is enough depends on whether the governance behind the alternative sovereignty actually reaches those publics. I take the tradition to make this available, not to guarantee it: engagement with the third box still has to be built in deliberately (in the twin’s mandate, in what it must document and for whom, as the port engagement is doing), even where the expertise sits in the system.

The Digital Twin Economy is a long way from built. The pieces it needs (European public-interest AI capability, firm-proprietary AI development, governance for both) exist in early form, but nowhere near the scale the scenario assumes. 💡

8.5 Scenario three: Artisanal Advisory

The lead partner has one standing instruction for her team: every analytical judgment in every client-facing document is owned by a named consultant with domain experience in the relevant sector. AI tools run the background research. The analytical judgment is human, identifiable, and answerable, and juniors draft beside the named consultant, defending their reading of the problem before any tool is opened. The firm charges a premium and has never lost a client who genuinely valued the distinction; it has had to explain the distinction to every new one. 💡

Artisanal Advisory keeps expertise in human practitioners, under alternative sovereignty conditions. European infrastructure advisory goes on reproducing itself through the integrated three-scale system Chapter 4 described; the firms, professional bodies and public buyers around it deliberately protect the formation pathway through which contributory expertise has always been reproduced, and the sovereignty conditions recognise European public-interest practice.

AI mediation is used, but bounded: it is folded into the three-scale system rather than replacing any part of it. Foundation-stage formation pathways are protected because the firms and bodies involved treat formation as constitutive of the profession. Applied-stage work stays where the firm’s voice and the individual practitioner’s analytical signature live. Chartered-stage advisory stays the seat of senior judgment, formed through sustained engagement with consequence over the arc of a career. The design principle that unifies these choices is what Xu et al. (2026) call scaffolded AI friction: the deliberate introduction of cognitive resistance into AI-mediated processes to force active engagement rather than passive acceptance 9 . Artisanal Advisory applies this principle institutionally rather than in the interface: the partner’s rule that every analytical judgment is owned by a named practitioner, formation pathways kept clear of generic mediation, supervision that makes practitioners engage rather than accept. Xu et al.’s design intervention for human-AI interfaces becomes, here, a way of forming people.

Acemoglu et al.’s formal model arrives at the Artisanal Advisory scenario from a policy-design direction that complements the philosophical reading here 6 : a moratorium on agentic AI followed by a permanent cap on its precision, the two-phase policy Chapter 9 (§9.6) sets beside this work’s recommendations. Artisanal Advisory is the political-economic analogue of what that welfare-optimal policy would produce: a long-run condition in which AI mediation runs inside rules that protect formation pathways, recognise the public-interest tradition, and cap how far AI substitutes for the formative engagements that reproduce contributory expertise.

The market has not waited for rules of that kind. The profession is already pricing this scenario, and the way it prices it is the scenario’s main risk. The HEC Alumni white paper (Chapter 2) expects that “certains clients paieront davantage pour un conseil sans IA, comme on paie une prime pour un sac Hermès ou un vinyle en édition limitée”, and warns in the same breath of “folklorisation” 10 ; among its recommended offers is a “Premium artisanal”, “intentionnellement sans IA pour les sujets sensibles ou critiques” 10 ; and its catalogue of cognitive biases lists overvaluing such AI-free offers “uniquement parce qu’elles sont rares” 10 . Read together, the three passages describe Artisanal Advisory as a luxury tier: formed judgment sold at a premium to clients who can afford the distinction, on subjects the firm deems sensitive, with the preference for it treated as a bias to be managed. That is a different thing from the scenario set out above, where the bounding is done by the institutions around advisory rather than by a segment of clients willing to pay for the handmade. The luxury version keeps the formation pathway open for a few firms and a few clients; the public-interest version keeps it open for the third box. The scenario’s plausibility turns on which of the two the market and the institutions select, and the white paper is evidence that, left to the market, the luxury version is the one already forming. A single firm can still hold the distinction on its own terms.

Copenhagen, 2025. A Danish infrastructure advisory practice of twelve consultants has built its entire market position on a single commitment: all regulatory analysis is produced by a named senior consultant with specific domain experience in the sector the client operates in. The firm does not use AI-mediated drafting for deliverable content. It uses AI extensively for background research, data aggregation, and document search. The distinction is a formation and quality assurance commitment rather than a refusal of technology: the analytical judgment in every client-facing document has been formed through genuine domain engagement. The firm is small. It charges a premium its larger competitors find implausible. Its client base is three major Danish infrastructure operators, one Scandinavian regulatory body, and two European infrastructure funds who have learned that when they need to understand what a regulatory development actually means for capital allocation, the Copenhagen practice will tell them something a platform cannot, in a voice they recognise as the firm’s. Its juniors spend their first two years in a single sector before they sign anything. 💡 💡

Of the four quadrants, this one serves the third box best. Advisory answers to its publics through practitioners whose contributory expertise was formed the long way, through the integrated three-scale system, and the European public-interest tradition survives because procurement and disclosure norms still treat advisory work as part of public-interest practice.

European public-interest practice keeps its national-cultural distinctiveness, and firms keep their own organisational voice: sovereignty holds at both scales, because the rules bounding AI mediation were written to preserve both.

Firms still externalise, but under terms that name what is leaving and to whom: extraction is contained rather than stopped, and the dark mirror of the SECI model is bounded by vendor terms drafted with the cost of unmitigated extraction in view.

Each ChMC stage is protected rather than eroded: Foundation pathways held open by design, Applied work still where firm voice and individual contribution are made, Chartered advisory still where senior contributory expertise is exercised and passed on.

Artisanal Advisory sits closest to the European public-interest tradition. It also asks the most of the institutions around advisory. Each piece it needs (formation pathway protection, procurement frameworks that recognise contributory expertise, disclosure norms that engage the third box, sovereignty arrangements that recognise European public-interest practice) exists somewhere in current practice, in fragments; nowhere are they assembled.

8.6 Scenario four: Disintermediated Clients

The utility has brought the work in-house. The advisory firm still has an engagement: four senior conversations a year, no deliverables. Nine years of engagement knowledge, regulatory judgment, and institutional memory sit partly in the utility’s new AI platform and partly in the heads of three advisors who have gone in-house. The platform vendor manages the architecture, and the utility’s board has not asked who owns the model weights, or who now answers to the regulator and the public for the analysis the platform produces. 💡

In the fourth quadrant, expertise stays in human practitioners while sovereignty is extractive Anglo-American commercial. Infrastructure clients build in-house AI capability that absorbs advisory work; advisory practice shrinks to compressed engagement scopes in which senior contributory expertise is still valued but the profession around it narrows. The axis location reflects the residual advisory relationship: the content clients continue to commission resides in the senior human practitioners whose judgment their platforms cannot replicate. The expertise axis describes what remains in the advisory relationship, not where the absorbed analytical work now resides.

Sophisticated infrastructure clients (energy utilities with substantial analytical functions, transport authorities with developed planning departments, regulatory bodies with technical capacity) build in-house AI capability that absorbs the analytical work that used to be bought as consulting engagements. The profession survives in reshaped form: engagements compress to what in-house capability cannot reproduce, which is senior contributory expertise, and everything that makes consulting consulting (firm structures, formation pathways) comes under pressure as they narrow. 💡

Lisbon, 2024. A sector regulator has built an AI-supported analysis platform that processes consultation documents, finds precedents in European regulatory decisions and drafts preliminary summaries within hours of a filing. It does not replace the regulator’s analysts; it replaces the pre-analysis that external advisers used to supply. The advisory firm that briefed the regulator for seven years keeps the questions the platform does not answer, about the political economy behind the regulator’s formal positions, and loses the rest. Two of its Applied-stage consultants have moved to the regulator to run the platform, taking with them much of what the firm knew about the sector. The firm has stopped hiring at the level where those two learned it. The regulator is, by any analytical measure, more capable than it was three years ago; the firm advises a better-informed client with a smaller team and no obvious route by which its next sector specialists will be formed. 💡

Extraction is the scenario’s defining dynamic. Expertise that belonged to advisory flows into in-house AI capability shaped by the client’s organisation rather than by the profession. The firm’s engagement knowledge (accumulated over years of sustained advisory relationship) becomes training material for the client’s platform, governed by the technology vendor, subject to the vendor’s ownership and reuse conditions. The risk to the profession here is absorption through the client side, accumulating engagement by engagement, rather than direct substitution by AI.

Sovereignty is lost at both scales here as well, by a different route. At the national-cultural scale, European public-interest practice ends up running on extractive Anglo-American commercial AI even when the capability is nominally the client’s own (the vendor manages the platform; the board has not asked who owns the weights). At the organisational scale, firm voice thins as the firms themselves are reshaped under disintermediation pressure.

Stage by stage, the picture is uneven. Foundation-stage work shrinks in volume as the client’s platform absorbs the analytical tasks that used to be junior content, and formation pathways compress with it. Applied-stage work shrinks as deliverables narrow, with the same consequence for formation. Chartered-stage work continues wherever senior contributory expertise is still bought, but inside a smaller firm doing a different kind of work.

As in the Digital Twin Economy, the third box is left in doubt. Advisory still answers to its publics through the senior contributory expertise that survives in the scenario. But the channels that answerability has historically run through (the firm-mediated engagement, the integrated three-scale system, European public-interest advisory practice) are all under pressure. Whether it survives the reshaping depends on whether the client’s in-house capability can carry the third box. That is a harder test than it looks: the judgment that moves in-house does not stay whole once it arrives. The client’s executives are exposed to the same mediation: Pomme’s account of strategic decision-making under AI finds that the loss of autonomy at the top can take the form of “a silent redistribution of the power to judge, interpret, and set boundaries for action” 11 . And the assurance the client used to buy with the external engagement does not travel with the output. In the Beyond the Prompt episode quoted in Chapter 7, Utley, recalling a chief executive about to hire BCG for a five-month study, put it this way: “when you know BCG has done it, you assume they fact checked. Now you have this weird thing of like, is this true?” 11 . 💡

Of the four, Disintermediated Clients is the one most threatening to the profession as a profession, even while senior contributory expertise stays valuable. It is also the one already emerging from procurement practice in some infrastructure subdomains, where in-house AI capability is being built and engagement scopes are compressing.

8.7 A supplementary thought-experiment: the Knowledge Commons

A European infrastructure advisory consortium of nineteen members (transport authorities, energy regulators, water utilities, two national infrastructure banks, six private advisory practices across four national jurisdictions) has operated a shared AI platform for seven years. Eleven years of aggregated project data: regulatory submissions, infrastructure assessments, procurement strategy documents, post-project evaluations from engagements in transport, energy, water, and port governance across the consortium’s member jurisdictions. The platform is governed collectively, its training data contribution rights distributed by sector and jurisdiction, its model weights owned by the consortium’s legal entity, its governance architecture designed by a Dutch public-law firm specialising in intellectual commons. The sovereignty question is answered. The formation question is on the agenda of none of its governance committees. 💡

A junior analyst in Rotterdam, twenty-six years old, three years into her career, works with the platform daily. She is competent. She produces contextually specific, sector-appropriate regulatory analysis that her supervisors recognise as professional work. She has never built an analysis from scratch that she did not abandon when the platform suggested something better, because the platform, trained on eleven years of aggregated expert judgment across nineteen organisations, consistently suggests something better. What she cannot do is say what she would have seen without the platform’s prior framing. She cannot say whether the judgment she exercises is hers, or the consortium’s, averaged across nineteen organisations. 💡

The Knowledge Commons sits outside the four-quadrant structure. Rather than a fifth position on the two axes, it is a different kind of institutional arrangement, one that resolves some of the concerns the four scenarios raise, leaves others intact, and makes one worse. It is here because a commons that resolves sovereignty and still fails formation shows what is distinct about the formation question: it is not a sovereignty problem, and sovereignty solutions do not touch it.

What the commons resolves. A well-designed commons addresses the sovereignty concerns identified across the four scenarios (the dual-scale threat to national-cultural and organisational sovereignty). Extraction becomes symmetric: knowledge flows into the shared platform and returns to contributing institutions, governed by the consortium’s collective ownership architecture. The dark mirror of the SECI model is turned; what firms externalise they co-own, and the extractive asymmetry that the dark-mirror argument identifies is designed out. National-cultural sovereignty is preserved through governance design: the consortium’s training data architecture ensures that European public-interest practice, regulatory traditions and institutional patterns remain the epistemic content the platform runs on. Organisational sovereignty persists through sector- and jurisdiction-differentiated contribution rights, under which firms contribute expertise within their domains and draw from the collective across them. The third box is partially addressed: the platform’s governance architecture is expressly public-interest in design, and the consortium’s mandate can be constitutionally oriented to the infrastructure publics the third box names. The productivity floor rises across all nineteen organisations, and junior practitioners work at a level of contextual sophistication that would have required years more formation to reach independently.

What the commons does not resolve. The sovereignty resolution is real, but it does not touch the formation deficit, which is not a sovereignty problem.

Cross-scenario Finding 1 (§8.8) holds. The commons produces interactional expertise at scale, applying the domain’s regulatory precedents and producing outputs that read as professionally competent, because that is what eleven years of aggregated expert judgment, encoded and made queryable, can support. It cannot produce contributory expertise, whose formation needs the accumulation of Erfahrung across encounters with difficulty, the discomfort of analytical failure, the forced development of independent judgment where the platform is not available or has not yet encountered the problem 12 13 . The commons has no mechanism for this. It makes Erfahrung unnecessary, and the forming of independent judgment difficult to pursue. The analyst in Rotterdam is interactionally fluent and contributorily hollow.

Cross-scenario Finding 2 holds, and under commons conditions becomes harder to see. In the four scenarios, the formation deficit is at least legible: in Commodity Advisory, it is visible as the collapse of Foundation-stage pathways; in Artisanal Advisory, seeing it is what motivates the scenario’s defining choices. In the Knowledge Commons, the formation deficit is concealed by the contextual sophistication of the commons’s outputs. Because the platform is trained on European public-interest practice rather than generic Anglo-American patterns, its outputs are analytically appropriate, jurisdictionally accurate, institutionally calibrated. The junior analyst’s work, shaped by the platform, reads as formed. The concealment is dangerous because it is profession-wide: where individual firms in Commodity Advisory might notice that their juniors are producing thin work, the commons makes thin formation look like thick competence across all nineteen organisations at once. The false proficiency that Ke et al. document at the individual level 14 now operates at the scale of an ecosystem.

Cross-scenario Finding 7 holds. The commons is built on specialised contribution within domains and free draw across them, and that shape forecloses integrative judgment. Practitioners put their own domain in and take other domains out, but never form the capacity to hold several domains together, because the commons makes the sustained cross-domain engagement that would form it unnecessary. The Applied stage has no safe trajectory under commons conditions any more than in the four scenarios: the analytical function that connects Foundation formation to Chartered authority, where the practitioner learns to hold multiple domains in view at once and judge under uncertainty, is bypassed rather than developed. Over a decade, the ecosystem stratifies into specialists whose integrative capacity is platform-dependent, and therefore unavailable where the platform has no precedent: the novel regulatory problem, the multi-sector interaction with no prior characterisation in eleven years of data.

What the commons intensifies. Beyond the deficits it leaves intact, the commons worsens one concern the four scenarios hold partly in check.

The Morozov circularity (Chapter 3, §3.5) is most acute under commons governance 15 . In the four scenarios, this circularity operates with some external reference: in Artisanal Advisory, the pre-commons tradition of formative apprenticeship provides an evaluative standard not itself produced by the mediation; in Commodity Advisory, the recognition that something has been lost presupposes a reference point outside the current system. In the Knowledge Commons, no external reference point remains at the level of the profession. The commons is constituted by the accumulated professional judgments of its contributing institutions; the evaluative standard for those judgments is produced by the commons; the governance choices about how to develop the commons are made by practitioners whose evaluative frameworks the commons has shaped. The circularity is closed. Whether the commons is adequate cannot be answered from outside it.

The commons and the Artisanal pairing. The commons also sharpens Artisanal Advisory by contrast. Unlike the commons, Artisanal Advisory does not resolve sovereignty comprehensively; it bounds the risk through choices about how AI mediation is used. It is defined by its further choices about how formation happens: protecting Foundation-stage pathways, refusing to absorb the formative parts of Applied-stage practice, governing AI mediation as a tool rather than a substitute. Those choices are independent of any sovereignty arrangement. They would be needed inside a commons as much as outside one.

The European infrastructure advisory ecosystem, under Knowledge Commons conditions, becomes uniformly adequate. Nineteen organisations, across four national jurisdictions, produce contextually appropriate, jurisdictionally calibrated, institutionally attentive advisory work at a level of sophistication that would have required decades of accumulated expertise to reach without the platform. The adequacy is real, the productivity floor is higher, and the third-box constituency is better served by adequate professional advice than by its absence; none of this is to be dismissed.

But the analyst in Rotterdam, at twenty-six, cannot form what the commons makes unnecessary to form. At thirty-six, she will be interactionally fluent at the level of a practitioner with fifteen years of formation, and contributorily adequate at the level of a practitioner who has never needed to form independently. The gap will not be visible in her outputs. It will not be visible in her organisation’s outputs. It will not be visible across the ecosystem. The formation question is unresolved and now invisible, at the scale of an ecosystem whose uniform adequacy keeps its institutions from recognising that adequacy and formation are not the same thing.

The thought-experiment pools a platform. Beneath it sits a second commons, the stock of formed practitioners, and the profession’s own forecasts already record it being drawn down. The HEC Alumni white paper’s prospective annex concludes that “les talents sont la future ressource rare” 10 . The same document reports that McKinsey’s internal assistant is used monthly by “75 % des salariés des premiers niveaux de la pyramide” 10 , predicts that firms pursuing “juniorisation” to protect their margin will be replaced 10 , and recommends automating the formative tier throughout. The forecast and the practice are not reconciled anywhere in its pages. They are the two halves of a commons problem: each firm’s automation of its own Foundation tier is individually rational, and the stock it depletes, the seniors of fifteen years from now, is shared by every firm in the market, including the ones that automated first. The Knowledge Commons resolves who owns the platform and leaves this stock untouched; Artisanal Advisory protects the stock and leaves the platform to others. The profession’s forecast that talent will be scarce is, on the white paper’s evidence, a forecast about what the profession is doing. 💡

That second commons has a name in the human-resource-development literature. Lovett’s cognitive-commons account (Chapter 7) starts from Hardin’s tragedy and asks what happens when the shared resource is a profession’s pool of deep expertise: “rational AI adoption decisions can deplete the shared expertise pool professions require for renewal”, because “training needs and professional expertise regeneration were coupled through the operational necessity of junior labor, and AI decouples them” 16 . His distinction between Internalized Mastery, domain knowledge built through sustained practice, and Distributed Mastery, fluency in orchestrating human-AI systems, is the contributory/interactional distinction in another discipline’s vocabulary, and his six-node causal chain formalises what the white paper’s two halves show in practice: reinforcing loops and time delays “that prevent individual actors from recognizing their collective impact on the regeneration mechanism”. Where his account goes further than this section is on governance. He follows Ostrom rather than Hardin: commons survive where there is “boundary definition, monitoring, graduated sanctions, and collective choice”, and those mechanisms “remain largely absent in the most vulnerable professional sectors”. That absence is what Chapter 9’s recommendations to chartering bodies and policymakers try to fill. 💡

8.8 Cross-scenario findings: what holds across all four

Seven concerns hold across all four scenarios.

First cross-scenario finding: the contributory/interactional distinction is the primary structural determinant of professional value. Across all four scenarios, AI mediation reproduces interactional expertise at scale: the capacity to speak a domain’s language, engage its practitioners, and produce outputs that read as competent. It cannot reproduce contributory expertise: knowledge formed through sustained engagement with consequence that enables a practitioner to contribute to the domain’s development 12 . The findings that follow all rest on this distinction. Formation (Finding 2) is the pathway by which interactional competence becomes contributory expertise. Sovereignty (Finding 3) settles whose posture structures the mediation, and with it whose contributory framework operates. For the third box (Finding 4), contributory expertise rather than interactional fluency sustains public-interest engagement under genuine uncertainty. And explicit legitimacy narration (Finding 5) is how contributory expertise is made legible once agents can reproduce interactional fluency. In each, the stake is whether the conditions under which contributory expertise is reproduced and made legible can be kept.

Second cross-scenario finding: the formation question becomes inescapable. In every trajectory, how the consulting profession reproduces contributory expertise is at stake. Formation pathways collapse in Commodity Advisory as Foundation-stage work is absorbed; in the Digital Twin Economy they are altered to operate within technically mediated systems; in Artisanal Advisory they are protected through institutional engagement; in Disintermediated Clients they compress under reduced engagement scope. The question of what is being formed, how, and with what support is live in all four. It can only be answered, by choices that firms, professional bodies and buyers make or fail to make.

This second finding is the philosophical and political-economic correlate of the learning externality in Acemoglu et al.’s model (Chapter 5, §5.6) 6 .

Third cross-scenario finding: sovereignty pressure operates across multiple scales. In all four trajectories, where AI mediation operates and under whose sovereignty is a live question. Pressure on national-cultural sovereignty (the distinctiveness of European public-interest practice) and on organisational sovereignty (the distinctiveness of firm-level analytical voice) is present in all four, with different force. The scenarios differ in which rules (procurement, professional standards, ownership terms) recognise sovereignty at each scale. Since Chapter 5 reads sovereignty as identity at risk, the finding also says whose distinctiveness survives in each trajectory. At firm scale, Commodity Advisory compresses distinctive ways of seeing under a mediation that works the same way in every firm; the Digital Twin Economy carries them inside firm-proprietary systems; Artisanal Advisory keeps them, and the individual practitioner’s analytical signature, in the work itself; and in Disintermediated Clients they thin as the firms themselves are reshaped.

Fourth cross-scenario finding: the third-box question persists. The third box does not dissolve in any scenario. The publics who bear consequences without holding contracts, including the future generations whose conditions of life are being shaped now, are engaged with different force in each trajectory. The question of how cannot be avoided; what varies is who, in each, is made to answer for them.

Fifth cross-scenario finding: explicit legitimacy narration becomes inescapable. In all four scenarios, legitimacy has to be narrated explicitly, on the terms of Chapter 6’s fourth finding; only the vehicle of the narration changes from one trajectory to the next. The narration has already begun, and its first audience is not the third box. In the earnings calls Blangeois and Lebrument studied (Chapter 4), firms narrate their legitimacy as control, and narrate it to markets 17 . That answers for the firm’s viability, not for what the work owes the publics who live with its results. 💡

Sixth cross-scenario finding: extraction operates structurally across all four scenarios. Knowledge flows outward from European advisory practice into the generic mediation apparatus in all scenarios; the variables are the channel and the ownership. In Commodity Advisory, extraction is unmediated: junior outputs feed training processes, accelerating the substitution that displaces them. In the Digital Twin Economy, it runs through the technical layer: the firm’s analytical posture is encoded in the twin’s configuration, portable and legible. In Artisanal Advisory, it is actively contested: the scenario’s defining institutional choices are those that limit its scope and govern the terms on which the posture circulates. In Disintermediated Clients, extraction is client-side: the client’s direct AI tools absorb the advisory posture the firm historically provided. The question shifts from ‘are we at risk of extraction?’ (all four scenarios answer yes) to ‘what are the ownership conditions of the extraction already happening?’

Seventh cross-scenario finding: the Applied stage has no safe trajectory in any scenario. In Commodity Advisory, Applied practice is commoditised from below as AI mediation absorbs deliverable production, and squeezed from above as Chartered-level authority becomes the defensible asset. In the Digital Twin Economy, Applied practice is technically reconfigured: the consultant becomes a system-integration and quality-assurance function, exercising judgment through configuration rather than direct analytical contribution. In Artisanal Advisory, the Applied stage is the scenario’s defended constituency, which is to say it is otherwise at risk, and the design choices that define the scenario include the refusal to accept Applied-stage substitution. In Disintermediated Clients, the Applied stage is disintermediated outright. Applied-stage vulnerability is distinct from both the Foundation-stage formation argument and the Chartered-stage senior-judgment argument: it concerns the intermediary analytical function (where interactional fluency and early contributory capacity combine) that connects formation to authority, and deserves its own institutional design logic. 💡

8.9 The cross-scenario ground

The four scenarios, built on two axes, differ in how the seven findings configure, but the findings hold in all of them. That is the ground Chapter 9 builds on: its recommendations answer to what holds across the four, not to any one anticipated future, and go to the constituencies whose decisions shape which one arrives.