Chapter 9 — Conclusion: What This Means, and What to Do About It (plain-language version)

What the thesis has argued

The central claim, stated plainly: Generic AI tools, as currently deployed, risk gradually hollowing out the situated, answerable judgment on which the legitimacy of European infrastructure advisory rests — and on which the profession’s obligations to the public and to future generations depend.

Eight chapters have assembled the evidence and argument for this claim. The philosophical framework (Chapters 2–3) explained what’s at stake and which concepts help name it. The domain description (Chapter 4) showed how expertise is actually built. The political economy analysis (Chapter 5) showed whose tools are doing what and who captures the value. The comparative evidence (Chapter 6) showed the same pattern in professions that have been through this longer. The stage-by-stage analysis (Chapter 7) traced the mechanism through Foundation, Applied, and Chartered practice. The futures analysis (Chapter 8) stress-tested the argument against four alternative trajectories.

Together, these chapters produced something the existing research on AI in consulting hasn’t provided: a philosophical account of what’s at stake in the profession’s transformation that addresses formation, sovereignty, and public accountability simultaneously.

A warning about fake solutions

Before the recommendations: there’s a specific failure mode these recommendations are designed to avoid.

The French cultural critic Guy Debord called it récupération — the way institutional systems absorb critical challenges by converting them into their own spectacular forms. The critique produces a representation of a response, which satisfies the demand for action without delivering it.

Applied to this argument, it looks like this. A firm hears the formation concern. They introduce a “responsible AI use policy.” It specifies that human judgment must be applied before AI outputs are used. Juniors are now required to review AI-generated drafts and confirm them. The box is ticked.

None of this restores what the argument requires. The junior validating the AI draft is applying judgment formed through the same AI-mediated conditions the validation was meant to correct. The senior approving the validated output is approving a staged performance of analytical engagement. The policy specifies human oversight. It cannot specify the one thing that actually matters: the time of genuine formative difficulty, the unassisted encounter with analytical challenge, the sustained engagement with consequence through which practical wisdom develops.

A governance response that produces the representation of engagement with the formation argument, without restoring the temporal structure of formation, is precisely the failure the recommendations below are designed against.

What this means for individual consultants

The recommendations here apply to any practitioner in any advisory context where AI mediation has begun to absorb the formative base of the profession. They’re developed in European infrastructure advisory terms, because that’s the research site — but the obligations apply broadly.

1. Seek formation, not just output. The distinction between doing work and having the work done matters most at the beginning of a career. Foundation-stage consultants face a choice, often unconscious, between producing competent outputs quickly with AI assistance and wrestling with problems in ways that are slower but formative. The recommendation is to seek the wrestling, even where the shortcut is available. For Applied-stage consultants: recognise the formative role you play for juniors, and take the supervisory engagement that role requires seriously — not reviewing AI-mediated output with your own AI-mediated review, but genuinely engaging with what the junior understands and doesn’t understand.

2. Practise reflexively. Stay conscious of when your own analytical contribution is being mediated, what the mediation is doing to the work, and what is being preserved or eroded in your own grasp. This doesn’t mean refusing AI tools. It means noticing the difference between using a tool to extend your judgment and outsourcing your judgment to a tool.

A Brussels example. A Foundation-stage consultant is brought by a senior partner to a European Commission technical working group on cross-border energy infrastructure governance. Her formal contribution is minor; the partner considers leaving her behind. He brings her anyway, with a brief: watch what happens when the Commission official asks a question we haven’t prepared for. Forty minutes in, an unanticipated question arrives about the interaction between the Energy Union framework and a specific bilateral treaty obligation. The partner handles it. In the taxi afterwards, they spend forty minutes talking through it: what he did, why, what he couldn’t say and why. She describes that conversation five years later as the most formative single afternoon of her career. She’s certain she couldn’t have had it via an AI tool.

3. Attribute. When AI mediation has substantially shaped your work, say so. The Foundation-stage consultant whose draft was substantially AI-generated owes attribution to the person who reviews it. The Applied-stage consultant whose deliverable was substantially AI-generated owes attribution to the client. The Chartered consultant whose advice was substantially AI-shaped owes attribution to the public who bears the consequences. Legitimacy depends on honesty about what was actually done.

4. Cultivate your distinctive voice. A senior regulatory adviser has, over fifteen years, developed a way of entering every engagement: she consistently asks what the regulatory body is trying to protect itself from — not what it wants, not what the legislation mandates, but what institutional fear is shaping its behaviour. It’s a question that produces different analytical entry points than standard approaches. She tested it once against an AI stakeholder analysis. The model produced a correct and comprehensive answer. It hadn’t asked what the regulator was afraid of. The question hadn’t occurred to it.

The distinctive analytical voice — the framings, the vocabulary, the way of seeing a problem that marks your work as yours — is part of what makes contributory expertise valuable. It requires deliberate cultivation in an environment that continuously pulls toward generic defaults.

5. Take responsibility for the next generation. For Chartered consultants especially: the long-term substance of the profession depends on whether current senior practitioners invest in formation pathways or allow them to erode. The firm’s capacity two decades from now depends on choices being made now. Senior responsibility isn’t only your own work. It’s also who comes after you.

What this means for advisory firms

Again, these recommendations apply to firms across advisory sectors where AI mediation has reached the formative base. European infrastructure is the research site; the institutional logic is general.

1. Protect formation pathways. Make explicit institutional choices about which engagements are routed through formation tracks — where junior consultants engage with the work before AI tools do, where supervisory review is formative rather than editorial. These choices have costs. They produce a different kind of practitioner. The economic research suggests — formally, not just philosophically — that protecting the labour-intensive components of work is the institutional lever that prevents entry-level automation from severing tacit knowledge transmission across generations.

2. Protect organisational knowledge from extraction. The structural pathway through which firm knowledge flows outward into vendor training data is real. Vendor contracts can be negotiated to exclude client materials from training datasets; deployment configurations can be chosen to mitigate the risk. Paris, 2024: an infrastructure advisory managing partner installs a non-negotiable condition in every enterprise AI platform evaluation — no client analytical content in vendor training data, contractually enforced, with audit rights. One vendor can’t meet the condition. The firm takes the more expensive option. Three months later, two clients mention the data governance clause unprompted in contract renewal conversations. The protection isn’t just principled — it’s commercially visible.

3. Treat distinctive firm voice as a strategic asset. Generic AI mediation produces pressure toward average outputs — the statistical centre of the corpus it was trained on. The firm that doesn’t actively cultivate its distinctive way of seeing problems will find it gradually eroded. This requires deliberate institutional investment: in senior practitioners whose analytical voice is recognisably theirs, in practices that reproduce firm-distinctive content across generations, in the recognition that what makes the firm valuable is the content that took twenty years to develop, not the productivity that any AI tool can provide.

4. Develop attribution practices. Clients are starting to ask whether their advisory firm’s deliverables are human-generated, AI-generated, or some combination. Getting ahead of this question through explicit attribution norms is better than being caught without an answer.

5. Engage the formation question at industry level. The formation pipeline is not a firm-level question alone. Chartering bodies, professional associations, training frameworks, and regulatory architectures shape it. Firms that contribute actively to those institutional conversations — rather than treating them as someone else’s problem — are helping build the infrastructure on which their own future practitioners will depend.

What this means for infrastructure clients

Infrastructure clients — utilities, regulators, transport authorities, government departments — are not passive recipients of advisory services. Their procurement decisions, disclosure norms, and capability-building choices actively shape what advisory practice becomes.

1. Procure for judgment, not just for output. When clients procure on price and turnaround time alone, they’re accelerating the Commodity Advisory trajectory. When they procure for the judgment behind the deliverable — the contextual grasp, the institutional knowledge, the relational intelligence that distinguishes one firm’s reading from another’s — they’re sustaining the institutional conditions under which that judgment is reproduced.

What’s being procured, when it’s working, is what Walter Benjamin called “counsel”: practical wisdom that comes from somewhere — from a specific person’s sustained engagement with a specific regulatory tradition. Not information (which any adequately calibrated system can now produce), but counsel. The test is whether the advice bears the marks of genuine formation in the tradition the client’s problem belongs to.

2. Develop disclosure norms. As AI mediation becomes more widespread in advisory work, clients have a legitimate interest in knowing how their work was produced. In infrastructure, where the third box — communities and future generations affected by decisions they didn’t commission — has formal standing, disclosure is an institutional accountability matter, not just a client preference.

3. Think strategically about in-house capability. Vienna, 2025. ÖBB (Austrian Federal Railways) has established an internal AI centre of competence for regulatory and strategic analysis. They’re not attempting to replace external advisory input for major proceedings. They’re attempting to be a more intelligent client — arriving at advisory engagements having completed the preparatory analysis themselves, so the conversation can begin at the strategic question. Advisory firms who work with ÖBB have noticed: initial sessions run shorter, questions arrive harder. One senior partner, after an unusually demanding opening session: They’ve done our homework. Now they want to discuss the exam.

Being a more capable client is legitimate. But clients who build in-house AI capability while systematically eliminating the Foundation and Applied engagement scope through which the wider advisory ecology’s formation pipeline runs are building on ground they’re eroding beneath themselves. The Chartered-stage judgment they’ll need in twenty years is being produced now — or not — in the firms they’re choosing to engage or not to engage.

4. Audit the formation consequences of your capability decisions. Does your in-house development systematically leave only the high-judgment work for external advisers? If so, the pipeline that produces practitioners capable of high-judgment work is being narrowed in ways that affect you.

What this means for European policymakers

These recommendations have the broadest institutional reach, and the most institutional consequence. They’re addressed to European policymakers specifically because the EU institutional architecture offers the most developed available instruments — but the general logic applies wherever public policy has an interest in the professional formation of practitioners who serve public-interest clients.

1. Extend the strategic-autonomy logic from defence to infrastructure advisory broadly. Chapter 5 showed that European policymakers already recognise sovereignty as a strategic concern in defence: domestic procurement preferences, technology transfer requirements, restrictions on foreign ownership. The same underlying logic — some decisions are too consequential to depend on systems and institutions outside European political control — applies in modulated form to energy, water, transport, and urban planning advisory. The extension doesn’t require defence-equivalent architecture. It requires recognising that the same logic operates.

2. Develop procurement architectures that recognise sovereignty. Current EU AI governance operates mostly at the procedural level: data residency, cloud infrastructure, regulatory compliance. Mügge’s analysis showed that EU strategy has treated procedural sovereignty as a proxy for epistemic sovereignty — the genuine empowerment of European practitioners — without delivering it. Procurement frameworks that ask not “where is the data stored?” but “what did this model learn from, and does that make it suitable for European public-interest advisory work?” would engage epistemic sovereignty rather than just procedural compliance.

Paris, 2025. The French Secrétariat général pour l’investissement publishes revised procurement criteria for AI-assisted advisory tools on public infrastructure decisions above a defined threshold. The criteria require that such tools: use training data that doesn’t include confidential French public-sector content without explicit consent; provide explainability of outputs on request; be subject to French data protection law for all data processed. A legal adviser reviewing them describes them as “the EU AI Act applied to advisory procurement, before the AI Act applies to it.” They’re asking not whose tool it is, but what it knows and whether it can be held to account.

3. Engage the formation question as policy. The formation pipeline isn’t a private firm matter. It’s an infrastructure for the profession’s capacity — and the profession’s capacity is an infrastructure for the public-interest decisions infrastructure advisory supports. Support for chartering and accreditation architectures, educational policy that addresses European public-interest professional formation, regulatory architectures that treat the formation question as a policy concern — these are available levers that aren’t currently being used.

4. Deploy existing EU instruments. The AI Act’s high-risk system provisions already apply to AI systems used in critical infrastructure sectors. The Green Deal’s sectoral governance frameworks are already in place for energy, transport, and water. What’s missing is the explicit linkage between high-risk AI deployment in infrastructure advisory and the formation pathway whose disruption such deployment occasions. The instruments exist; the linkage needs to be made.

5. Give the third box institutional standing. The communities and future generations whose lives are shaped by infrastructure decisions are the constituency that makes this argument urgent. They didn’t hire anyone. Often they have no formal voice. But the legitimacy of infrastructure advisory work ultimately depends on whether it’s answerable to them. Explicit institutional architectures through which that standing is recognised — in public consultation requirements, in regulatory documentation, in policy frameworks that take intergenerational responsibility seriously — are available. They need to be built.

The limits of what this thesis has done

Three limits are acknowledged honestly.

Single domain. The framework was built for European infrastructure advisory. The mechanism it describes appears in software and law too, and the general claims are more broadly applicable — but the specific content of the argument (the third box, the public-interest tradition, the European regulatory architecture) is suited to this domain. How the framework extends to other contexts is a residual question.

Conceptual, not empirical. This thesis contains no interviews, no surveys, no longitudinal data. Its contribution is conceptual and normative — what is structurally at stake, what the underlying mechanisms are — not a measurement of exactly how much has already happened. Systematic empirical validation in the European infrastructure context is the most consequential outstanding work.

Philosophical depth deferred. Chapter 3 developed the philosophical framework at a working level adequate for the analytical chapters. A full philosophical treatment would go deeper into Stiegler’s apparatus, Simondon’s account of individuation, Jonas’s relationship to Heidegger and Ellul, and the philosophical ontology of large language models. That deeper framework was deliberately deferred. It’s a limit, and a direction for further work.

Five questions for further work

1. Empirical validation. Do the formation patterns described here show up in systematic interviews, longitudinal studies, and comparative research across European infrastructure firms? The conceptual argument needs empirical testing.

2. Comparative work across contexts. How does the framework extend to strategy consulting, financial advisory, and non-European contexts? Which parts of the argument are general and which are specific to this domain?

3. The deeper philosophy. What does the fuller philosophical apparatus — Stiegler’s wider framework, Simondon on individuation, Jonas’s relationship to earlier technology critics — add to the working-level framework deployed here?

4. Operational specifics. The recommendations are principled, but not operational. What exactly should a procurement framework that recognises sovereignty look like? What should formation pathway protection standards say? Working these out in collaboration with the relevant institutional actors is necessary work.

5. Longitudinal validation of the long-horizon predictions. The most consequential prediction — that formation disruption now propagates into the senior practitioners of two decades from now — cannot be tested in any single research programme. It requires tracking cohorts over decades. It’s the residual question with the longest horizon, and the most institutionally important.

Closing

The European public-interest tradition in infrastructure — the orientation toward serving populations, not just clients; the recognition that a railway or a water system or an electricity grid answers to the people who depend on it — has been one of the post-war period’s achievements. It’s produced infrastructure decisions that have shaped how people in Europe live across generations.

The advisory profession that serves this tradition has been one of the institutional architectures through which its values are exercised in practice. The practitioner who reads a regulator’s hesitation and recognises what public interest it’s protecting. The senior consultant who tells a client that the technically correct answer is insufficient given what it would mean for the communities bearing the consequences. The formation system that produces practitioners capable of those judgments.

Whether that formation system survives the transformation that AI mediation has begun to produce is the question this thesis has tried to make visible. The answer is not given. It depends on the choices that practitioners, firms, clients, and policymakers make — many of which are being made now, often without explicit recognition that a formation question is at stake.

The trajectory is not determined. The conditions under which it resolves can be engaged. The engagement requires seeing clearly what is at stake.

That is what the thesis has tried to provide.