Chapter 8 — Four Possible Futures (plain-language version)

Why think in scenarios

Nobody knows how European infrastructure advisory will look in twenty years. That uncertainty doesn’t mean thinking about the future is pointless — it means thinking about it carefully requires a different approach than prediction.

This chapter constructs four scenarios: not forecasts, not most-likely projections, but analytical thought-experiments that ask if the world goes this way, what happens? Each scenario is a combination of two key variables: where does expertise live (in human practitioners or in AI systems?), and who controls the AI systems (commercial platforms operating under Anglo-American logic, or alternatives — European public-interest bodies, firm-owned systems, something else)?

The combinations produce four quadrants. Running the framework’s analysis through each one produces a picture of what the profession looks like under those conditions — and more importantly, identifies what concerns persist across all four scenarios, not just one.

Those persistent concerns are what Chapter 9’s recommendations are designed to address. Recommendations suited to only one scenario are fragile. Recommendations suited to concerns that show up everywhere are more useful.

Scenario 1: 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. The work passes quality checks. Nobody has said out loud what the graduate is not learning.

In this scenario, expertise increasingly concentrates in AI systems rather than in human practitioners — and those AI systems are owned and operated by large commercial platform vendors, mostly American, whose priorities are commercial rather than public-interest.

What it looks like. Generic AI mediation matures across most components of advisory work. The Foundation-stage pipeline collapses — junior analytical work is absorbed before it can form anyone. Applied-stage work is commoditised — firms compete on price and turnaround time, with distinctiveness compressed into a narrow band at the senior level. Chartered-stage advisory narrows — senior consultants still add value, but the scope of what they’re engaged on shrinks as everything else gets handled by the platform.

Milan, 2025. 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 the particular character of the Italian regulator ARERA, 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. One partner prepares a presentation for the firm’s annual strategy day titled: The market is pricing our distinctiveness at zero. Nobody disputes the diagnosis. Nobody has a confident answer.

Why it’s the most concerning trajectory. This is where current conditions are pointing, in the absence of deliberate institutional engagement. The formal economic model developed by Acemoglu, Kong and Ozdaglar predicts that under unchecked AI substitution, a professional community can tip into a “knowledge-collapse steady state” where collective expertise vanishes — even as individual outputs continue to look competent. The work passes quality checks. Nobody has said out loud what the graduate is not learning.

Scenario 2: 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.

In this scenario, expertise still concentrates in AI systems — but different AI systems. European public-interest AI capabilities, firm-proprietary AI tools trained on firm-distinctive material, governance architectures that recognise what data is flowing where.

What it looks like. The technical transition is similar to Commodity Advisory, but the institutional architecture around it is European. AI tools are trained on European regulatory data, governed under European frameworks, owned through institutional structures that recognise public-interest purposes. Firms might develop proprietary AI capabilities that preserve their distinctive analytical approach rather than defaulting to generic platforms.

Brest, 2024. A French port authority has commissioned a digital twin of its logistics infrastructure under an EU-funded port digitalisation programme, built on a platform developed by a French technology consortium. The expertise of an 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 judgment, and how the twin’s advice should be documented for regulatory purposes. The engagement is about the epistemic infrastructure for governing port infrastructure — a distinction the firm’s engagement scope had not previously needed to make.

The complication. This scenario requires substantial institutional development that doesn’t yet exist — European AI capability of this kind, governance frameworks that work, procurement architectures that specify these conditions. The scenario is achievable, but it requires deliberate political and institutional choices. Chapter 9’s recommendations are partly aimed at creating the conditions for this trajectory.

Scenario 3: 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. The firm charges a premium. It has never lost a client who genuinely valued the distinction.

In this scenario, expertise remains concentrated in human practitioners. AI tools are present and used, but they’re bounded — integrated into the formation system rather than replacing its core components. Formation pathways are protected. The European public-interest tradition is preserved.

What it looks like. Foundation-stage formation still happens through engagement with consequential work, because the firm has made institutional choices to protect it. Applied-stage work still demonstrates distinctive analytical voice, because the practitioners exercising it have been genuinely formed. Chartered-stage advisory still has the depth it requires, because the pipeline that produces Chartered practitioners hasn’t been ruptured.

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 doesn’t use AI-mediated drafting for deliverable content. It uses AI extensively for background research, data aggregation, and document search. The distinction is not a refusal of technology; it is a formation and quality assurance commitment. The firm 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.

The formal-model convergence. The Acemoglu et al. model identifies what they call a welfare-optimal policy: a period of restricting AI substitution to rebuild the collective knowledge stock, followed by a permanent cap at a level that supports rather than depletes expertise. Artisanal Advisory is the political-economic analogue of what that policy would produce — a sustained condition in which AI tools operate within institutional boundaries that protect formation.

Scenario 4: 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 advisers who have gone in-house. The platform vendor manages the architecture. The utility’s board has not asked who owns the model weights.

In this scenario, expertise remains with human practitioners — but increasingly those practitioners are inside infrastructure clients rather than in advisory firms. Clients build AI capability in-house, absorb the analytical components of what they used to commission externally, and continue to engage advisory firms only for the narrowest senior judgment work.

What it looks like. Sophisticated clients — major energy utilities, developed transport authorities, well-resourced regulatory bodies — build internal AI platforms trained on their own regulatory histories, past submissions, and accumulated data. What they used to pay advisory firms to do, their internal platforms now do. They still engage advisers, but only for the conversations their platforms can’t have.

Lisbon, 2024. ANACOM, the Portuguese regulatory authority for telecommunications, has deployed an AI-supported regulatory analysis platform. An advisory firm that has provided ANACOM with regulatory intelligence briefings for seven years has seen its engagement scope compress: the quarterly briefings, once analytical documents, are now primarily concerned with questions the platform doesn’t answer — the relationship between ANACOM’s formal regulatory positions and the political economy of the current government’s infrastructure priorities. The relationship persists. The firm is smaller. ANACOM is analytically more capable than it was three years ago, and the advisory firm is providing a different kind of advice to a more informed client.

The hidden risk. This scenario appears more benign than Commodity Advisory — expertise is still in human hands, just different hands. But the institutional architecture of the consulting profession is quietly being dismantled. If advisory firms are only engaged for the narrowest senior judgment calls, the pipeline that produces practitioners capable of those calls shrinks. And the knowledge that’s been built into the client’s in-house platform? That’s now owned partly by the platform vendor — whose terms of service the utility’s board hasn’t thought to question.

Seven findings that hold across all four scenarios

Four scenarios, two structural axes. But when you run the framework’s analysis through all four, seven concerns appear in every quadrant, regardless of which trajectory unfolds.

Finding 1: The contributory/interactional distinction is what determines professional value. In every scenario, what AI mediation can replicate at scale is interactional expertise — the ability to speak the profession’s language, produce competent-looking outputs, engage practitioners in their own terms. What it cannot replicate is contributory expertise — knowledge built through sustained engagement with consequential work that enables genuine independent contribution. This distinction is the load-bearing beam in the entire framework. Everything else depends on it.

Finding 2: The formation question becomes inescapable. In Commodity Advisory, formation pathways collapse. In Digital Twin Economy, they’re altered to operate within AI-mediated systems. In Artisanal Advisory, they’re actively protected. In Disintermediated Clients, they compress under reduced engagement scope. Every scenario forces a choice about formation. The choice can be made deliberately or by default, but it’s always being made.

Finding 3: Sovereignty pressure operates across multiple scales. In every scenario, the question of whose cultural assumptions are embedded in the AI tools that advisers use, and who owns those tools, matters for what European infrastructure advisory becomes. The national-cultural scale (does European public-interest practice survive or get absorbed into Anglo-American defaults?) and the organisational scale (does the firm’s distinctive analytical voice survive or get homogenised?) operate in all four scenarios, with different intensities.

Finding 4: The third box question persists. Across all four scenarios, the communities and future generations whose lives are shaped by infrastructure decisions — who didn’t hire anyone and often have no formal voice in the advisory process — are either engaged or not. The advisory work’s legitimacy ultimately depends on whether it’s answerable to them. That answerability operates differently in each scenario, but it never disappears as a question.

Finding 5: Explicit legitimacy narration becomes inescapable. In every scenario, the implicit legitimacy that has historically come from the profession’s institutional architecture — from the assumption that someone who’s been practising for twenty years and carries a senior title has earned that title — is insufficient. When interactional fluency is widely available, legitimacy has to be narrated: what this practitioner’s actual contribution is, what AI was used for, what the human judgment was that the client is paying for.

Finding 6: Extraction operates structurally in all four scenarios. The variable isn’t whether firm knowledge flows outward into the generic AI landscape — it does, in all scenarios. The variable is through what institutional architecture and under whose ownership. In Commodity Advisory, extraction is unmediated and accelerating. In Artisanal Advisory, it’s actively contested through deliberate institutional choices. In Digital Twin Economy, it operates through technically governed channels. In Disintermediated Clients, it’s client-side. But in every scenario, the dark mirror of the knowledge cycle is operating: what firms know tends to flow outward. The question is only how fast, and who captures it.

Finding 7: The Applied stage has no safe trajectory in any scenario. At every turn, Applied consultants — the mid-career practitioners who bridge Foundation formation and Chartered judgment — are under specific pressure. In Commodity Advisory, their deliverable moat is commoditised. In Digital Twin Economy, they become system-integration and quality-assurance functions. In Artisanal Advisory, they’re the explicitly defended constituency — which itself signals they’d otherwise be at risk. In Disintermediated Clients, they’re the most directly disintermediated. The Applied stage occupies a structurally exposed position, caught between the automation of Foundation work and the continued value of Chartered judgment, without a stable landing zone.

What the scenarios establish

These aren’t probability assessments. No claim is being made that Commodity Advisory is more likely than Artisanal Advisory, or that any specific scenario will happen.

What the scenarios establish is that the seven concerns above are robust to trajectory uncertainty. Chapter 9’s recommendations are designed for those seven concerns — not for one anticipated future, but for the structural challenges that show up everywhere.

The trajectory is not fixed. The conditions under which it resolves are being shaped now, by the decisions practitioners, firms, clients, and policymakers are making.