This is an ENSI Foresight Division analysis, the third in a sequence. Report №01 made the growth case; Report №02 made the agentic-operations case. This report fuses them: it argues that a country's single highest-return investment is a standing foresight capability — and that AI agents are what finally make such a capability continuous, affordable, and whole-of-government in scope.
Scan → model → stress-test → route-to-decision. A continuous loop that reads the global signal space, builds and refreshes scenarios, attacks every plan, and pushes the relevant slice into the decision being made anyway — the budget, the legislative programme, the security review.
Human-in-the-loop, audit, provenance. Agents widen the aperture; humans keep the accountability. Every signal is traceable to source, every judgement is owned by a named officer, and every model run is logged — so the capability earns trust rather than assuming it.
| Depth | Time | What you get | Where |
|---|---|---|---|
| Depth 1 | 90 seconds | The whole thesis — cover, executive summary, the dashboard, the engine in one diagram, twelve findings. | §1 · §4–§7 |
| Depth 2 | 20 minutes | The nine analytical angles — the report's main contribution, where the understanding actually forms. | §8–§16 |
| Depth 3 | reference | The 32-area catalog — each a seven-part operating brief, grouped into four tiers of leverage. | §17–§20 |
| Depth 4 | the proof | Cases, the 12-month build roadmap, the steelman of objections, methodology, the 188-source library. | §21–§26 |
Front-loaded by design: a reader who never reaches the catalog still owns the full thesis. Movements A and B carry the intellectual payload; Movement F then specifies the engine itself — the Agentic Foresight Architecture, 26 components in four layers.
Most governments are run as if the future were a rumor. Budgets are set one year at a time, ministers are rewarded for what lands inside an electoral cycle, and the slow, certain forces that actually decide a country's fate — ageing, the energy transition, the arrival of machine intelligence, the drift of supply chains — are treated as someone else's problem until they arrive as a crisis. Then the same governments discover, expensively, that the future was knowable all along. The pandemic was on every risk register. The energy shock was in every scenario. The skills gap was a demographic certainty twenty years out.
The reframe that organises this whole report is simple and, in practice, radical: foresight is not a report, it is a capability — a standing function that continuously scans, models, and stress-tests the country's possible futures, and routes what it learns into the decisions being made anyway. The states that have built this — Singapore, Finland, the UAE, Canada — did not get better forecasts than everyone else. Forecasts are usually wrong. What they got was option value: the ability to see a shift early, to have already rehearsed the response, and to act while action is still cheap.
Why a country, and not a company? Because the state is the only actor with the time horizon, the breadth, and the obligation to think in decades. A firm that misreads the future loses money and exits; a country that misreads the future loses a generation. And foresight's payoff is highest exactly where markets are blindest — in long-lived infrastructure, in public goods, in systemic risk, in the slow accumulation of national capabilities. That is why a foresight capability is a growth strategy, a stability strategy, and a quality-of-life strategy at once: it lets a country invest ahead of demand, absorb shocks rather than be broken by them, and choose the future it wants rather than inherit the one it failed to see coming.
Foresight is not the art of predicting the unpredictable; it is the discipline of refusing to be surprised by the predictable — and of staying nimble in the face of the genuinely uncertain.
What is new — and what makes this an ENSI document rather than a restatement of classical futures studies — is the agentic engine. Traditional foresight is bottlenecked by human attention: a handful of analysts can scan only so many signals and refresh them only so often, so foresight arrives as an occasional, expensive set-piece.
AI agents dissolve that bottleneck. A standing fleet of scanning agents can read the entire global information space every day; scenario agents spin and stress-test hundreds of futures; simulation agents run agent-based models of the economy, the labour market, or an epidemic; red-team agents attack every plan; and briefing agents deliver the relevant slice to the right decision-maker at the moment of decision. Foresight stops being a quarterly artefact and becomes a continuous, living nervous system for the state — with humans keeping every judgement and every accountability.
The 32 areas are ranked by leverage and grouped in four tiers. Tier I is the foresight core — the central capability plus the macro-fiscal, growth, technology, labour, demographic, energy and risk functions every other decision hangs from. Tier II is productive capacity and competitiveness. Tier III is society, resilience and quality of life. Tier IV is environment, security, and the enabling layer — the data architecture and the futures-literacy the whole system runs on. The ranking is not just a list; it is a sequencing instruction: build the engine once, on a shared data backbone, staffed by people who can think in futures, then point it in turn at each tier.
Twelve figures and two charts that carry the argument before a word of analysis. Facts we own are tagged ENSI; institutional figures carry their source.
One continuous loop — scan → model → stress-test → route-to-decision — run by an agentic layer and pointed, in turn, at the four tiers of national life. Build the engine once; apply it 32 ways.
The argument compressed to twelve one-sentence claims, each carrying the evidence chip that backs it.
Budgets are set one year at a time. Ministers are rewarded for what shows up inside an electoral cycle. The long, slow, certain forces that actually decide a country's fate are treated as someone else's problem — until they arrive as a crisis, and the state discovers, expensively, that the future was knowable all along.
The pandemic was on every risk register. The energy shock was in every scenario set. The skills gap was a demographic certainty visible twenty years out. None of these were unknowable; all of them were unattended. The failure was not of forecasting but of posture — a state organised to react to the present cannot see the future even when the future is plainly written down. Ageing, the energy transition, the arrival of machine intelligence, the drift of supply chains: each moves on a clock measured in decades, and each is routinely ignored on a clock measured in months.
This is not a failure of individual ministers. It is a structural property of how a government allocates attention. The annual budget rewards the extrapolation of the year that just ended. The legislative programme rewards what can be delivered before the next election. The career incentive rewards the absence of visible failure on your watch, not the avoidance of a crisis that lands on your successor's. The predictable, slow-moving threat is precisely the one no part of the machine is built to act on while action is still cheap — and so the cheap window closes, unnoticed, every single time.
Foresight is not the art of predicting the unpredictable. It is the discipline of refusing to be surprised by the predictable — and of staying nimble in the face of the genuinely uncertain.
That sentence does most of the work in this report. It splits the future into two halves and assigns each a different obligation. The predictable half — demography, asset lifespans, diffusion curves, fiscal arithmetic — carries an obligation to provision early, because the cost of acting in time is a fraction of the cost of reacting late. The genuinely uncertain half — shocks, tipping points, technological discontinuities — carries a different obligation: to stay rehearsed and nimble, so that when one branch resolves, the response is already on the shelf. A state that confuses the two — demanding certainty before it acts on the predictable, or freezing before the uncertain — fails in both directions at once.
Every comfortable belief a government holds about the future is paired with a mechanism that quietly contradicts it. The myths are reassuring because they license inaction; the mechanisms are inconvenient because they demand provisioning ahead of demand. Naming both, side by side, is the first act of foresight — because you cannot stress-test a plan whose hidden assumption you have never said out loud.
| The comfortable myth | The mechanism that actually operates |
|---|---|
| "The future is unknowable, so planning for it is guesswork." | Most of what matters is not unknowable — it is unattended. Ageing, asset lifespans and diffusion curves are forecastable to a tight range; the failure is acting on them, not knowing them. |
| "We'll deal with it when it arrives." | By arrival the cheap options have expired. A second supplier, a trained cohort, a retrofitted grid take years; the shock takes days. Reaction buys the worst options at the highest price. |
| "Forecasts are usually wrong, so foresight is a waste." | True — and beside the point. The payoff is not an accurate forecast but option value: seeing the shift early, having rehearsed the response, acting while action is still cheap. |
| "Foresight is a report we commission every few years." | A report ages the day it ships and routes nowhere. The thing that works is a standing capability — continuous scanning, modelling and routing into decisions being made anyway. |
| "The risk register covers us." | A static binder reviewed annually is contradicted by the next surprise. Real crises cascade and compound; what's needed is a live system that catches interactions a register lists separately. |
| "Strategic sectors are obvious; we know our strengths." | Comparative advantage migrates. Yesterday's strength is often nostalgia dressed as strategy; the reachable high-value rung is defined by where demand and technology will be in fifteen years. |
| "This is for big states with slack to spare." | The asymmetry is sharpest for a mid-sized state with little fiscal slack: you cannot buy your way out of a crisis you failed to imagine. Anticipation is the affordable option, not the luxury one. |
The rest of this analysis takes the reframe at its word. If foresight is a capability rather than a report, then the next question is mechanical: what is the capability made of? Not which crystal ball, but which engine — what does it ingest, how does it process, and where does its output go so that it changes a decision rather than decorating a shelf. That is the subject of the angle that follows.
Singapore's risk-and-horizon machinery, Finland's megatrend work, Canada's disruption scanning — they look different on the surface and run the identical loop underneath: scan → model → stress-test → route-to-decision, refreshed continuously rather than commissioned occasionally. Learn the engine once and you can aim it at all thirty-two domains.
A report is a line: it ends. A capability is a loop: each decision and each new signal feed the next scan, so the picture is never stale and the response is always rehearsed. The loop is what converts forecasting — which is usually wrong — into option value, which compounds.
The same engine produces wildly different value depending on where it is aimed. The thirty-two applications are ordered by leverage — how much getting foresight right in that domain changes everything downstream — and grouped into four tiers. The order is not decoration; it is the build sequence. Get Tier I right and every tier above it inherits the advantage; get it wrong and no downstream foresight can compensate.
| Tier | What it governs | The contribution, and why it sits here |
|---|---|---|
| I | The Foresight Core Areas 1–8 | The central capability plus the macro-fiscal, growth, technology, labour, demographic, energy and risk functions — the spine every other decision hangs from. The engine room. |
| II | Productive Capacity Areas 9–16 | Innovation, trade, digital, financial, infrastructure, materials, the firm base and investment — whether the economy keeps climbing the complexity ladder. |
| III | Society & Quality of Life Areas 17–24 | Health, education, ageing, migration, cohesion, regional balance, housing and wellbeing — where foresight most directly touches how people actually live. |
| IV | Environment, Security & Enablers Areas 25–32 | Climate, food and water, nature, defence, cyber, regulation — and, load-bearing for all the rest, the data architecture and the human futures-literacy the whole system runs on. |
Classical foresight is bottlenecked by human attention: a handful of analysts can scan only so many signals, build only so many scenarios, and refresh them only so often, so foresight arrives as an occasional, expensive set-piece. A standing fleet of agents dissolves that bottleneck — and turns the same loop into a daily organ of the state. Five archetypes run the engine; humans curate, judge and own what goes up.
Scanning agents read the global information space every day and flag trend-breaks the moment they cross a threshold. Scenario-generation agents spin and refresh hundreds of futures and pressure-test the canonical set for staleness. Simulation / ABM agents run the economy, labour market or an epidemic across synthetic populations. Red-team agents adversarially attack every plan, hunting the assumption that breaks it. Briefing & early-warning agents compress it all into the decision-shaped brief and fire when a leading indicator crosses a line.
A foresight capability is a modest standing cost — a few dozen people and a fleet of agents — set against the large, lumpy, unbudgeted cost of reacting late. The entire economic case is one asymmetry: the cost of acting early is almost always a fraction of the cost of reacting late, and that gap, compounded across thirty-two domains, is measured in points of GDP.
The mistake most governments make is to treat foresight as a cost to be justified against an uncertain benefit. It is the opposite. Foresight is the purchase of option value — the right, but not the obligation, to act early on a future that may or may not arrive. Like any option, it is cheap relative to the loss it insures against, and its value is highest exactly where uncertainty is highest and the underlying stakes are largest. A country that holds the option can move while moving is still cheap; a country that does not must buy the underlying outright, at crisis prices, on the worst possible day.
The arithmetic is an arithmetic of asymmetries. In each case the early cost is small, recurring and chosen; the late cost is large, lumpy and forced. Three concrete shapes recur across the catalog:
| The predictable shift | Acted on early (cheap, chosen) | Reacted to late (lumpy, forced) |
|---|---|---|
| A skills gap visible a decade out in the demographic and exposure data | Redirect training places and apprenticeship pipelines toward forecast demand — a budget reallocation, not new money | Becomes structural mass unemployment — transfer payments, lost output, a region in decline, a political emergency |
| A risk sitting quietly on the register with a faint early signal | A day of warning and a rehearsed response — preparedness at modest standing cost | Becomes a national crisis — emergency spending, cascading failure, the catastrophic cost of surprise |
| An asset built for a world that is ending — a chokepoint, a fuel, a plant | Stage the bet, qualify a second source, keep the option to switch — real-options discipline | Becomes a stranded asset — capital written off, a dependency weaponised, a decade of lock-in |
The instinct is that anticipation is a luxury for large, rich states with slack to spare. The reverse is true. The asymmetry between early and late cost is sharpest precisely where fiscal slack is thinnest, because a state that cannot afford to buy its way out of a crisis cannot afford to be surprised by one. For an open, mid-sized economy embedded in larger value chains, a distant chokepoint becomes a domestic factory standstill within weeks, and a demographic spending wave that was visible for thirty years arrives with the budget already committed elsewhere. Anticipation is not the expensive option here. It is the only affordable one.
Notice what the option-value frame does to the usual debate. The objection to foresight is always "but forecasts are wrong" — and the objection is correct and irrelevant. You do not buy an option because you are certain the future arrives; you buy it because you are not. The value lives in the asymmetry of payoffs, not the accuracy of the prediction. A foresight capability that is wrong about which scenario lands, but right that some scenario will, still pays for itself many times over — because it kept the response cheap, staged and reversible while everyone else was committing.
A foresight capability is a lean directorate of eight to twelve, a signals registry, and a fleet of agents that does the high-volume scanning and modelling no human team could. It is a recurring line item measured in the low millions — set against contingent liabilities, stranded assets and crisis spending measured in points of GDP. The cost side of the ledger is the easy side.
The expensive surprises are never the headline number. They are the off-balance-sheet liability nobody priced, the chokepoint nobody mapped, the cohort nobody trained — each a predictable force met with a posture built only to react. Foresight moves those costs from "forced and large" to "chosen and small."
You cannot afford to buy your way out of a crisis you failed to imagine. Imagining it early is the cheapest line in the budget.
Read the catalog end to end and a reader expects thirty-two separate programmes — a fiscal model here, an energy dashboard there, a skills observatory somewhere else. That reading is wrong, and the mistake is expensive. Beneath the thirty-two briefs runs a single machine. Only the domain changes; the engine, the backbone, the agents and the horizons are the same. Build it once and every new area is a low-cost extension of a system that already works.
The thirty-two areas were written to be read independently, each a complete seven-part operating brief. But they were designed to share a chassis. The same loop — scan → model → stress-test → route to decision — turns in Area 2's debt fan charts and Area 25's flood maps alike. The same data backbone (Area 31) feeds them all; the same human futures-literacy (Area 32) decides whether anyone acts on what they say. The same fleet of agents widens the aperture in every domain. And the same three horizons — near, medium, long — recur, decision for decision, whether the subject is a winter adequacy margin or a fifty-year settlement pattern. The deep regularities are not a coincidence of style. They are the thesis.
Once you see the regularities, the prioritization stops looking like a ranked list of worthy projects and starts looking like what it is: a sequencing instruction for one build. You do not staff thirty-two teams. You staff one engine and point it, in turn, at the macro spine, at competitiveness, at society, and at the country's planetary and security boundaries. The five patterns on the facing pages are the load-bearing structure under the whole catalog.
The same engine — scan, model, stress-test, route to decision, repeat — runs in every one; only the domain changes.
If foresight were thirty-two unrelated projects, no single property would hold across them. It does. Map the five cross-cutting patterns against the four tiers and the cells fill in — not by accident, but because each area is the same engine wearing a different domain's clothes.
Classical foresight is rationed by human attention. A handful of analysts can scan only so many signals, build only so many scenarios, and refresh them only so often — so foresight arrives as an occasional, expensive set-piece, a workshop every quarter. That scarcity is the reason most governments treat the future as a rumor: not because anticipation is hard, but because it was expensive to do continuously.
A standing fleet of agents removes the constraint. Scanning agents read the global information space every day; scenario agents spin and stress-test hundreds of futures; simulation agents run agent-based models of an economy or an epidemic; briefing agents deliver the relevant slice to the right desk at the moment of decision. Foresight stops being a quarterly artifact and becomes a continuous nervous system — the shift Angle 6 develops in full.
Open any brief and the horizons line up: a near band (0–2 years) of operational moves already in play; a medium band (2–6 years) where curricula, reforms and permits are decided; and a long band (6–15 years and beyond) where capital, infrastructure and structural identity are committed. The classic policy error is confusing them — funding short retraining for a structural shift, or planning grand reform for a cyclical dip.
The countries that built this capability did not get better forecasts — forecasts are usually wrong. What they got was the ability to see a shift early, to have already rehearsed the response, and to act while action is still cheap. That is option value, and it is the payoff in every single brief: a skills gap caught before it becomes mass unemployment, a risk caught before it becomes a crisis, an asset caught before it becomes a stranded liability.
The prioritization is not just a ranking of worthy domains — it is a build sequence. A country starting from zero should stand up the core engine and its data and human foundations first, then extend into the high-leverage macro functions, and only then into society and the planetary boundaries. Skip the foundations and you do not get a faster foresight state. You get isolated reports nobody acts on.
There is a tempting mistake available to every government that discovers foresight: to commission the exciting, visible work first — the climate-adaptation scenario, the pandemic playbook, the AI-and-jobs study — because those are the topics that make headlines and win ministerial enthusiasm. It is the wrong order, and it fails predictably. A society-tier or environment-tier study built without the engine has nowhere to route its conclusions, no shared data backbone to keep its indicators live, and no futures-literate officials to act on it. It becomes a beautiful document that gathers dust — foresight theatre, not foresight capability.
The dependency logic runs the other way. Four areas form the minimum viable foresight state: the national foresight capability at the centre of government (Area 1), national risk and crisis anticipation (Area 8), the data architecture and decision-intelligence platform (Area 31), and futures literacy and capability-building (Area 32). The engine, its early-warning organ, its data spine, and its people. Stand those up and every subsequent domain is a low-cost extension of a system that already works — the scenario machinery, the agent fleet, the routing into the budget and the legislative cycle are all already running; a new area just points them at a new subject. Attempt the same domain without the foundations and you are rebuilding the whole apparatus, badly, thirty-two times.
Build the engine first and each new domain is a low-cost extension of a system that already works; attempt the society and environment tiers without the engine and you get isolated reports nobody acts on.
The four tiers are not parallel options to fund in any order. They are a dependency chain. Each stage inherits the engine, the data and the people from the stage before — and pays a falling marginal cost for every area it adds.
This is the move that makes the document an ENSI document rather than a restatement of classical futures studies. Traditional foresight is bottlenecked by human attention, so it arrives as an occasional, expensive set-piece. A standing fleet of agents dissolves that bottleneck — and in doing so changes foresight from something a country has into something a country continuously does. But the corollary is sharp: when the machine never sleeps, discipline matters more, not less.
A handful of analysts can scan only so many signals, build only so many scenarios, and refresh them only so often. That is why, for most of its history, foresight has been a quarterly workshop and a binder on a shelf. The agentic layer breaks the constraint at its root. Agents can read the entire global information space every day, spin hundreds of scenarios, run agent-based simulations overnight, attack every plan, and deliver the relevant two pages to the right minister at the moment of decision. The set-piece becomes a standing organ of the state — always on, always current, embedded in the decisions that are being made anyway.
What follows are the five agent archetypes that recur, in different combinations, in all thirty-two areas. Each does something that scarce human analysts structurally never could — not because the humans were less capable, but because attention was always rationed and now it is not. The humans do not disappear; they move up the stack, from doing the scanning to judging what the scanning found.
Foresight stops being a quarterly artifact and becomes a continuous, living nervous system for the state. Agents widen the aperture; humans keep the accountability.
The same five roles staff every area's engine. Read them as the standing crew of the foresight state — the watch, the imaginers, the modellers, the adversary, and the messenger.
Ingest the entire global information space — preprints, patents, funding rounds, vacancy text, trade and sensor feeds — every day, and flag trend-breaks the moment they cross a threshold. What humans never could: read everything, continuously, and never tire of the boring 99% where the weak signal hides.
Recombine signals into fresh narrative branches, spin and stress-test hundreds of futures, and pressure-test the canonical scenarios for staleness. What humans never could: explore the combinatorial space of futures at a breadth no workshop can reach, then prune to the plausible.
Run agent-based models of households, firms, regions, labour markets or an epidemic — testing how a policy or a shock propagates through the real system. What humans never could: run thousands of behavioural paths nightly and refresh the fan chart as the world moves.
Adversarially attack every major plan, hunting the assumption that, if false, breaks it — and generate the compound shocks no one war-gamed. What humans never could: attack every plan, relentlessly, without the institutional reluctance to embarrass a colleague.
Track the leading indicators that say which scenario the country is sliding into, fire calibrated alerts with lead time to act, and compress the whole picture into the two-page anticipatory brief the right decision-maker needs at the moment of choice — each tailored to the decisions that owner actually controls. What humans never could: hold the cross-hazard picture in one head, around the clock, and deliver its relevant slice to thirty desks at once without it going stale on the way.
An always-on agent fleet can be a continuous nervous system — or a continuous source of confident, unaccountable error at scale. The difference is not the model. It is the discipline wrapped around it. A foresight state that automates the scanning but not the governance has not built a capability; it has built a liability that briefs the prime minister.
The agentic shift does not relax the need for rigour — it raises it. A human analyst who makes a judgement makes one, slowly, with a name attached. A fleet of agents makes thousands, fast, and the temptation is to trust the output because it is fluent and arrived on time. That is exactly the failure mode to design against. The five guardrails below are the standing discipline that keeps an always-on system trustworthy — and they are non-optional precisely because the system is continuous.
Agents change foresight from a quarterly set-piece into a continuous nervous system — and that is exactly why the discipline has to be tighter, not looser. The imagination and the judgement stay human.
The 32 areas are not equal candidates for early investment. The first move belongs where two conditions meet: leverage is highest — getting foresight right there changes everything downstream — and institutional maturity is lowest, so the standing capability does not yet exist to be improved. Plotted on those two axes, the four tiers stop being a ranking and become a map of where the marginal foresight euro buys the most future.
The catalog ranks the 32 areas by leverage alone, and that is the right ordering for a reference layer. But leverage is only half of a start-here decision. A domain can be enormously consequential and already well-served — a country with a strong central bank does not need to stand up financial-stability foresight from scratch. The decision rule is the gap between how much the domain matters and how much capability is already in place to anticipate within it.
The 2×2 below makes that gap visible. The vertical axis is leverage / impact: how far a foresight advantage in this domain propagates into every decision taken downstream. The horizontal axis is institutional maturity / readiness: how much standing anticipatory capacity the typical mid-sized European state already has in the domain. The first foresight dollar belongs in the top-left quadrant — high leverage, low maturity — because that is where the same money moves the most and meets the least existing structure.
| Tier | Why it ranks here | The first move |
|---|---|---|
| Tier I — Core (1–8) The engine room |
Highest leverage of all: the central capability plus the macro-fiscal, growth, technology, labour, demographic, energy and risk spine. Get these right and every other tier inherits the advantage; get them wrong and no downstream foresight can compensate. | Stand up Area 1 at the cabinet centre — a Chief Foresight Officer, a lean directorate, the national scenario set and signals registry. Then wire Areas 2 and 8 (fiscal, risk) into the budget and crisis machinery. |
| Tier II — Competitiveness (9–16) Climbing the ladder |
Determines whether the economy keeps climbing the complexity ladder — innovation, trade, digital, finance, infrastructure, materials, the firm base, investment. High leverage, but most states already hold fragments (a statistics office, a central bank, an investment agency). | Connect, don't rebuild: anchor industrial-strategy and supply-chain foresight (Areas 3, 10) to the centre's shared scenario set so national bets compound rather than contradict. |
| Tier III — Society (17–24) How people live |
Health, education, ageing, migration, cohesion, regional balance, housing, wellbeing. Leverage is real but the time-constants are long and the existing institutions (schools, hospitals, pension funds) are dense — foresight here deepens an established system rather than creating one. | Seed selectively where a demographic certainty is already locked in — pension and long-term-care sustainability, the depopulating-region early-warning map — using the centre's shared population baseline. |
| Tier IV — Enablers (25–32) Boundaries + backbone |
Climate, food/water, nature, defence, cyber, regulation — and, load-bearing for everything, the data architecture (31) and futures-literacy (32). These two look like the bottom of the list but are the substrate the whole system runs on: a scenario is only as credible as the data feeding it. | Treat data architecture as strategic infrastructure from day one — stand up the data-mesh for three priority domains in parallel with Area 1, not after. Defence/cyber foresight wires into the national-security secretariat. |
They built standing capability and wired it into decisions. Singapore, Finland, the UAE, Canada, the EU and the UK do not out-predict everyone else — forecasts are usually wrong everywhere. What separates them is structural: a permanent unit, a fixed cadence, and a hard route from the foresight output into the budget, the legislative programme or the national strategy. The lesson is never the method. It is the plumbing.
Each exemplar below is profiled on three things, because those three are what actually transfers across borders: what they built (the standing institution), how it routes into decisions (the wiring that stops a report dying on a shelf), and the one lesson a country starting today should copy. The institutions and their signature outputs are drawn from ENSI's source library of the world's leading foresight bodies.
Built. The CSF, inside the Strategy Group of the Prime Minister's Office, and the Scenario Planning Plus / RAHS toolkit — a standing futures function at the very centre of the state, now past its fifteenth year.
Routes in. Foresight feeds directly into whole-of-government strategic planning; the biennial Foresight journal and scenario sets are shared infrastructure that ministries plan against, not advisory papers.
The one lesson. Put it at the centre, keep it permanent, and make the scenario set a shared asset the whole of government uses — once, well, rather than thirty times badly.
Built. A three-part system: Sitra (the independent futures fund and its Megatrends), the parliamentary Committee for the Future, and the statutory Government Report on the Future each electoral term.
Routes in. The Government must table a foresight report and Parliament's own committee must respond — foresight is written into the constitutional rhythm of governing, not bolted on.
The one lesson. A statutory hook beats good intentions. When law requires a future-facing report and a parliamentary answer, foresight survives the minister who championed it.
Built. A sovereign futures agency (DFF) with its own mandate, the Museum of the Future as a public anchor, and signature outputs such as The Global 50 future-opportunities report.
Routes in. Foresight is fused with delivery — opportunity-scanning flows into national strategies, accelerators and procurement, treating the future as a pipeline of bets to place, not a forecast to admire.
The one lesson. Foresight gains political force when it is visible and opportunity-framed — anchored in institutions citizens and ministers can see, not buried in a secretariat.
Built. Policy Horizons Canada, the federal government's dedicated foresight organization, with a repeatable annual cadence — its flagship Disruptions on the Horizon scan assesses dozens of potential disruptions across society, economy, environment and geopolitics.
Routes in. A standing methodology and an annual product give departments a common, refreshed disruption map to plan against, with training that spreads futures skills across the public service.
The one lesson. Cadence and method matter more than any single brilliant report — a repeatable annual scan that everyone reads beats an occasional masterpiece nobody acts on.
Built. An institutionalized cycle: the Joint Research Centre's foresight capacity, the inter-institutional ESPAS process, and an annual Strategic Foresight Report from the Commission (e.g. the 2023 edition on sustainability and open strategic autonomy).
Routes in. The annual report feeds the Commission's work programme and concrete actions — foresight is embedded in the Union's policy-making cycle, not a side study.
The one lesson. An annual, named report tied to the work programme turns foresight into a recurring obligation of governing — and gives member states a ready ecosystem to nest inside.
Built. GO-Science and its flagship Foresight programme, which run structured evidence reviews and "future worlds" exercises (e.g. the Future of Mobility, looking out to 2040), alongside the national risk-assessment practice.
Routes in. Foresight projects are commissioned to inform specific cross-departmental decisions, with a chief-scientific-adviser network carrying the evidence into each ministry.
The one lesson. Tie every foresight project to a live decision and give it a named official who owns the route in — foresight that is commissioned by a decision rarely dies orphaned.
None of these states bought a better crystal ball. They bought a standing function, a fixed cadence, and a statutory or institutional route from the scenario set into the decision — and that route, not the forecast, is the asset.
| State | The standing institution | How it routes into decisions | The one lesson |
|---|---|---|---|
| Singapore | Centre for Strategic Futures (PMO); Scenario Planning Plus / RAHS | Whole-of-government scenario set; biennial Foresight journal as shared infrastructure | Put it at the centre, keep it permanent, share the scenarios |
| Finland | Sitra + parliamentary Committee for the Future | Statutory Government Report on the Future + a parliamentary answer each term | A statutory hook outlasts the champion |
| UAE | Dubai Future Foundation (sovereign agency) | Opportunity-scanning fused with strategy, accelerators and procurement | Make foresight visible and opportunity-framed |
| Canada | Policy Horizons Canada | Annual Disruptions on the Horizon scan + futures training across the service | Cadence and method beat the occasional masterpiece |
| EU | JRC / ESPAS + Strategic Foresight Report | Annual report feeds the Commission work programme & concrete actions | A named annual report becomes an obligation of governing |
| UK | Government Office for Science — Foresight programme | Decision-commissioned reviews carried in by the CSA network | Tie every project to a live decision and a named owner |
Czechia has real foresight ingredients — they are simply not yet assembled into an engine. The honest reading is neither "we have nothing" nor "we are fine": it is that the country holds capable parts and a rich external ecosystem, but no standing central foresight function to run them once for the whole of government.
What is already there. A capable national statistical office — the Czech Statistical Office (ČSÚ) — already engaged with the UNECE HLG-MOS modernisation agenda, giving the country a credible data backbone to build on. Sectoral pockets of anticipation exist: sector-skills councils on the labour side, the R&D council (RVVI) and Technology Agency on the science side, a digital agency, and a transmission-system operator that already models energy adequacy. And Czechia sits inside a dense EU foresight ecosystem — the JRC, ESPAS, the annual Strategic Foresight Report, Horizon Europe — that supplies scenarios, methods and signals it can nest inside rather than rebuild.
What is missing. There is no standing central foresight engine — no unit at the cabinet centre that owns a shared national scenario set, runs the scanning machinery once for the whole government, and holds line ministries to wind-tunnelling their plans against more than one future. The capable parts therefore operate as thirty private spreadsheets, not one shared spine. As a frontline-adjacent NATO and EU member with limited fiscal slack and a sharp demographic transition already locked in, the asymmetry is unforgiving: Czechia cannot afford to buy its way out of a crisis it failed to imagine.
The report is written and nobody acts. That single sentence accounts for more dead foresight units than any analytical error ever did. Foresight rarely fails because the scan missed the signal or the scenario was wrong; it fails because there was no route from the output to a decision — no owner, no cadence, no mandate, no data to refresh on. The failures are organisational, and so are the fixes.
Six failure modes recur across the institutions that tried foresight and lost it. Each has a predictable cause and a fix that ties straight back to the three things this report argues a foresight capability must be: an engine rather than a report, kept current by an agentic refresh, and owned at the centre of government (Area 1) with a hard route into the decisions being made anyway.
| Failure mode | Why it happens | The fix — and where it ties back |
|---|---|---|
| Orphaned reports written, shelved, forgotten |
The output has no pre-agreed destination — no decision, budget line or law it feeds. A brilliant scenario set lands on a desk with no obligation attached, and the next crisis quietly contradicts it. | Wire every product to a live decision before it is written (the UK GO-Science discipline). The centre-of-government engine (Area 1) owns a statutory hook into budget and legislative cycles, so foresight is a required input, not a courtesy. |
| No data backbone scenarios built on sand |
A scenario is only as credible as the indicators feeding it, and an early-warning system blind across domains cannot see where modern crises form. Without integrated data, the unit improvises from private spreadsheets. | Treat data architecture (Area 31) as strategic infrastructure from day one — the data-mesh and signals registry built beside the engine, so every scenario refreshes from one shared, reusable spine. |
| No mandate / wrong altitude real but powerless |
The unit is buried in a line ministry with no convening power, or it has authority but sits too far from the budget. It cannot make other departments wind-tunnel their plans, so its work is advisory at best. | Place it at the cabinet centre reporting to the cabinet secretary, with embedded ministry liaisons (Area 1). Finland's statutory route shows mandate is what outlasts the champion who created the unit. |
| Captured by short-termism the electoral clock wins |
Ministers are rewarded for what shows up inside a four-year cycle; the long, certain forces — ageing, the energy transition, machine intelligence — are treated as someone else's problem until they arrive as a crisis. | Hard-wire the long horizons into the calendar: a statutory future report each term and a standing weekly anticipatory brief make the 25-year view a recurring obligation, not a discretionary luxury. |
| One-off exercise no continuous refresh |
Foresight arrives as an occasional, expensive set-piece — a workshop, a glossy report — then goes stale because no one can afford to run it again. The world moves; the scenarios do not. | Replace the set-piece with the agentic refresh: a standing fleet of scanning, scenario and red-team agents keeps the scenario set live and fires when a signal crosses a threshold. Cadence, as Canada shows, beats the occasional masterpiece. |
| Foresight as theatre signalling, not deciding |
The unit exists to be seen — a museum, a launch event, a strategy document — with no obligation that anything change as a result. Activity substitutes for influence, and the theatre is defunded the moment budgets tighten. | Bind every exercise to an owned decision and a metric: share of major decisions formally wind-tunnelled, lead time from signal to response. The agentic engine industrialises the work so the unit earns its keep in decisions, not appearances. |
Read the failure table as a column, not six rows, and one pattern resolves: every fix ties back to the same three structural commitments this report has argued from the first page. The engine — foresight as a standing capability, not a report — kills the orphaned-report and one-off-exercise failures. The agentic refresh — a continuous fleet of scanning, scenario and red-team agents — kills staleness and theatre by making the work cheap enough to run forever. And Area 1 ownership at the centre of government — with a statutory hook into budget and legislation — kills the mandate, altitude and short-termism failures by giving foresight a home with the authority to be acted on.
This is why the diagnosis matters more than it first appears. A country that believes its foresight failed because the forecast was wrong will respond by buying better analysis — and watch the next unit die the same death. A country that understands the failure is plumbing will instead invest in the route from output to decision: the owner, the cadence, the mandate, the data backbone. The cheaper, more durable fix is almost always the organisational one.
Foresight does not die at the moment of prediction. It dies in the gap between a finished report and a decision no one obliged anyone to make — and that gap is closed with plumbing, not with a better forecast.
A permanent strategic-foresight unit in the cabinet centre that routes anticipatory intelligence straight into the budget, the legislative programme and the security review — infrastructure, not advice.
Why it ranks first. Every other area runs on this one: absent a home, a method and a route into power, foresight is reinvented ad hoc, politicised, or defunded the moment its champion moves on. The payoff is not better forecasts — governments are mediocre forecasters — but institutional reflexes: the habit of asking "and if the world turns out otherwise?" before money is committed.
Budgeting for the futures that have not happened yet — scenario-based budgets, debt stress-tested against shocks, and a live register of the contingent liabilities priced before they crystallise.
Why it ranks here. Fiscal space is the master constraint: it decides whether a country can invest, absorb a shock, or fund the transitions every other area demands. The binding risks are almost never in the headline deficit — they are the off-balance-sheet pension promises, guarantees and care bills that reactive budgeting discovers only at the moment of crisis, when options are worst and borrowing costs highest.
The demographic spending wave is the single most predictable — and most under-provisioned — fiscal event of the next three decades.
Treating growth as navigation on the map of what a country can plausibly become next — and using anticipation to place scarce industrial and innovation bets on the right rungs of the complexity ladder.
Why it ranks here. Long-run prosperity is a story of structural transformation, not of doing the same things more efficiently. Growth is path-dependent: a country can only easily move into products "nearby" in capability space to what it already makes — so diversification is a foresight problem, defined by where global demand and technology will be in fifteen years, not today. Pick the wrong rung and you waste a decade of subsidy.
A continuous, structured watch on emerging technologies — and on their second- and third-order consequences — to detect disruption while it is still cheap and reversible to act on.
Why it ranks here. Technology is the fastest-moving force on growth, security and quality of life, and it respects no electoral calendar. The cost of acting rises steeply along the diffusion curve: intervene early on standards, skills and regulation and you shape the trajectory for almost nothing; intervene late and you buy influence over an entrenched system at ruinous cost. The expensive surprises are the second-order effects — what AI does to labour, what cheap biotech does to biosecurity, what quantum does to encryption.
Skills are the slowest variable in any economy. Foresight treats the workforce as something a country builds a decade ahead of demand — turning automation panic into managed transition and demographic change into a wage dividend.
Why it ranks here. The real threat is skills mismatch, not job destruction — and mismatch is forecastable. The automation debate fixates on net job counts, where it is wrong in both directions; the violent churn happens underneath stable totals, and the displaced are rarely the people the new roles need. Generative AI is the first wave to expose cognitive and white-collar work at once, compressing transitions that once took a generation into years.
Demography is the most predictable force shaping a nation and the most chronically ignored. The job is forcing slow, compounding certainties into decisions taken on fast political clocks.
Why it ranks here. Demography's danger is its very predictability: because nothing happens suddenly, nothing feels urgent, and the cheap early window closes unnoticed. There is no fast fix — you cannot conjure forty-year-olds, and pro-natal or migration responses take a generation to register. Most of Europe is at the inflection where post-war cohorts retire en masse as small younger cohorts enter work; the choices that soften the next thirty years must be made this decade.
Energy sits at the intersection of growth, security and climate, and the next two decades rebuild it end to end. This is where getting foresight wrong is most expensive — and most visible.
Why it ranks here. Security and decarbonisation are not opposites to be balanced but a single system to be sequenced — and sequencing is a foresight problem. Energy assets have 20-to-40-year lifespans, so a transmission line or nuclear unit commissioned today locks in cost or risk for decades. Electrification of transport and heat is about to surge demand precisely as variable renewables rise and thermal plant retires; the reliability margin thins before it improves, and that vulnerable middle passage must be planned through, not stumbled into.
Almost every state keeps a national risk register and treats it as a document — reviewed annually, shelved, contradicted by the next surprise. This reimagines it as a living, always-on early-warning system.
Why it ranks here. Real crises are interconnected and arrive faster than annual review cycles can track. Anticipation beats reaction more starkly here than anywhere: in a crisis, a day of early warning is worth enormous amounts and decays fast, while the cost of preparedness is modest and the cost of surprise catastrophic. The capability gap is between the speed of modern shocks and the cadence of the institutions meant to see them coming.
Pointing the science budget at tomorrow's strengths — funding the capabilities that compound into national edge fifteen years out, not the fields that produced last decade's citations.
R&D is the slowest-acting and highest-multiplier lever a state holds: a misallocated decade of research wastes a generation of talent and forfeits the spillovers that make small economies rich. For a mid-sized open economy inside Horizon Europe, foresight is the discipline that earns the right to say no to thirty me-too priorities and concentrate.
Mapping chokepoints before they close — the single mines, fabs, ports and protocols on which prosperity silently rests, and how decoupling and friend-shoring will redraw them.
For a small, deeply open, export-led economy braided into German automotive and European value chains, trade is the circulatory system. The lead time to qualify a second supplier or reshore a process runs to years; the lead time of a shock runs to days. Foresight closes that gap before the bill arrives.
Provisioning the national digital backbone ahead of demand — treating compute and data capacity the way a serious state treats grid and water: a strategic stock to be planned, not a service procured when the lights flicker.
It is the substrate beneath innovation, public-service quality, financial supervision and defence. Lead times are long and dependencies sovereign: GPU supply is rationed globally, data-centre power collides with grid and climate limits, and cloud dependence becomes geopolitical exposure. Digital infrastructure is a capacity-planning problem on the scale of energy policy, not an IT budget line.
Seeing the next crisis form — watching the system the way an epidemiologist watches a population: not whether each institution is sound, but where fragility is quietly accumulating and how a small failure cascades.
Financial stability is the precondition for everything else: a banking or sovereign-debt crisis vaporises years of growth faster than any other shock. The damage is non-linear and the warning window is the thing always missing. Stress tests ask "can the system survive the last crisis?"; foresight asks "what is the next one, and is anyone mandated to look?"
Financial crises are never new in substance and always new in costume.
Building 50-year assets for a future we can only imagine — reframing the capital pipeline as a portfolio of options under deep uncertainty: not "what will demand be in 2055?" but "which assets perform across the futures we cannot rule out?"
Infrastructure errors are uniquely unforgiving: a stranded grid corridor or an under-sized lock chamber is a fifty-year liability cast in concrete, with sunk cost defending it against correction. Decarbonisation is rewiring demand, climate is rewriting the load envelope, and reshoring is redrawing the European freight map — capital is being committed now against the most uncertain fifty years in living memory.
Securing the inputs the future economy runs on — lithium, rare earths, gallium, high-purity silicon, copper — whose supply is concentrated, politicised and slow to scale. The supply-side insurance policy for every industrial ambition.
Materials are a chokepoint multiplier: a single export restriction can stall an entire downstream industry that took a decade to build. The 2023 gallium and germanium controls showed how a refining monopoly converts into geopolitical power overnight. Opening a mine takes 10–15 years; a supply shock lands in weeks — foresight is the only instrument operating on the timescale of the cure.
Anticipating the future of the firm base — where the productive capacity and jobs of 2035 actually form, which kinds of firm, in which sectors, at what scale — and shaping the environment to meet that emergence, not subsidise its predecessor.
The firm base is where every other lever cashes out into productivity and employment — and OECD business dynamism has fallen for two decades: fewer high-growth entrants, more zombie incumbents. Czechia's maturing manufacturing base, thin scale-up layer and retiring founder-owners make this a live strategic question. AI now threatens to reshape firm size faster than any institution can react.
Getting ahead of where global capital moves next — positioning to attract the strategic sectors of the future while screening the inbound flows that threaten sovereignty. National positioning in a contest already underway.
Investment-attraction is a winner-take-much contest with durable lock-in: a gigafactory or AI datacentre cluster anchors suppliers, skills and follow-on capital for decades, and there are only so many to win. Reacting to an investment wave is losing it — the site-selection, incentives and skills pipeline must be ready before the capital looks. The EU FDI-screening framework exists because some inbound capital is a vector for strategic dependence.
Designing a health system for the demand of 2040, not 2010 — treating demography, epidemiology, biosecurity and workforce as one coupled system, and asking not "is it efficient now?" but "is it solvent against 2040?"
Health is a growth engine, a stability anchor and the most visceral component of quality of life — and its failure modes are non-linear. A pandemic erases years of GDP in months; an under-projected workforce gap closes hospital wards a decade after the missed recruitment decision. WHO, ECDC and HERA exist because reactive health policy is ruinously expensive.
Teaching for a world that doesn't exist yet — a child entering school this year will work into the 2070s. Teach forward: anticipate the skills, the shape of work, and the rhythm of learning across a whole life.
Skills are the binding constraint on everything else in this report — you cannot run an energy transition, a digital economy or a modern health system without the people to staff them, and people take fifteen to twenty years to grow. Get it wrong and you import the skills, watch your young emigrate, or stall the transitions you committed to. The recurring question: are we teaching content AI will commoditise, or capabilities that compound?
Turning a demographic burden into an economic frontier — making care, pension and fiscal systems solvent against the curve while capturing the longevity-economy market that older, healthier, wealthier populations open up.
Ageing is the master demographic driver beneath health, care, pensions, labour supply and public finance — and unusually certain: the people old in 2040 are already alive and counted. Left reactive, it is a slow fiscal landslide; handled with foresight, the same demographics become a stability story and a growth story. For Czechia, with one of the EU's steeper ageing curves, this is the central fiscal question of the next two decades.
Anticipating flows and making integration work — reading the drivers that move people before flows crest, then ensuring arrivals find their way into the labour market and social fabric. Arrival and integration as one system to be designed.
Migration is a sharp instrument that cuts toward growth or instability depending entirely on management. For ageing, shrinking European societies, managed inward migration is among the few levers that ease workforce and fiscal pressure. The long-run outcome is decided not at the border but in the decade after — and integration outcomes shape social trust for a generation.
Spotting the fracture lines before they break — treating the fabric of society as a forecastable system, watching where trust thins and modelling how it propagates while intervention is still cheap and legitimate.
Cohesion is the precondition for everything else: growth strategies, green transitions and demographic adjustments all demand that citizens accept disruption today for benefit tomorrow — and acceptance runs entirely on trust. Where trust erodes, even good policy is read as elite capture and reforms stall. A country that anticipates where cohesion thins gains the social licence to actually govern.
Keeping the whole country in the future — refusing the comfort of the national average, surfacing regional divergence before it hardens into permanent decline, and investing in opportunity geographies while the window is open.
The geography of opportunity has become the defining political and economic cleavage of the age — the "places that don't matter" striking back through the ballot box. For Czechia, the Prague-versus-Ústí/Karlovy Vary/Moravian-Silesian cleavage is acute, and EU Cohesion Policy moves enormous sums blind to long-horizon trajectories. Divergence destroys productivity, corrodes cohesion and degrades quality of life at once.
Planning the city of 2050 while it is still affordable to plan — cities are slow machines built from fast decisions; a zoning rule or height limit set today locks in form, cost and liveability for fifty years.
Housing is where the macroeconomy meets the kitchen table: unaffordable housing suppresses fertility, traps labour, transfers wealth violently across generations, and drives the cohesion and regional-divergence crises directly. Prague's affordability is among Europe's worst relative to incomes, and supply responds on a five-to-ten-year lag — by the time a problem is obvious it is a generation too late to fix cheaply.
Steering by quality of life, not just output — anticipating the trajectory of healthy-life-expectancy, mental health and lived quality of life, and building the dashboards that let governments optimise for what citizens experience.
Wellbeing is both end and means: healthy, secure, mentally well populations are more productive, cohesive and resilient — quality of life is a precondition of sustainable growth, not a luxury bought after it. The deterioration is underway: rising youth mental-ill-health, stalling healthy-life-expectancy, an epidemic of loneliness, and a widening gap between rising output and flat life-satisfaction.
Pricing in a future climate that is already locked in — sizing roads, grids, hospitals and fiscal buffers for the climate of the 2050s, not the 1990s in which they were planned.
Adaptation is the resilience floor under every growth ambition: a flooded substation or a low-flow river starving a plant of cooling water erases years of output in a week. For a landlocked, river-dependent industrial economy — Elbe and Morava flooding, Moravian drought, grid heat stress — adaptation is macro-fiscal policy. Physical risk transmits into transition risk, insurance withdrawal and sovereign borrowing cost.
Securing the non-substitutable basics under stress — turning "we are mostly self-sufficient" complacency into a scenario-tested map of where the real fragilities sit.
Food-price spikes are among history's most reliable precursors of unrest, and even import-secure economies are exposed through global commodity, fertiliser and energy markets — the 2022 shock chained them together fast. Drought-prone, CAP-dependent Central Europe, with falling groundwater and stressed soils, faces this as a direct driver of rural stability and food-price inflation that monetary policy cannot easily tame.
Managing the productive assets that don't appear on the balance sheet — forests, soils, pollinators and wetlands that are silently liquidated until the services they provide fail.
A very large share of economic activity depends moderately or highly on nature, and the danger is non-linearity: an ecosystem degrades gradually, then collapses, wiping out services assumed free and permanent. The bark-beetle devastation of spruce monocultures was a vivid, expensive lesson — a managed ecosystem crossing a threshold and imposing sudden costs on the timber economy and landscape, often irreversible on policy timescales.
Reframe environmental policy from a compliance chore into asset management of the stock the economy quietly depends on — a stock that, once past a threshold, cannot be rebuilt with money.
Reading the strategic environment before it turns — matching slow-to-build force design, posture and warning to the war that might come, not the last one fought.
Security is the precondition for everything else, and being wrong here is least recoverable: forces take a decade to build into a threat environment that may already have transformed. The post-2022 environment — Russian aggression, rapid rearmament, the drone-and-EW revolution, a contested NATO eastern flank — gives a frontline-adjacent member a direct stake in getting force design and warning right. Long lead times make foresight structurally necessary, not optional.
Defending the information space and the democratic core — staying ahead of a threat surface that mutates monthly and runs at machine speed.
This is a precondition, not a sector: a compromised election, a poisoned public conversation or a paralysed hospital network corrodes the trust every other policy depends on. Because attacker innovation outpaces defender procurement cycles, only an anticipatory posture closes the gap — reactive cyber-defence guarantees you are always one campaign behind. The question is whether democratic resilience compounds faster than adversary capability.
Regulating emerging technology ahead of the harm — flipping the default from legislating after the crash to shaping rules, sandboxes and institutions before harm is locked in.
Regulation is the interface between every emerging technology and the real economy; a slow or brittle interface throttles both safety and growth. Reactive rule-making delivers the worst of both worlds — harm runs ahead unchecked, then a panicked overcorrection chills investment. Inside the EU's regulatory gravity (AI Act, GDPR, DSA/DMA), anticipatory governance is how a smaller state punches above its weight: the jurisdiction where adaptive rules make responsible innovation faster, not slower.
The shared backbone that powers every other area — a national data architecture and decision-intelligence layer that turns raw statistics, sensors and registers into integrated, near-real-time, decision-ready signals.
Nothing else works without it. A scenario is only as credible as the indicators feeding it; an early-warning system that cannot see across health, energy, mobility and finance in one view is blind precisely where modern crises form. The World Bank's Data for Better Lives made the core case: data is a public good whose value compounds when integrated and reused.
For the Czech Republic, with a capable national statistical office (the Czech Statistical Office, ČSÚ) and the UNECE HLG-MOS modernisation agenda already in play, the backbone is within reach — but only if treated as strategic infrastructure with sustained investment, not a perennial IT afterthought.
A data-mesh approach lets each ministry own its domain data while exposing it as a governed, interoperable product — federation without a single brittle warehouse. Knowledge graphs provide the semantic layer linking people, places, events and indicators across silos. Nowcasting converts high-frequency sources — satellite, mobility, payments, sensors, web-scraped prices — into live reads of the economy, health and environment. The standard scanning, scenario and cross-impact methods all draw from this common, trusted substrate.
The statistical office evolves from periodic publisher into the operator of an always-on, trustworthy decision-intelligence platform — the engine room of the foresight state.
The human operating system for a foresightful state — the skills, habits and culture that turn the other thirty-one areas from reports into reflexes.
Foresight fails most often not because the analysis was wrong but because no one in the room knew how to use it. A state can buy the best scanning platform and commission brilliant scenarios, yet if officials default to short-term, reactive, single-future thinking, the investment is wasted. UNESCO made futures literacy a recognised competence precisely because the bottleneck is human, not technical.
The differentiator is whether anticipation is a specialist unit's job or a shared way of working. The countries that win are not those with the most foresight reports, but those where foresight has become an ordinary, embedded capability of the people who govern.
UNESCO-style Futures Literacy Laboratories let people rigorously imagine and use multiple futures; reframing exercises surface and challenge the implicit assumptions behind today's decisions. Participatory scenario workshops and the Futures Wheel build the muscle by doing, not lecturing. Causal Layered Analysis trains officials to see beneath events to the deeper narratives — paired with communities of practice and "learning by intervening" on real policy problems, not abstract rehearsal.
Coaching agents prompt the anticipatory question inside real decision workflows — "what would have to be true in ten years?" — until anticipation becomes the path of least resistance.
Four governments did not get better forecasts than everyone else — forecasts are usually wrong. They built the capability: a standing engine that scans, models, stress-tests, and routes what it learns into the decisions being made anyway. Each proves a different transferable lesson about how anticipation is wired into power.
The Prime Minister's Office runs the Centre for Strategic Futures (CSF), paired with the Strategic Futures Network of permanent secretaries. Its method lineage — RAHS (Risk Assessment & Horizon Scanning) and Scenario Planning Plus — turns scanning into the national scenario set every ministry plans against.
| Routes via | The PMO seat plus the permanent-secretary network make foresight a required input to the budget and strategy cycle — not a courtesy memo. |
| Lesson | Put the engine at the centre, with a standing convening mandate, so it cannot be defunded when a champion moves on. |
2009 CSF stood up after the RAHS programme · scenario practice traces to the 1990s Shell-influenced Scenario Planning Office · Foresight journal now in its 15th-anniversary cycle.
Finland is unique in giving foresight a parliamentary home: the Committee for the Future, a standing committee of MPs that debates the government's quadrennial Report on the Future. The innovation fund Sitra supplies the megatrends and futures capability that feed it.
| Routes via | The government must publish a long-horizon Report on the Future each term; Parliament must formally respond — anticipation becomes a constitutional habit, not an option. |
| Lesson | Anchor foresight on both sides of power. A legislative hook outlasts any single administration. |
1993 Committee for the Future established · 1988 Sitra's futures work matures · Government Report on the Future now a fixed once-per-term obligation.
Policy Horizons Canada is the federal government's own foresight organisation, producing the flagship annual Disruptions on the Horizon — a structured, ranked scan of plausible disruptions across society, economy, environment, politics and health, built to be reused across departments.
| Routes via | A repeatable annual product plus open methods and training spread futures literacy across the public service — foresight as a shared craft, not a black box. |
| Lesson | A disciplined, published cadence and method beats brilliant one-offs. Repeatability is what makes foresight credible. |
1996 origins as the Policy Research Initiative · matured into Policy Horizons Canada · the 2024 edition assesses 35 potential disruptions across five domains.
The Dubai Future Foundation treats the future as something to be prototyped, not just studied — convening ministries, running the Museum of the Future, and pushing foresight straight into pilots, regulation-light testbeds and sovereign moonshots.
| Routes via | High-level political sponsorship plus a bias to build: scenarios convert into pilots, testbeds and public-facing commitments fast. |
| Lesson | Pair foresight with delivery. A futures function that can prototype turns anticipation into momentum, not paper. |
2016 Dubai Future Foundation established under direct leadership sponsorship · Museum of the Future opened 2022 as its public anticipatory front door.
Read side by side, the cases are not four models to choose between — they are four answers to the same question: how does anticipation get a guaranteed route into the decision? Singapore puts it at the executive centre; Finland anchors it in the legislature; Canada makes it a disciplined public craft; the UAE fuses it with delivery. Each solves the failure mode the others are exposed to.
| State | What they built | The route into decisions | Transferable lesson |
|---|---|---|---|
| Singapore | CSF in the PMO; RAHS + Scenario Planning Plus; Strategic Futures Network of perm-secs. | Cabinet-centre seat; shared national scenario set feeds budget & strategy. | Build the engine at the centre — one scan for the whole of government. |
| Finland | Parliamentary Committee for the Future; Government Report on the Future; Sitra megatrends. | Statutory once-per-term future report; Parliament must respond. | A legislative hook makes foresight survive elections. |
| Canada | Policy Horizons Canada; Disruptions on the Horizon (35 disruptions, 2024). | Repeatable annual product; open methods + training across departments. | Cadence and method beat one-off brilliance. |
| UAE | Dubai Future Foundation; Museum of the Future; testbeds and moonshots. | Leadership sponsorship; scenarios convert into pilots and regulation-light testbeds. | Pair foresight with delivery — prototype the future. |
The countries that built foresight did not get better forecasts than everyone else. What they got was the ability to act while action was still cheap.
The order is the strategy. A country starting from zero stands up the minimum viable foresight state first — the centre-of-government engine, the national early-warning system, the data backbone, and futures literacy — then extends along the macro spine. Build the engine once; every later domain is a low-cost extension of a system that already works.
| First move | Owner | Horizon | Metric to watch |
|---|---|---|---|
| 1 · Stand up the foresight engine — a lean directorate at the cabinet centre; publish the first national scenario set; wind-tunnel the next budget. | Chief Foresight Officer → cabinet secretary Area 1 | Q1 | Share of major decisions wind-tunnelled; signal-to-response lead time. |
| 8 · Stand up national early-warning — always-on cross-hazard scanning; rebuild the risk register as a living, cascade-based system. | Resilience / national-security secretariat (PMO) Area 8 | Q1–Q2 | Warning lead time on materialised risks; preparedness coverage of priority risks. |
| 31 · Lay the data backbone — data-mesh + knowledge graph for three priority domains; one nowcasting pilot; codify the privacy framework. | National statistical office + digital-gov authority Area 31 | Q1–Q3 | Decision latency (event-to-indicator); interoperable data products published. |
| 32 · Build futures literacy — a curriculum + a cohort of trained facilitators seeded across ministries; a cross-government community of practice. | Civil-service academy + centre-of-gov foresight Area 32 | Q2–Q3 | Share of officials with working futures literacy; foresight-to-decision ratio. |
| 2 · Extend to the fiscal spine — replace single-baseline budgeting with a three-scenario stress test; build the contingent-liability register. | Finance ministry + independent fiscal council Area 2 | Q3–Q4 | Contingent liabilities priced; medium-term forecast error narrowed. |
| 3 · Extend to growth & industry — build the economic-complexity dashboard; shortlist reachable high-value rungs; deliver a diversification map. | Economy ministry strategy fn + development bank Area 3 | Q4 | Complexity index trajectory; relatedness density around target activities. |
| 5 · Extend to labour & skills — stand up the live vacancy-and-exposure dashboard; publish a first three-horizon skills outlook. | Labour + education ministries; PES Area 5 | Q4 | Skills-mismatch index; share of training places aligned to forecast demand. |
Four moves make a foresight state that functions: an engine at the centre, eyes on the horizon, a spine of data, and people who can think in futures. Stand these up in the first year and the macro spine plugs straight in. The phasing below turns the roadmap into a calendar.
Engine + scenario set + signals registry + cross-hazard watch. The state stops being surprised by the predictable.
Liaisons in ministries; first wind-tunnelled budget; futures literacy seeded. Foresight reaches the decision.
Data backbone live; macro spine plugged in. Each new domain is now a cheap extension.
A foresight capability that cannot survive its own critics does not deserve a budget. So here are the five objections raised most often — each stated at its strongest, the way a sceptical finance minister would put it — and then answered. The honest answer to most of them is not "you're wrong" but "you've described foresight done badly, which is exactly what this design prevents."
| The objection (steelmanned) | The answer |
|---|---|
| "Foresight is just guessing." The future is unknowable; dressing up guesses as scenarios lends false authority to what is, at bottom, speculation. | Correct — and beside the point. Foresight does not claim to predict; it claims to prepare. Its product is not a forecast but option value: the rehearsed response, the early move, the assumption tested before money is committed. You do not need to know which crisis comes to be less surprised by all of them. |
| "Agents hallucinate, and automation bias makes it worse." An AI fleet that confidently invents signals, scaled across government, is more dangerous than no foresight at all. | Which is why the design is explicit: agents widen the aperture; humans keep the accountability. Agents scan, generate and red-team; a human directorate curates, judges and owns what goes up. Every claim is traceable to source, and red-team agents are pointed at the system itself. Discipline is a design choice, and we made it. |
| "It'll be captured by short-termism — or become theatre." Foresight units get colonised by the politics of the moment, or survive as ritual workshops that change nothing. | The single most common failure mode — and the reason for the statutory hook (Finland) and the centre-of-government seat (Singapore). The KPI that matters is the ratio of foresight that changes a decision to foresight that merely gets published. Measure that, and theatre has nowhere to hide. |
| "We can't afford it." Fiscal slack is tight; a new unit is a cost we can defer until conditions improve. | The asymmetry runs the other way. A founding directorate is eight to twelve people; the cost of one un-anticipated crisis — a skills gap that becomes mass unemployment, a risk that becomes an emergency, an asset that strands — dwarfs it. For a mid-sized economy with little fiscal slack, the asymmetry is sharper, not weaker. |
| "Scenarios never come true." We built scenarios last cycle and none described what actually happened — so why build more? | A scenario that "comes true" was a forecast, and a lucky one. Scenarios are not bets on one future; they are stress tests for many. The pandemic was on every risk register and the energy shock in every scenario set — the failure was never imagination, it was routing. This design fixes the routing. |
Behind the five objections sits one real danger: not that foresight is wrong, but that it is built badly. Foresight units die from a small, repeating set of causes. The guardrails below are the response — each a structural answer, not a slogan.
A foresight engine would be a luxury if the future were noise. It is not. The technologies that reshape economies ride exponential cost-and-capability curves — and those curves are read-ahead-able. Across 62 technologies, Wright's law (cost falls as a power of cumulative production) forecast progress better than calendar time; the same shape governs solar, batteries and machine-learning compute. The disruption was a surprise only to institutions that were not reading the curve. The agentic architecture exists to read every such curve, continuously, and route what it reads into the budget while the move is still cheap and reversible.
Reading a curve early buys the cheap, reversible move; reacting late buys the expensive, irreversible one. Oxford INET's empirically grounded forecasts found a fast technology-led energy transition is multi-trillion-dollar net cheaper than a slow one — the same arithmetic every area in this report describes, now made continuous.
Way, Ives, Mealy & Doyne Farmer (Oxford INET); RethinkX cost-curve disruption framework.
| # | Component | Layer | What its agents do | Anchor source |
|---|---|---|---|---|
| A1 | Agentic Foresight Operating System | I | Orchestrate the fleet; own the shared futures picture | OECD–WEF AI in Strategic Foresight |
| A2 | National Signals Registry & ingestion | I | Curated, deduplicated weak-signal feed | OECD Strategic Intelligence Tools |
| A3 | Scanning agents | I | Continuous horizon scanning & weak-signal detection | SciMON; WISDOM |
| A4 | Forecasting agents | I | Calibrated probabilities on dated questions | Halawi et al; Schoenegger |
| A5 | Scenario-generation agents | I | Recombine signals into stress-tested branches | Ritchey morphological analysis |
| A6 | Simulation & ABM agents | I | Synthetic populations of households & firms | Park et al, Generative Agents |
| A7 | Red-team & adversarial agents | I | Attack every plan and every forecast | CSET; Intl AI Safety Report 2026 |
| A8 | Discovery agents | I | Automated tech/science discovery as early signal | Lu et al, The AI Scientist |
| A9 | Briefing agents | I | The two-page weekly anticipatory brief | OECD–WEF; Singapore CSF |
| A10 | Cost-and-capability curve library | II | The forecastable-disruption substrate | Nagy–Farmer; Epoch AI; IRENA |
| A11 | Patent & bibliometric tech-mining | II | Read the frontier before it reaches market | WIPO; OECD Patent Manual |
| A12 | Technology-readiness & roadmapping base | II | Turn a signal into a timed trajectory | NASA TRL; Cambridge IfM |
| A13 | Provenance & the evidence ledger | II | Traceability of every claim the fleet emits | OECD Anticipatory Governance |
| A14 | Human curation & judgement layer | II | Analysts as curators and judges; accountability | Schoenegger, AI-Augmented |
| A15 | AI & frontier-compute fleet | III | Read the curve that bends every other curve | Epoch AI; Stanford HAI Index |
| A16 | Quantum & advanced-computing fleet | III | Price a slow curve with a sharp crypto cliff | WEF; CSET; Lau et al |
| A17 | Engineering-biology & bioeconomy fleet | III | Falling cost curve meets rising risk curve | OECD Bioeconomy; Nat Academies |
| A18 | Energy & clean-tech transition fleet | III | The domain where cost-curve foresight is proven | IEA; IRENA; Oxford INET |
| A19 | Technology-convergence (NBIC) fleet | III | Catch disruption where domains compound | NSF NBIC; OECD STI Outlook 2025 |
| A20 | Critical & dual-use / security fleet | III | The geopolitics of the frontier | US CET List; NATO STO |
| A21 | Anticipatory-governance wrapper | IV | Route into rules without freezing innovation | OECD; WEF Agile Governance |
| A22 | Assurance of the agents themselves | IV | Reliability, calibration & red-teaming of the fleet | Intl AI Safety Report 2026 |
| A23 | Technology-assessment & parliament | IV | The democratic-legitimacy layer | EP STOA; US OTA legacy |
| A24 | Wiring into budget, STI & procurement | IV | A required input, not a courtesy | OECD STIP; EU RIS3 |
| A25 | KPIs & the option-value ledger | IV | Measure lead time & cost avoided, not reports | Singapore CSF; Sitra |
| A26 | The build sequence | IV | First 24 months, talent, the mid-sized state | UNIDO; UK GO-Science |
The twenty-six components are built from a small, reusable kit of agent archetypes. The evidence that they work is now real, not promissory: language-model forecasters approach the human-crowd aggregate on real questions, an ensemble of models rivals the crowd, and an LLM assistant lifts a human forecaster's accuracy by roughly a quarter. The discipline is equally real — every archetype runs under a named human owner, a calibration score, and an adversarial check.
| Archetype | What it does | Evidence it works |
|---|---|---|
| Scanning | Reads the whole information space; flags weak signals & trend-breaks | SciMON; WISDOM weak-signal detection |
| Forecasting | Puts calibrated probabilities on dated questions | Halawi; Schoenegger Silicon Crowd |
| Scenario | Recombines signals into stress-tested narrative branches | Morphological analysis; OECD–WEF |
| Simulation / ABM | Runs synthetic populations of households, firms, regions | Park et al, Generative Agents (25-agent society) |
| Red-team | Attacks every plan; hunts the assumption that breaks it | CSET national-power work; Intl AI Safety Report |
| Discovery | Generates & tests new tech/science as an early signal | Lu et al, The AI Scientist (end-to-end) |
| Briefing | Compresses the fleet's output into the weekly two-pager | OECD–WEF; Singapore CSF practice |
No agent output reaches a budget desk or a statute without a named human standing behind it. The curation-and-judgement layer (A14) is not a concession to caution — it is where the architecture's legitimacy lives. Machines flag and model; accountable officials decide and act.
The engine room (A1–A9): the orchestration layer, the shared signal feed, and the seven agent families that scan, forecast, generate scenarios, simulate, red-team, discover and brief. Build this and a state has a working nervous system, however thin — a continuous loop where a quarterly workshop used to sit.
The fabric's first job is to run foresight once for the whole of government rather than thirty times badly. A single operating system owns the shared futures picture; a single signals registry is the evidentiary spine every ministry scans against. On that spine sit the agent families — each cheap to run at the margin, each tireless, none trusted without a human owner. The reframe is that capacity, not insight, was always the binding constraint: a handful of analysts could watch only so much. The fabric lifts the ceiling on attention while leaving the ceiling on judgement exactly where it belongs.
| # | Component | The agentic move & its guardrail |
|---|---|---|
| A1 | Foresight Operating System | Orchestrates the fleet and owns the one shared futures picture — run once for the whole of government, not thirty times badly. Guardrail: a Chief Foresight Officer accountable for what goes up. |
| A2 | National Signals Registry | A curated, deduplicated weak-signal feed the whole fleet scans against — the shared evidentiary spine, not thirty private spreadsheets. |
| A3 | Scanning agents | Read the full reachable information space daily; flag trend-breaks the moment they cross thresholds. Guardrail: human triage of signal vs noise. |
| A4 | Forecasting agents | Put calibrated probabilities on dated questions; track forecast-vs-outcome. Guardrail: published calibration scores, never a bare number. |
| A5 | Scenario-generation agents | Spin and stress-test narrative branches; pressure-test the canonical national scenarios for staleness. |
| A6 | Simulation & ABM agents | Run policy options across synthetic households, firms and regions to capture behaviour standard models miss. |
| A7 | Red-team agents | Adversarially attack every major plan and forecast; forbidden from authoring the plans they attack. |
| A8 | Discovery agents | Generate and test new technological possibilities — what is becoming feasible before it shows up in patents. |
| A9 | Briefing agents | Compress everything into the PM's two-page weekly anticipatory brief; the human directorate curates what goes up. |
Every later layer plugs into the fabric. Stand up the operating system, the registry and two or three proof-carrying agents, and the data, domain and governance layers become low-cost extensions of a system that already works. Attempt them first and you get isolated tools nobody trusts.
The living substrate the agents reason over (A10–A14). At its centre is the cost-and-capability curve library — the thing that makes disruption forecastable — flanked by the patent and bibliometric mining layer, the technology-readiness base, the evidence ledger that makes every claim traceable, and the human curation layer where accountability lives.
Watching more signals only changes outcomes if the signals resolve into timing. The curve library is what converts a scanning agent's "this is happening" into a forecasting agent's "and it gets cheaper than the incumbent in roughly N years" — the one sentence a budget can act on. Around it, the tech-mining and readiness layers locate where a technology sits on its trajectory, the evidence ledger version-stamps every claim, and the curation layer is where a named human turns the machine's output into an institutional position. This layer is the difference between a fleet that talks and a fleet a finance ministry will believe.
| # | Component | What it holds & why it matters |
|---|---|---|
| A10 | Cost-and-capability curve library | Wright's-law experience curves for every tracked technology — the substrate that turns "what's new" into "when it gets cheap." Grounded in Nagy–Farmer, Epoch AI and IRENA cost data. |
| A11 | Patent & bibliometric tech-mining | Reads the frontier before it reaches the market — patent families, preprints, weak-signal topic detection. WIPO, OECD Patent Manual, CWTS Leiden. |
| A12 | Technology-readiness & roadmapping base | Turns a signal into a timed trajectory via TRL and roadmapping. NASA/DoD TRL, Cambridge IfM, UNIDO. |
| A13 | Provenance & the evidence ledger | Every claim the fleet emits is traceable to its source and version — the precondition for routing agent output into power. |
| A14 | Human curation & judgement layer | Analysts as curators and judges. The RCT evidence is that human + LLM beats either alone — so this is a performance layer, not just a control. |
Agents are commodities; the curated, provenance-stamped, continuously-updated knowledge base they reason over is not. A state that owns its curve library and signals registry owns a compounding asset — every month of operation makes the next forecast sharper.
Frontier coverage (A15–A20): specialised fleets that each read one domain's curves, plus a convergence fleet that catches the disruptions no single-domain scan would see. The library behind them is real — 102 documents spanning AI & compute, quantum, biology, energy, convergence and dual-use security.
Each fleet is the Core Fabric pointed at one frontier, fed by that frontier's slice of the curve library. The AI & compute fleet reads the curve that bends every other curve; the energy fleet works the domain where cost-curve foresight is already proven; the biology fleet tracks a falling cost curve against a rising risk curve; the convergence fleet exists precisely because the highest-leverage disruptions fall between domains. For a mid-sized industrial state, two of these — AI/compute and energy — carry the most option value and are where a Czech-sized build should start.
| # | Fleet | The frontier signal it reads | Anchor source |
|---|---|---|---|
| A15 | AI & frontier-compute | Training-compute & capability curves — the curve that bends every other curve | Epoch AI; Stanford HAI |
| A16 | Quantum & advanced computing | A slow capability curve with a sharp cryptographic cliff (harvest-now-decrypt-later) | WEF; CSET; Lau et al |
| A17 | Engineering biology & bioeconomy | A falling cost curve meeting a rising biosecurity risk curve | OECD Bioeconomy; Nat Academies |
| A18 | Energy & clean-tech transition | The most proven cost-curve domain — solar, wind, batteries, electrolysers | IEA; IRENA; Oxford INET |
| A19 | Technology convergence (NBIC) | Where AI×bio×materials×quantum compound into disruption | NSF NBIC; OECD STI 2025 |
| A20 | Critical & dual-use / security | The geopolitics of the frontier — critical-tech lists & economic security | US CET List; NATO STO |
The disruptions that broke incumbents in the last decade — AI-designed proteins, software-defined everything, electrified transport — lived between disciplines. A scan organised purely by domain misses them by construction; the convergence fleet (A19) is the deliberate correction.
The trust and the route to power (A21–A26). The first three layers build a machine that sees; this layer decides whether anyone may act on what it sees — routing foresight into rules without freezing innovation, assuring the agents themselves, keeping the whole apparatus democratically legible, and hard-wiring it into the decisions that spend money.
A foresight fleet that scans, forecasts and red-teams produces a torrent of anticipatory judgement — and anticipatory judgement is worthless, or worse, unless it is governed on three fronts at once. It must route into rules without freezing the innovation it is reading (A21). The agents producing it must themselves be reliable, calibrated and adversarially tested, because routing an over-confident agent into power is a faster way to make a bad decision (A22). And the whole apparatus must be answerable to a parliament, not just a minister and a vendor (A23). Only then does the wiring (A24) — making the engine a required input to the budget and procurement — become safe to switch on.
| # | Component | The job it does |
|---|---|---|
| A21 | Anticipatory-governance wrapper | Routes foresight into adaptive regulation and sandboxes — acting on the Collingridge dilemma without freezing the future. OECD; WEF Agile Governance. |
| A22 | Assurance of the agents themselves | Reliability, calibration and red-teaming of the fleet — you cannot route an over-confident agent into power. Intl AI Safety Report 2026. |
| A23 | Technology-assessment & parliament | The democratic-legitimacy layer — answerable to a parliament, not just a minister and a vendor. EP STOA; the US OTA legacy. |
| A24 | Wiring into budget, STI & procurement | Makes the engine a required input to the decisions that allocate money — not a courtesy report. OECD STIP; EU RIS3. |
| A25 | KPIs & the option-value ledger | Measures the engine by signal-to-response lead time and cost avoided by acting early — not by reports produced. |
| A26 | The build sequence | The dated 24-month plan, the talent, and the discipline of self-funding each increment from the last. |
Anticipatory judgement routed into power ungoverned is a faster way to make a bad decision, not a slower one. Assurance and legitimacy are the difference between a clever internal tool and an instrument of the state — which is why the wiring into the budget is switched on last, not first.
The architecture is a single continuous loop. Raw signals are ingested and deduplicated into the registry; the curve library and tech-mining layer turn them into timed trajectories; the agent fleet forecasts, simulates and red-teams; and the briefing layer routes a curated two-pager into the decision — every step provenance-stamped and human-owned. The loop runs daily, where a foresight set-piece used to run once a year.
Signals Registry (A2) — preprints, patents, procurement, financing, standards.
Curve library (A10) + patent mining (A11) → timed trajectories.
Scan · forecast · scenario · simulate · red-team (A3–A8).
Briefing agent (A9) → the budget, STI strategy & procurement (A24).
The evidence ledger (A13) version-stamps every claim, and the assurance layer (A22) scores every agent's calibration. Nothing reaches the brief unprovenanced — the precondition for letting agent output touch a budget.
The curation layer (A14) owns adjudication — which signals matter, which forecasts to trust, which recommendation to sign. The fleet widens the aperture; the human keeps the call, and carries the accountability for it.
| Stage | Agents involved | What it emits | The human control |
|---|---|---|---|
| Ingest | Registry & scanning (A2–A3) | Deduplicated weak signals | Triage: signal vs noise |
| Ground | Curve library, tech-mining (A10–A12) | Timed cost/capability trajectories | Sign-off on the curve fit |
| Reason | Forecast, scenario, sim, red-team (A4–A8) | Calibrated forecasts & stressed scenarios | Adjudicate disagreement |
| Route | Briefing & wiring (A9, A24) | The two-page anticipatory brief | Owns & signs the recommendation |
Movement F is built on a dedicated, real, downloaded source library — 102 primary documents across fifteen angles, ~527 MB on disk — pulled from the world's leading foresight and technology institutions. Every claim above is traceable to a named document; none is invented.
A mid-sized EU member state — the running example is the Czech Republic — does not buy foresight with a thousand analysts. It builds the nervous system in the right order, on a modest budget, and proves value against a real decision within the year. The figures below are ENSI design targets, not measured facts: they exist to make the architecture buildable.
| Window | What ships | The proof it must produce |
|---|---|---|
| Months 1–6 | Foresight OS (A1) + Signals Registry (A2); first scanning & forecasting agents (A3–A4); seed the curve library (A10) | The fleet can sense and forecast at all, with humans curating every output |
| Months 7–12 | Briefing agent (A9) + budget-annex wiring (A24); first option-value ledger entries (A25) | One anticipatory brief lands inside the budget round |
| Months 13–18 | One or two priority domain fleets (for a Czech-sized industrial state: AI/compute A15 + energy A18); assurance & calibration (A22) | A live, dated readiness assessment in two frontier domains |
| Months 19–24 | Red-team & evaluation agents (A7); hardened provenance & evidence ledger (A13); first full option-value portfolio | Year two funded by demonstrated value, not faith |
Build the engine and the wiring before the domain fleets; never scale agents faster than the curation team can govern them; and let each increment fund the next from the value the ledger books. For a state with thin fiscal slack, that is not the expensive option — it is the only affordable one.
This is an analysis-first report built on a real, downloaded source library. The method was deliberate: assemble the evidence base first, derive the 32-area analysis from it, then front-load the report with the argument so a reader who never reaches the catalog still owns the thesis.
Every claim in this report stands on a real, downloaded library — 188 primary documents across 20 angles, each a PDF pulled from the original institution and recorded with its source URL. This is the source mix: where the evidence comes from, and how it is distributed across the questions the report answers.
The library is weighted toward the institutions that actually run foresight — OECD, the EU JRC and Commission, the UK Government Office for Science, Policy Horizons Canada, Singapore's CSF, Sitra, the Dubai Future Foundation, the US National Intelligence Council — balanced by the UN and World Bank data literature and the seminal economics of growth.
OECD EU JRC UK GO-Science Policy Horizons Singapore CSF Sitra Dubai Future Fdn US NIC UNDP UNESCO UNECE HLG-MOS World Bank Eurostat PARIS21 WEF RAND Bruegel CSIS McKinsey GI Millennium Project ČSÚ arXiv
The 188 documents are organised into 20 angles across five clusters — what foresight is and how to do it; foresight in and for the state; foresight × AI × agents; the data and statistics literature; and the economics of growth. The counts below are read straight from the library index.
| # | Angle | Cluster | Docs |
|---|---|---|---|
| 01 | Foundations of Foresight (What Is Foresight) | Foresight & method | 10 |
| 02 | Foresight Methodologies and Methods | Foresight & method | 11 |
| 03 | Strategic Foresight in Government and Public Policy | For the state | 12 |
| 04 | National and Government Foresight Institutions | For the state | 8 |
| 05 | Horizon Scanning and Emerging Issues | Foresight & method | 10 |
| 06 | Scenario Planning Practice and Frameworks | Foresight & method | 9 |
| 07 | Megatrends and Global Trends Reports | Data & trends | 8 |
| 08 | Foresight and AI (AI-Augmented Futures) | AI & agents | 11 |
| 09 | Agentic Foresight (AI Agents for Futures) | AI & agents | 9 |
| 10 | Anticipatory Governance and Innovation | For the state | 10 |
| 11 | Best Think Tanks — Futures and Foresight Output | For the state | 8 |
| 12 | Data Architecture for Foresight and Decision Intelligence | Data & trends | 10 |
| 13 | Data Governance and What Data to Collect | Data & trends | 9 |
| 14 | National Statistical Offices — Strategy and Modernization | Data & trends | 10 |
| 15 | Official Statistics and Data for Policy | Data & trends | 8 |
| 16 | Theories of Economic Growth | Growth economics | 10 |
| 17 | Drivers of Growth — What Makes Countries Grow | Growth economics | 8 |
| 18 | Economic Complexity and Industrial Policy | Growth economics | 10 |
| 19 | Technology and STI Foresight | AI & agents | 9 |
| 20 | Foresight for Resilience, Risk and Long-Term Decisions | For the state | 8 |
Seven terms recur throughout this report. Each names a discipline, not a buzzword — a specific way of using the future to make a better decision in the present.
The continuous, structured search for early signals of change — weak signals, trend-breaks, emerging issues — before they are obvious. The radar of a foresight system.
Building a small set of distinct, plausible futures to stress-test strategy against — not to predict which arrives, but to find the choices that hold up across all of them.
Fixing a desired future and working backwards to the dated sequence of decisions required to reach it — turning a long horizon into present action.
A frame separating today's maturing system (H1), the emerging future already present in pockets (H3), and the contested transition between them (H2) — where most strategy mistakes are made.
Faint, early indicators of a change that is not yet established — the tremor before the shift. The discipline is telling signal from noise by tracking which ones strengthen.
Governing so that anticipation is built into how decisions are made — routing foresight into the budget, legislation and strategy cycles, rather than reacting after the fact.
The human capability — recognised as a competence by UNESCO — to use the future to see the present differently. The operating system that turns scenarios from reports into reflexes.
A country does not need better forecasts. It needs a standing foresight capability — scan, model, stress-test, route to decision — run continuously by AI agents, because the cost of acting early is a fraction of the cost of reacting late. Compounded across 32 domains and across decades, that gap is the difference between a country carried by the future and one that authors it.