Version 0.1 · January 2026
Strategize,
Then Build,
But Fast
An Action Plan for Europe
Europe has enough strategy documents. What it doesn't have is enough people building the actual systems, services, and institutions that will define the agent era. This is a plan for builders—people who ship things, not people who write about shipping things.
The builder's diagnosis
Europe doesn't have a strategy problem. It has an execution problem. Walk into any ministry, any university, any major company in Europe and you'll find smart people who understand exactly what needs to be done. What you won't find is anyone actually doing it.
The reason is structural. European institutions are optimized for process, not output. They're designed to produce documents, hold meetings, form committees, and manage stakeholders. They're not designed to ship products, run experiments, or iterate based on evidence. The incentives reward participation, not results.
This worked fine when the world moved slowly. It doesn't work when the technology cycle is measured in months, not decades. By the time a European working group finishes its consultation process, the technology they were consulting about has been superseded twice.
The American approach—move fast, break things, figure out governance later—has its own problems. But at least it produces things. Europe produces PDFs.
This action plan is different. It's organized around what needs to be built, who can build it, and how fast it needs to happen. Everything else is commentary.
What we mean by building
Building isn't about technology. Building is about creating things that work—things that solve real problems for real people and get better over time. A well-designed curriculum is a built thing. A government service that actually functions is a built thing. A civic institution that helps people deliberate is a built thing. A company that employs people and serves customers is a built thing.
Building requires certain things that European institutions often lack:
Iteration speed
You can't build anything good on the first try. You have to ship something, learn from how it works in reality, and improve it. This requires short cycles—weeks, not years. Most European institutions operate on annual planning cycles, which means they iterate once per year. Builders iterate once per week.
Ownership and accountability
Someone has to be responsible for whether the thing works. Not responsible for following the process correctly—responsible for the outcome. European institutions diffuse responsibility across committees and working groups until no one is accountable for anything.
Permission to fail
Most things don't work the first time. Builders expect this and plan for it. They run pilots, measure results, kill what doesn't work, and double down on what does. European institutions treat any failure as a scandal, which means no one tries anything risky, which means nothing new gets built.
Domain expertise at the center
The people who understand the problem best need to be the ones building the solution. Not consulted. Not stakeholdered. Actually building. This means teachers building educational tools, public servants building government services, doctors building healthcare systems. The technologists enable; the domain experts lead.
Platforms, not point solutions
Every problem gets solved once and reused everywhere. When a city figures out how to process permits faster, that solution becomes a platform other cities can adopt. When a school develops a curriculum that works, it becomes a resource every school can use. Builders think in terms of leverage and scale.
The 90-day discipline
Here's a simple rule: if you can't ship something useful in 90 days, you're not building—you're planning. Planning feels productive but it's not. Only shipping is shipping.
This doesn't mean everything gets finished in 90 days. It means something real goes into the world every 90 days. A pilot program with real students. A government service with real users. A tool with real data flowing through it. Something you can measure, learn from, and improve.
The 90-day discipline forces certain behaviors:
You have to scope ruthlessly
You can't do everything, so you have to pick the one thing that matters most and do that. This is painful but necessary. Most European projects fail because they try to solve every problem at once and end up solving none.
You have to work with reality
When you have 90 days, you can't wait for perfect conditions. You work with the systems you have, the people you have, the constraints you have. You find workarounds. You make pragmatic choices. Purists never ship.
You have to actually measure
At the end of 90 days, you need to know if the thing worked. Not whether people felt good about it. Not whether the process was inclusive. Whether the actual outcome improved. This requires defining success upfront and instrumenting the work to measure it.
You have to make decisions
There's no time for endless deliberation. Someone has to decide, and then everyone has to execute. The decision might be wrong, but you'll find out quickly and correct course. A wrong decision made quickly is better than a right decision made too late.
Every initiative in this action plan is structured around 90-day cycles. If we can't show progress in 90 days, we're doing it wrong.
What needs to be built
Not strategies. Not frameworks. Not guidelines. Actual things that work. Here's the build list, organized by who needs to build it.
Government services that actually work
Most government services are designed around the needs of the bureaucracy, not the needs of citizens. Filing a permit, claiming a benefit, registering a business—these things take months when they should take minutes. Not because public servants are lazy, but because the systems are broken.
AI agents can fix this. An agent can read a permit application, check it against requirements, identify missing information, request clarifications, and route the complete application to a human for final approval—all in hours instead of weeks. The human stays in control of the decision; the agent handles the drudge work.
But this requires building actual systems, not writing reports about systems. It requires putting a small team in a real government office, letting them build for 90 days, measuring what happens, and iterating.
The build spec
What: Agent-powered service delivery for high-volume government services (permits, benefits, registration, scheduling)
How: Embed small teams (3–5 people) inside actual government offices. Give them 90 days to build a working prototype. Measure cycle time, error rate, and citizen satisfaction. If it works, expand. If it doesn't, learn and try again.
First 90 days: Pick 5 services across 5 countries. Deploy teams. Build prototypes. Run pilots with real citizens.
Success looks like: At least 2 services with measurable cycle time improvements (>30%). Published case studies including failures.
Who builds it: Small teams of builders working inside public institutions—not consultants writing recommendations from outside.
Schools that teach people to build
The current education system is optimized for producing people who follow instructions well. This made sense when most jobs required following instructions. It doesn't make sense when AI can follow instructions better than any human.
What humans will still be uniquely good at: asking the right questions, exercising judgment in ambiguous situations, coordinating with other humans to solve novel problems, and building new things. These skills need to be taught deliberately, not assumed to emerge from a traditional curriculum.
This isn't about "adding AI to education." It's about fundamentally rethinking what education is for. The goal isn't to produce students who can pass exams; it's to produce people who can build things—systems, institutions, solutions, companies, communities.
The build spec
What: A "builder curriculum" that teaches reasoning, judgment, coordination, and creation—not just recall and compliance.
How: Partner with schools that want to experiment. Work with teachers who are frustrated with the status quo. Build lesson plans, projects, and assessments together. Test with real students. Measure outcomes. Iterate.
First 90 days: Establish 10 pilot schools across Europe. Develop initial curriculum modules for critical thinking, project-based learning, and deliberation skills. Train participating teachers.
Success looks like: Students who can construct and defend arguments, collaborate on complex projects, and use AI tools to enhance (not replace) their thinking. Teachers who want to keep using the curriculum.
Who builds it: Teachers and educators who want to change things, working with curriculum designers and assessment experts. Not education ministries issuing top-down mandates.
Infrastructure for democratic deliberation
Democracy depends on people being able to discuss, disagree, and decide together. This requires shared understanding of facts, the ability to hear opposing views, and mechanisms for reaching collective decisions. All of these are under attack.
Social media optimizes for engagement, which means outrage. AI can generate persuasive content at scale, which means manipulation gets cheaper. Filter bubbles mean people never encounter views they disagree with. The infrastructure of public discourse is corroding.
The solution isn't more moderation—you can't moderate your way to a healthy democracy. The solution is building new infrastructure for deliberation: tools that help people find common ground, platforms that expose people to different perspectives, institutions that facilitate productive disagreement.
The build spec
What: Tools and institutions that enable productive democratic deliberation at scale—citizen assemblies, deliberative polls, structured debate platforms, evidence-sharing systems.
How: Start with municipalities. They're small enough to experiment, close enough to citizens to get real feedback, and hungry for solutions. Build deliberation infrastructure for local decisions—budgets, planning, services. Prove it works at the local level before trying to scale.
First 90 days: Partner with 5 municipalities to run AI-enhanced citizen assemblies on real policy questions. Build the supporting tools (evidence packs, argument mapping, voting mechanisms). Document the process and outcomes.
Success looks like: Citizens who feel heard. Decisions that have broader legitimacy. A replicable model that other municipalities can adopt.
Who builds it: Democracy innovators, civic technologists, and municipal governments working together. Not tech companies imposing solutions from outside.
Safety infrastructure that's more than paperwork
Europe passed the AI Act. Great. But laws don't make systems safe; engineers do. Right now, most "AI safety" is compliance theater: checkboxes that get ticked, documents that get filed, audits that rubber-stamp whatever's already happening. None of this actually makes AI systems safer.
Real safety requires building real things: test suites that actually stress-test systems before deployment. Monitoring tools that detect when systems are behaving unexpectedly. Red teams that try to break systems before attackers do. Incident response playbooks that work in practice, not just in theory. Kill switches that actually kill.
The goal is to make "safe AI" a competitive advantage, not a compliance burden. European AI should be the AI you trust because it's been tested, monitored, and proven—not because it has a certification badge.
The build spec
What: Practical safety infrastructure—evaluation tools, monitoring systems, red team capacity, incident response protocols—that can be used across all AI deployments.
How: Build open-source tools that any organization can use. Create shared evaluation benchmarks that become industry standards. Fund independent testing labs that can evaluate systems objectively. Publish results, including failures.
First 90 days: Launch an open-source toolkit for agent safety (logging, monitoring, permissions, kill switches). Establish 3 independent evaluation labs with real testing capacity. Run first evaluations on public-sector pilot systems.
Success looks like: A practical safety standard that organizations actually adopt because it makes their systems better, not just because regulators require it. Published evidence from real evaluations.
Who builds it: Safety engineers, security researchers, and AI developers working together—not compliance consultants generating paperwork.
Companies that build with, not sell to
Europe has plenty of traditional industries with deep expertise and established relationships: professional services, healthcare, manufacturing, logistics. What these industries lack is the ability to build AI-native products and services themselves. So they become customers of American tech companies instead of builders in their own right.
The opportunity isn't to sell AI to European businesses; it's to build AI businesses with European domain experts. The lawyer who understands contract law becomes a co-founder of the legal AI company. The manufacturer who understands supply chains becomes a co-founder of the logistics AI company. The domain expertise is in Europe; it just needs to be combined with building capability.
This creates a different kind of company—one that's rooted in real expertise, solving real problems, with real customers from day one. Not a startup chasing product-market fit, but a partnership between builders and domain experts creating something neither could create alone.
The build spec
What: AI-native companies built through partnerships between technology builders and domain experts in traditional industries.
How: Identify domain experts in language-intensive industries (legal, healthcare, professional services) who see the opportunity for transformation. Partner with them as co-founders, not customers. Build the AI product together, using their expertise and existing relationships. Scale from there.
First 90 days: Establish 5 partnerships in priority verticals. Build first products with real users. Validate business models with real revenue.
Success looks like: European AI companies that are globally competitive because they combine world-class AI with world-class domain expertise. Founders who are operators, not just technologists.
Who builds it: Builders who know how to ship products and operators who know their industries—working together as equals.
Transition infrastructure that actually works
AI will change what work looks like. Some tasks will be automated. Some jobs will disappear. New jobs will emerge. This transition will be painful for many people if it's not managed well. The question is whether the gains from AI productivity accrue only to capital, or whether they're shared broadly.
Europe has a choice. It can pretend this isn't happening and let the market sort it out—which will mean concentrated gains and dispersed losses. Or it can build the infrastructure for a fair transition: training that leads to real jobs, income support that enables risk-taking, new forms of work that leverage uniquely human capabilities.
This isn't about stopping automation; it's about making sure automation serves everyone, not just shareholders. The goal is an economy where AI makes everyone more productive and everyone shares in the gains.
The build spec
What: Training programs that actually lead to employment. Transition support that enables people to take risks. Worker voice in automation decisions.
How: Partner with companies deploying AI to design transition plans from the start. Work with unions and worker representatives as co-designers, not afterthoughts. Build training programs tied to real job opportunities. Measure outcomes—not just training completion, but employment and wages.
First 90 days: Establish 5 "transition labs" in industries undergoing rapid AI adoption. Co-design training and support programs with workers and employers. Track outcomes rigorously.
Success looks like: Workers who transition successfully to new roles with equal or better pay. Companies that adopt AI with full worker buy-in. Unions that see AI as an opportunity, not a threat.
Who builds it: Employers, unions, training providers, and government working together—with workers at the center.
Resilience against information warfare
The next generation of propaganda will be AI-generated, personalized, and indistinguishable from authentic content. Deepfakes of politicians. Synthetic news articles. Armies of fake social media accounts that sound human. Targeted messages designed to exploit individual psychological vulnerabilities.
You can't fact-check your way out of this. By the time you've debunked one fake, a thousand more have been generated. The only thing that scales is building resilience: citizens who have the judgment to question suspicious content, institutions that can detect and respond to coordinated campaigns, and a media environment that makes it harder to manipulate.
This is infrastructure for democratic survival. If we don't build it, the information environment will become so polluted that democratic discourse becomes impossible.
The build spec
What: Detection and response capacity for AI-enabled manipulation. Media literacy at scale. Transparency infrastructure for political content.
How: Build a European capacity to detect coordinated manipulation campaigns in real-time. Require transparency for AI-generated political content. Invest massively in media literacy—not just for kids, but for adults. Run realistic exercises before every major election.
First 90 days: Establish a cross-border "information resilience cell" with real detection capacity. Run first realistic exercise simulating an AI-enabled influence campaign. Deploy media literacy programs in partnership with libraries and community organizations.
Success looks like: Coordinated campaigns detected and responded to within hours, not weeks. Citizens who can identify manipulation attempts. A public that maintains trust in democratic institutions despite sustained attacks.
Who builds it: Security agencies, researchers, platforms, and civil society working together—not in silos.
Why Europe can actually lead
This isn't European delusion. Europe has real advantages that aren't being exploited. The question is whether we'll capitalize on them or keep complaining about American dominance.
Regulatory maturity
GDPR was annoying for tech companies. It was also world-leading. The AI Act will be the same. Europe knows how to build regulatory frameworks that actually work—frameworks that set rules without killing innovation. This becomes a massive advantage when the world wakes up to AI risks and needs governance models. We have them.
Deep domain expertise
The most valuable AI companies won't be the ones building foundation models; they'll be the ones applying AI to real problems in real industries. Europe has world-class expertise in manufacturing, healthcare, professional services, logistics, energy. This expertise is the raw material for building AI applications that actually matter.
Market fragmentation as forcing function
The American market is one language, one legal system, one culture. The European market is 27 countries with different languages, laws, and customs. This is annoying—but it forces you to build robust, adaptable platforms instead of point solutions. Products that work across Europe can work anywhere.
Underserved markets
Silicon Valley is obsessed with digital-native companies. But the biggest opportunity isn't building for startups; it's digitizing the giants. Across Europe, there are thousands of established businesses with real customers, real revenue, and real expertise—hungry for AI transformation but ignored by American tech. That's our market.
Patient capital culture
American VC wants 10x returns in 5 years. This pressure creates companies that optimize for growth over sustainability, hype over substance. European capital can be more patient—building real companies with real business models, not just growth stories for the next funding round.
The question isn't whether Europe can compete. It's whether we'll stop talking about competing and start actually building.
The build timeline
Every 90 days, something real should ship. Here's what the timeline looks like if we're serious.
First builds ship
→ 5 government service pilots running with real citizens
→ 10 pilot schools using first curriculum modules
→ 5 municipal citizen assemblies with AI-enhanced deliberation
→ Open-source safety toolkit v1 released
→ First domain-expert partnerships established in priority verticals
First lessons learned
→ Published results from government pilots—including failures
→ Curriculum iteration based on teacher and student feedback
→ First products from domain-expert partnerships in market
→ Safety evaluations completed on all public-sector pilots
→ Transition labs established in 5 high-automation industries
Scale what works
→ Successful government pilots expanded to additional services/countries
→ Curriculum adopted by 50+ schools based on proven results
→ Deliberation infrastructure used by 20+ municipalities
→ Safety standards adopted by early movers in private sector
→ First information resilience exercise run before national election
Critical mass
→ AI-powered public services are standard, not experimental
→ Builder curriculum available to any school that wants it
→ Deliberation infrastructure becomes go-to for contested decisions
→ European safety standard recognized as global gold standard
→ 20+ European AI companies built through domain-expert partnerships
The goal
→ European institutions are globally recognized as high-trust, high-capability
→ European education produces builders, not just employees
→ European democracy has survived the AI-enabled manipulation era
→ European companies compete globally on quality, not just price
→ AI productivity gains are broadly shared, not concentrated
What happens if we don't build
The failure mode isn't mysterious. It's already visible. If we keep producing strategies instead of products, here's what happens:
Government stays broken
Other countries automate and improve; European citizens wait months for permits while algorithms serve citizens elsewhere in minutes. Trust in public institutions continues to erode. The best people leave public service. The state's capacity to respond to crises—pandemics, climate, economic shocks—gets weaker.
Education fails a generation
Kids graduate with skills optimized for a world that no longer exists. They can follow instructions but can't solve novel problems. They can pass exams but can't build anything. Youth unemployment rises. Brain drain accelerates. The next generation of European builders never emerges.
Democracy gets hacked
Without resilience to AI-powered manipulation, elections become increasingly influenced by whoever has the best synthetic propaganda. Public discourse fragments into manufactured bubbles. Trust in institutions collapses. Authoritarian movements rise on waves of artificial outrage.
Europe becomes a technology colony
All critical AI infrastructure gets owned and operated by American and Chinese companies. European data flows to foreign servers. European regulations become irrelevant because there's no European capability to enforce them. "Digital sovereignty" becomes a punchline.
The transition is brutal
AI productivity gains accrue to capital while workers get displaced. Inequality rises. Political backlash intensifies. Either we get a neo-Luddite movement that tries to ban AI, or we get an oligarchy that captures all the gains. Neither is a good outcome.
This isn't inevitable. But avoiding it requires building now, not strategizing for later.
The builder's creed
If you want to be part of this, you need to believe certain things:
Shipping is more important than planning.
A mediocre thing that exists is worth more than a perfect thing that doesn't. You can improve what exists. You can't improve what's still in a planning document.
Iteration is how you get quality.
Nobody gets it right the first time. Quality comes from shipping, measuring, learning, and improving—over and over. The goal is fast cycles, not perfect launches.
Domain expertise beats technology expertise.
The hardest part isn't the technology; it's understanding the problem well enough to build the right solution. Work with the people who understand the problem deeply.
Platforms beat point solutions.
Every problem should be solved once and reused everywhere. Think in terms of leverage. Build things that enable others to build more things.
Accountability is non-negotiable.
Someone's name is attached to every initiative. Not a committee. Not a working group. A person who will answer for whether it worked.
Speed is a moral imperative.
Every day we delay, the window narrows. The technology keeps advancing. The stakes keep rising. Moving fast isn't reckless—it's necessary.
Europe can lead if it chooses to.
We have everything we need—the expertise, the capital, the regulatory capacity, the values. What we lack is the will to build. That's what needs to change.
Join the build
This isn't a document to admire. It's a recruitment call.
If you're inside a government and you want to build services that actually work—we need you. If you're a teacher who wants to build a curriculum for the future—we need you. If you're a domain expert who sees the opportunity to build something transformative in your industry—we need you. If you're a builder who wants to work on problems that matter—we need you.
The strategy is simple: find people who want to build, give them permission to build, and get out of their way. The rest is execution.
Europe's future won't be written in strategy documents. It'll be built by people who ship things. Are you one of them?