Europe’s Centicorn Opportunity
If you had asked me five years ago, what AGI would look like, I am pretty sure today’s LLMs would pass my then-expected bar. We blew through the Turing test, ungodly sums are being spent on compute, and every week another AI company blows through the “time to $100M revenue” record. However, for most people outside of tech, the world largely still looks and feels the same (aside from unsolicited frequent AI cold calls).
The largest pools of inefficiency in the global economy sit mostly untouched. They are not hiding in some undiscovered frontier. They are in the most boring places imaginable: Europe’s legacy industries, the unsexy backbone of the continent’s economy.
Government procurement offices, construction sites and manufacturing floors have been operating on roughly the same principles for decades. A public sector administrator in 2025 does fundamentally the same work as one did in 1995, just with slightly better software instead of paper. A construction project manager still struggles with the same coordination nightmares, equipped with software for procurement or collecting bids on “Leistungsverzeichnisse”.
Every now and then I step out of our little AI bubble and try to shadow a real-world occupation, usually my partner’s job managing large construction sites. I am shocked by how crowded “AI scribe for X” feels, yet how manual most of the actual processes remain. In their case, tasks are recorded and delegated via dictaphones, in-house secretaries handle scribing and typing out emails, schematic plans are printed on A1 paper and hung on construction sites (versioning is estimated by memory) and communication largely runs through WhatsApp. The systems haven’t changed because they have been “good enough”, inefficient yes, but functional.
Parts of the heavily manual nature of these businesses, and the resulting lack of digital processes, make it incredibly difficult for AI to extract context or embed itself into existing workflows. On a scale from bits to atoms, AI has excelled at automating businesses that operate primarily on the bits side (engineering, legal, customer support). But the more a business depends on the atoms side, the more AI falls apart. As a result, large parts of these atom-driven industries still rely on “CAPTCHA-like tasks” that humans perform daily. Until now, the cost of transformation has outweighed the visible benefit.
It is only a matter of time before that changes, not through the promise of AGI or robotics, but through practical solutions available today. OCR can now read handwritten documents; scribing solutions paired with agents can delegate tasks; procurement agents can manage vendor delays.
Startups in this space typically take one of two approaches: they either go horizontal, entering a wide range of markets through common tasks and transferable applications, or they go deeply vertical. The former usually comes with the “unlimited TAM” story, making it more immediately appealing to investors. The latter often looks like consulting in the early stages, given the level of hand-holding and “FDE” work required to understand workflows, access data, map processes and become embedded in daily operations.
For now, AI doesn’t need to do everything, it just needs to be better than the current alternative: humans spending half their week on repetitive admin tasks they openly despise.
The Efficiency Killer
The numbers, when you add them all up, are depressing. Studies show that 60% of knowledge workers spend at least a quarter of their week on routine administrative tasks: data entry, email management, form filling and copy-pasting between systems. The global average is even worse: office employees lose roughly 60 hours per month to easily automatable digital busywork. That’s nearly 4.5 months of productive time vanishing into the void every single year.
This is not a marginal problem. It is a structural tax on productivity that we have collectively accepted as the cost of doing business. They are mind-numbing tasks that prevent people from doing the work they were actually hired to do. Medical professionals didn’t train for years, just to then fill out forms or schedule appointments, yet that’s where much of their time goes.
Importantly, this is not about cutting headcount, it is about unlocking latent capacity, enabling the existing workforce to produce dramatically more value. A few percent of annual productivity growth would be transformative for Europe. Compounded over a decade, that differential adds up to trillions in economic value.
So why has this opportunity been sitting there, largely untapped?
Two reasons. First, until very recently, the technology was not quite good enough. Early RPA and automation tools were rigid, brittle and required extensive customisation. But more importantly, legacy industries move slowly. They are risk-averse, heavily regulated and organisationally complex.
Humans as an API
Over the past thirty years, companies have accumulated a patchwork of “systems of records”: ERP for operations, CRM for sales, HRM for people, plus dozens of specialised applications for procurement, compliance, analytics, etc. Each one was purchased and siloed to solve a specific problem, without sharing or talking to the other systems.
Software accounts for roughly 34% of corporate IT budgets on average. And much of that spend is pure waste. Studies find that 30% of software licenses go completely unused, and another 8% are used less than once a month. In the US and UK alone, an estimated $34B evaporates annually on unused licenses.
Employees spend their days flipping between disconnected systems, manually re-entering data because the applications don’t integrate. They become “human APIs”, copying information from one database to another. It is absurd, and everyone knows it, but still, we all do it daily, myself included.
AI startups promise to solve this: Instead of discrete systems that each demand their own interface and manual input, everyone pitches the AI layer that sits on top of all your data sources. Users interact through natural language or intelligent agents. The AI pulls information from wherever it lives, synthesises it and executes workflows across systems automatically (we deeply believe in this future, which is why we backed Spinnable AI 🖤 at pre-seed, building fully autonomous co-workers that conduct recurring multi-step tasks across your stack).
In this model, the specific “system of record” becomes less important. You don’t need separate software for every business function if AI agents can orchestrate tasks end-to-end, dynamically querying and updating the underlying databases as needed. The interface that used to matter so much, gets abstracted away. In a world where the AI handles transactions and maintains state, the legacy database becomes a background utility, not the primary tool.
At least, that is the VC narrative and dream of 2025. Now apply this dream to legacy industries.
Where the Money Actually Is
Many of the largest industries are already seeing extremely promising startups from Europe. Those with highly digital processes like law, insurance, certain segments of healthcare and banking seem to be the first to be disrupted. Yet most are still in their infancy.
Government and public procurement represent the single biggest “industry” in Europe. Government procurement alone accounts for roughly 14% of EU GDP, or about €2T annually. Two trillion euros, every year, flowing through tender processes, contract management and bureaucratic workflows that are still largely manual.
Procurement officers spend weeks reviewing bids, often defaulting to the lowest price because they lack the bandwidth to assess quality or risk. “Single bidding” rates have doubled in the last decade, meaning that 40%+ of tender processes now have only one provider bidding, which is never a good sign for buyer power or pricing.
Achieving even a 5% efficiency gain, achievable through AI that can make contracts more visible and accessible to suppliers, streamline tendering, flag anomalies and automate contract compliance, would free up roughly €100B.
Healthcare represents another 10% of European GDP. Doctors and nurses routinely spend too much time on documentation, typing, and bureaucratic tasks instead of patient care.
AI promises relief on both fronts. On the clinical side, diagnostic models trained on massive datasets can analyse medical images with expert-level accuracy, catching early-stage signs or conditions a human eye might miss. The more obvious and popular use-cases we are seeing are within the administrative side: NLP/ ASR transcribing doctor-patient conversations into structured medical records, and AI back-office assistants generating billing documentation (much of which has been enabled by Corti 🖤, which Heartcore backed at pre-seed in 2016 and is currently valued at $250M+). We are starting to see AI triage systems that prioritise emergency cases, telemedicine bots for preliminary consultations and remote monitoring devices that catch issues early. Doctors get more face time with patients because the paperwork is handled. Patients get faster, more personalised and hopefully better care.
Construction is the laggard. Productivity in European construction has been essentially flat for decades, worse than agriculture, worse than almost any other sector. It has remained stubbornly manual, fragmented and resistant to standardisation. Which also means the potential upside from AI and digitisation is enormous. AI agents can help pre-qualify sites and automate feasibility studies (being tackled by a stealth company 🖤, which Heartcore backed at seed in 2025), Building Information Modeling combined with AI can simulate construction projects and identify design clashes and cost overruns in advance. On-site, we are seeing early deployments of robotic bricklayers, drone-based inspections, and even pre-fab/ 3D-printed structures. Given construction’s trillion-euro scale in Europe, even incremental improvements matter.
Manufacturing accounts for roughly 15% of EU GDP. This sector has seen waves of automation before: predictive maintenance, IoT, PLCs, lean manufacturing, supply chain automation etc., but AI promises more. It is currently not there yet in terms of time series analysis or general forecasting beyond basic extrapolation, but we can see first signs of pattern recognition, optimise production schedules and dynamically reroute logistics around disruptions.
Despite extremely different use-cases, the pattern is consistent: these are massive markets, operating with decades-old processes, where even modest automation yields enormous returns.
Our Best Guess
Over the next 2-5 years, we expect a rapid proliferation of AI deployments across Europe’s legacy sectors. Many of the pilots and proofs-of-concept that have been running will mature into core operations. We will see hybrid systems initially, closer to AI augmentation, of AI handling routine tasks and humans maintaining critical oversight.
Gradually, the balance will shift. AI will take on more autonomy as trust builds and systems prove reliable. The workforce won’t shrink dramatically, but its composition will change. Fewer people doing data entry and paperwork, more people doing oversight, creative parts of the job and relationship management. The companies and institutions that navigate this transition well will see extraordinary productivity gains. Those that don’t will struggle to compete on pricing or margins. There is another thesis we are pursuing here, around AI first services, but to be discussed at another time.
What’s clear is that applied AI in Europe’s legacy industries will likely be one of the largest GDP impacts, with absolute justification for local champions and centicorn outcomes.