As of late 2025, 88% of organizations are using AI in at least one business function, up from just 55% in 2023. But while this is the case, only 7% have fully scaled AI across their organization, while 62% are still experimenting or piloting. Most companies are drifting with AI—using tools because they’re available or trendy, not because they’re part of a coherent strategy.

While some may hope that the AI wave will pass, the truth is the opposite: AI is here to stay. Companies that don’t embrace AI will likely fall behind. But the companies that are building intentionally will pull ahead.

Gartner predicts that by 2028, organizations with sustained AI-first strategies will achieve 25% better business results than their peers. As we enter 2026, the window to move from drifting to building is closing. It’s a certainty that AI will reshape your industry—the only question is whether you’ll direct that change or react to it.

Why strategy can’t wait

Many organizations still question the need for an AI strategy. They don’t see direct applications, are concerned about the unknowns, or think AI is only for coding.

But AI is already embedded in industries far beyond traditional tech. In healthcare, for example, AI systems can already examine stroke patients’ brains with twice the accuracy of human professionals and detect epilepsy lesions that radiologists miss. In education, 57% of higher education institutions are prioritizing AI for digital acceleration and personalized learning.

Here’s an uncomfortable truth: your organization is already using AI, whether you have an AI strategy in place or not. Many employees are using AI tools to complete their work. Some are using approved tools according to clear protocols. Others are pasting proprietary data into the free version of ChatGPT and copying the results blindly.

The absence of a strategy creates two problems. First, employees using AI haphazardly create security vulnerabilities, compliance issues, and what’s now called “workslop”—low-quality output that looks professional but lacks depth and substance. Second, employees who should be using AI aren’t, either because they don’t know which tools are approved or because they fear being replaced by them.

Strategy solves both problems. It channels AI use toward value creation while establishing guardrails. It helps employees transform anxiety into capability. And most importantly, it purposefully redefines what good work means in your organization—because AI is already redefining roles, whether you acknowledge it or not

3 strategic imperatives for 2026

Building an intentional AI strategy requires more than just declaring “we’re AI-first” on your homepage. It requires structural changes across three key dimensions:

1. Treat AI as infrastructure, not innovation.

The first mistake organizations make is treating AI as something special—like a center of excellence. This thinking leads to AI remaining peripheral.

Instead, AI must become infrastructure: as unremarkable as email, as essential as your Slack channels. Your leadership team should actively identify processes to automate, not to reduce headcount, but to redirect human effort toward higher-value work. When AI handles the routine tasks, your people can focus on the complex, creative, and strategic work that actually moves your business forward.

This shift must come from a cultural change. Software engineers should reach for AI coding assistants by default. Finance teams should assume AI will flag anomalies in real-time. Support teams should expect AI to handle basic questions. The goal is to make AI invisible, not impressive, as it’s woven into how work gets done every day.

2. Establish baseline AI competency standards.

You wouldn’t hire engineers who can’t use version control. You shouldn’t hire engineers who can’t effectively use AI coding tools.

As AI streamlines routine coding tasks, the role of software developers is evolving. Developers who master AI tools can focus on architecture, complex problem-solving, and work that requires deep contextual understanding—the work AI can’t handle. Developers who don’t will struggle to keep pace.

Having baseline standards doesn’t mean everyone needs to be an AI researcher. It means establishing basic competencies: understanding which tasks AI excels at, how to write effective prompts, how to validate AI output, and when to override AI recommendations. These should be hiring criteria, onboarding requirements, training mandates, and promotion considerations.

The alternative is a growing skills gap within your own organization, resulting in team members who’ve embraced AI becoming more productive, while those who haven’t struggle with the basics.

3. Implement guardrails before incidents happen.

AI offers organizations transformative productivity and efficiency gains. It also introduces serious risks: copyright infringement lawsuits, data leakage, hallucinated information presented as fact, and security vulnerabilities, to name just a few. The question isn’t whether these risks exist, but rather whether you’ll address them proactively or reactively.

Being reactive is expensive. It means learning about data leaks from customer complaints, discovering copyright issues through legal threats, or finding that proprietary code was used to train external models after the fact.

Being proactive means establishing clear policies now: which AI tools are approved, what data can be shared with them, how to verify AI-generated output, and what happens when the policies are violated. This requires training employees not just on how to use AI, but on how to use it responsibly within your organization’s risk tolerance.

Guardrails, when properly designed and implemented, aren’t about limiting AI use, but about enabling it safely at scale.

Building, not chasing

Becoming AI-first requires changes that feel difficult because transformation of any kind is challenging. But organizations that invest in deploying AI capabilities to existing operations are fundamentally redesigning how work gets done.

At MacPaw, we don’t treat AI as an add-on or a trend to ride. It’s the foundation of how we build products, support customers, and operate as a company. This didn’t happen because AI became fashionable. It happened because we recognize that the companies that integrate AI deepest and earliest will have compounding advantages that competitors can’t easily replicate.

The difference between 2026 and 2025 won’t be more AI tools; rather, it will be which organizations stopped drifting and started building. The strategy you implement now determines which category you’ll be in.

By Volodymyr Kubytskyi, Director of AI at MacPaw

UK-based qualitative analytics startup WholeSum has raised £730,000, combining grant funding from Women TechEU with a pre-seed round led by Twin Path Ventures. The round also included participation from SFC and strategic angel investors via Ventures Together, including founders and operators from JustPark, Episode 1, ClearScore, and Prolific.

A significant share of organisational data is unstructured, yet many teams lack consistent methods for analysing large volumes of text. Key information is often found in transcripts, open-ended survey responses, online discussions, and customer feedback, while analysis frequently relies on time-intensive manual processes or automated summaries that are difficult to verify or reproduce.

WholeSum provides an AI-based analytics layer for qualitative data, converting large amounts of free text into statistically supported, auditable outputs. Founded by Emily Kucharski and Dr Adam Kucharski, the company combines experience in commercial and public sector insights with expertise in statistical inference and machine learning.

The platform is designed for direct API integration and produces quantified, reproducible insights that can be incorporated into existing analytics workflows. Tasks that would typically take weeks of manual analysis can be completed more quickly and then examined further using statistical tools to support decision-making.

Through collaborations with organisations including Imperial College London and Female Founders Rise in partnership with Barclays, WholeSum has found that high-value insights are often contained within unstructured data rather than simplified survey metrics. However, identifying these insights consistently and at scale has historically been difficult.

John Spindler, Partner at Twin Path Ventures, said that qualitative analysis has long depended on manual processes and inconsistent approaches. He noted that WholeSum introduces a more systematic and automated framework and described the platform as an important foundation for how qualitative evidence may be produced and relied upon in the future.

The platform is primarily aimed at sectors such as research, healthcare, and financial services, where reliable qualitative evidence plays an important role in decision-making and outcomes. According to internal evaluations, WholeSum demonstrates improved performance compared with several established reasoning models on certain datasets, offering faster processing and lower theme attribution error while maintaining reproducible results.

The company plans to use the new funding to advance product development, expand its science and engineering teams, and scale early enterprise deployments.

It’s remarkable to think that a year or so ago, securing a coveted page-one search engine ranking was considered the marketer’s holy grail. No easy feat, getting there required absolute precision, constant fine-tuning and the perfect mix of keywords and backlinks.

Of course, today that world looks rather different. As AI-driven search experiences dominate, they have blown the doors off the old search-first mindset. Brand discovery is no longer about ranking for blue links; it is about appearing in AI-generated answers. And to do that, brands need content that machines can interpret just as easily as humans, something many are far from ready for.

The consequences are already clear. Traffic is dipping, zero-click answers are overshadowing genuine expertise, and outdated website content is reappearing in AI-generated responses. Brands are losing credibility and control of their own narrative long before a customer ever reaches their site.

Cue an inevitable onslaught of debate over whether AI has, in fact, “killed the website”, as the internet supposedly heads towards a future where machine-led noise takes precedence, and quality information and authentic human interaction fade away.

But the truth is far less bleak. AI isn’t a website killer. Rather, it has simply exposed the brittle foundations, outdated structures and messy content models that have been holding websites back for years. AI is not an ending; it is a turning point, offering brands a chance to rebuild their digital foundations and finally do content right.

Be relevant, not just ranked

Previously, brands could get away with a content ecosystem that was outdated and internally inconsistent. Take a look at almost any site, and chances are you’ll find old news releases or dusty product specs. While humans might overlook a stale blog post from 2018 or incorrect pricing buried three levels deep, AI systems do not.

Algorithms now crawl and compare information at a scale and speed no person can match. They evaluate how current your content is, whether your messaging stays consistent across pages and channels, and how authoritative your knowledge base looks against competitors. In other words, AI is no longer just reading what you publish; it is judging whether your entire digital presence adds up.

Generative Experience Optimisation (GEO) is the roadmap for this new landscape. At its core, GEO is about ensuring that content is relevant, coherent and authoritative so that AI can accurately interpret and surface it. For marketers, this translates into designing content that prioritises clarity, consistency and authority: clear language, cohesive messaging and demonstrable expertise that generative AI can reliably recognise and elevate in its responses.

Crucially, GEO also shifts the focus from individual pages to the overall logic of your digital estate. It asks whether every touchpoint, including product pages, help centres, blogs and social media copy, tells the same story and uses the same definitions. When that ecosystem hangs together, AI systems can more easily map questions to your answers, quote you as a primary source and elevate your brand in aggregated, AI-generated experiences.

Historic bottlenecks

This shift is also placing a sharper focus on content governance. Across many organisations, individual channels still operate in isolation. Web teams publish one thing, marketing says another, and brand information is fragmented.

AI interprets this inconsistency as a lack of authority and reduces trust accordingly. This is further compounded by legacy CMS systems that were never designed for the dynamic, multi-channel, AI-driven environment we now inhabit, making timely updates slow and consistency nearly impossible.

To remain visible and trusted in this discovery environment, brands need content systems that are flexibly structured and built for speed. This strengthens the case for composable, API-driven architectures, technology stacks made from interchangeable, best-in-class components rather than rigid, all-in-one systems.

Composable CMS platforms allow marketers to update information once and distribute it everywhere, ensuring that every channel, region and touchpoint reflects the same up-to-date source of truth. They also encourage more thoughtful content modelling and governance, which is essential for maintaining accuracy and coherence across increasingly complex digital ecosystems.

A composable stack also makes it easier to introduce new tools, such as personalisation engines, automation layers, AI assistants or creative workflows, without disrupting the entire system. This level of flexibility gives marketing teams the agility they need to operate confidently in an AI-first discovery landscape, where freshness, structure and adaptability determine relevance.

A catalyst for change

The rise of AI-driven discovery does not signal the end of the web; it signals an opportunity. Brands that embrace coherent and authoritative content, supported by flexible, composable systems, will not only survive but thrive. By prioritising relevance over rankings and designing digital ecosystems built for clarity and consistency, businesses can regain control of their narrative, earn trust and increase visibility in an AI-first world. In this way, AI should be viewed not as a disruptor but as a catalyst for positive and much-needed change.

The post AI didn’t break your website: It exposed its weaknesses appeared first on EU-Startups.

Hybrid-casual puzzle game Pixel Flow has closed a seed round with participation from Akın Babayiğit and e2vc. The amount of the investment was not disclosed.

Three months after launch, the game has reached seven-figure daily revenue and entered the Top 25 of the US App Store Top Grossing chart, reflecting the team’s accelerated scaling efforts and the commercial potential of the hybrid-casual category.

Pixel Flow was developed by Kübra Gundogan and Emre Çelik, an Istanbul-based team previously involved in the game Twisted Tangle. Drawing on their experience in product development, go-to-market execution, and scaling within the hybrid-casual category, the team has supported the game’s early adoption and growth.

Co-founder and CEO Kübra Gundogan said the project emerged from the team’s long-term focus on creating original game mechanics. She noted that the game’s reception reflects their analysis of player behaviour and the insights derived from that work, and that Pixel Flow represents a new core mechanic and design direction for the team.

Commenting on the investment, Akın Babayiğit said that Pixel Flow has attracted strong attention across the industry, describing it as a notable recent success from the region and pointing to both its performance metrics and the originality of its core mechanic.

Following the investment, the team expanded its scaling efforts, indicating that hybrid-casual games can reach wide audiences while generating sustainable revenue.

In 2026, many startup founders are facing the same uncomfortable truth. Their product may be technically solid, and their team may be shipping fast, but growth stalls the moment AI agents become the first touchpoint in the customer journey. The interface has changed, and with it, so should we.

In previous years, you optimised for the App Store or Google search. Today, AI agents, AI-first browsers such as Atlas, and workflow tools inside Slack, Teams, and Notion are the default interfaces for knowledge and software. The first user of your product is now an AI system deciding whether humans will ever see you. If AI agents cannot understand or operate your product, you become invisible, no matter how good the human UX is.

As a result, you need to optimise for the AI layer that sits between you and your customer. But how do you speak the language that teams care about? You become AI-native.

Becoming AI-native is one of the best chances for startups to punch above their weight against incumbents. To help you get ahead of the market, this piece offers a practical definition of AI-native, a simple self-assessment blueprint, and a founder’s view on what needs to change in hiring, team structure, and culture in this new AI-powered era.

What AI-native actually means in practice

AI-native is a confusing term. Most startups have integrated some form of AI to speed up their day-to-day operations. That is not being AI-native. That is being AI-enhanced. The difference is fairly straightforward.

• AI-enhanced: This is internal. AI is used inside your company to speed up work, but the product itself still assumes a human user.
• AI-native: Your product is built so that AI systems outside your company can reliably read, query, and act on it.

Essentially, AI-enhanced makes you faster, while AI-native makes you discoverable and interoperable. The difference is fundamental to how you operate as a business, from messaging to product design, sales, marketing, and partnerships.

How to be AI-native

So how can you tell whether your product is AI-native or not? Here is what you need.

Machine-consumable surfaces

• Consistent structured outputs, stable schemas, and robust APIs.
• Semantic clarity with clear names, types, and contracts so agents can reason without hacks.

Documentation and knowledge for machines

• Documentation and FAQs written so that LLMs can parse them. They should be up to date, structured, and low in ambiguity.
• Internal knowledge formatted as graphs, schemas, or clean text, not just slide decks.

Agent-friendly interfaces

• Interfaces that support programmable navigation through links, IDs, and action endpoints, rather than relying only on visual affordances.
• Clear ways for agents to trigger workflows and retrieve results without scraping pixels.

Workflows optimised for AI decisions

• A default assumption that an agent will orchestrate multiple steps, not a human clicking through screens.
• Predictable timings, idempotent actions, and observable states so agents can recover from failure.

Predictability and clarity in responses

• Stable response shapes and clear error modes so agents can integrate once and trust the system.
• Think contract testing for agents, not just something that is good enough for a human reading a blog.

As you can see, becoming AI-native is a fundamental structural choice. It cannot be an add-on or a feature.

How startups can win big

You might be thinking that this gives startups a massive advantage over incumbents, and you would be right.

Startups do not have to overcome legacy systems. They are not carrying ten years of UI conventions, data debt, and one-off integrations. They can design clean schemas, transparent logic, and agent entry points from day one. Startups also tend to have smaller teams, which enables cheaper and faster experimentation with schemas, APIs, and AI-facing documentation.

This means startups can regularly test how well AI agents route to them in real workflows. In incumbents, everything runs through committees. They cannot pivot quickly, and they cannot test in the same way.

We have already seen this at Tastewise. When ChatGPT’s browser, Atlas, launched, many competitors had to scramble to adapt their content to this new AI-driven environment. Tastewise had already built an approach designed to thrive in AI environments, which put us in a strong position to scale in this new era.

AI agents tend to choose their preferred tools and stick with them. If you become an AI agent’s go-to option in your category, your ability to scale increases rapidly, as the agent does much of the heavy lifting. By making this transition early, you position yourself ahead of the industry and ahead of major changes that will shape it going forward.

Five questions to ask yourself

1. Can an AI agent understand what our product does from our public documentation in under 30 seconds?
2. Are our main outputs and events available as structured data with stable contracts?
3. If a copilot inside a customer’s workspace searched for tools like ours, would it find us and know how to call us?
4. Do we know which parts of our product are hardest for a machine to interpret today?
5. Is there a named owner responsible for AI legibility across product, documentation, and data?

If a few of these questions made you uncomfortable, that is a useful signal. Most teams are still designing for humans and hoping AI agents will improvise around the gaps. They will not. The shift to AI-native starts inside the company, long before it appears on your roadmap or homepage.

What changes inside your company

Hiring: An AI-native product needs fewer people obsessing over pixels and more people obsessing over structure. You want engineers who think in contracts, schemas, and events, not just screens. You want product managers who understand how LLMs read, rank, and chain calls. You also want people who enjoy naming things clearly and documenting why systems behave the way they do.

Front-end work still matters, but it sits on top of a stable, machine-readable core. When you are AI-native, the surface is the final layer you polish, not the only layer you invest in.

Team structure: Instead of organising purely around features, you begin organising around knowledge surfaces. For example, one team might own pricing logic and every surface where pricing appears, including APIs and documentation used by agents. Another might own customer state and lifecycle events and expose them in predictable ways. Another might own documentation, taxonomies, and examples and treat them as a product.

Each team has a clear mandate. Humans should understand their domain, and AI agents should be able to navigate it without hacks.

Culture: AI-native is a mindset as much as a technology stack. In practice, that means writing documentation and internal notes with headings, definitions, and context that a model can follow without guessing. It means treating internal decisions as things that will be read later by both a machine and a new teammate. It means defaulting to observable systems where you can explain, in plain language, what happened when an agent interacted with your product.

Transparency stops being a slogan and becomes the way you make your product legible to both humans and machines.

Why this becomes your edge

When AI browsers and agents started to matter, many companies discovered they had a visibility problem. Their content was locked in formats that worked for humans and little else. They had to rush to restructure their knowledge so agents could even find them.

At Tastewise, we felt the advantage of building for AI consumption early. When tools like Atlas entered the picture, our structured, machine-friendly approach meant AI environments could use our outputs without a rebuild. That did not make us smarter than our competitors. It meant we had done the groundwork.

The same opportunity exists for any startup willing to design for AI as the first user.

AI-native as the default

Over the next few years, AI agents will scan your documentation, test your APIs, compare you to alternatives, and decide what to surface to the humans you care about. Human UX still matters, but AI UX determines whether anyone ever sees that beautiful interface.

Start small. Pick one area of your product, make it fully legible to an AI agent, and give someone ownership of that work. Then repeat.

The real question for 2026 is simple. When an AI system looks at your product, does it know what to do with you? If the answer is yes, you are already ahead.

The post What AI-native means for startups in 2026, and why it is not just for big tech appeared first on EU-Startups.

There are around 350,000 people diagnosed with primary brain cancer every year, and 250,000 people die from it. Despite decades of progress in medicine, the tools primarily used to access, diagnose, and treat the brain remain limited — until now.

Robeauté is a Paris-based MedTech startup developing a new class of therapeutic microrobots designed to diagnose, treat, and monitor the brain with unprecedented flexibility. Operating at the intersection of robotics, physics, materials science, chemistry, biology, and medicine, the company has developed a modular medical device built around a universal robotic core with interchangeable micro-extensions.

I spoke to co-founder and COO Joana Cartocci to learn all about it. 

From targeted drug delivery to live data collection

Roughly the size of a grain of rice, Robeauté’s microrobots can navigate curved, non-linear paths through the brain’s extracellular matrix, safely reaching multiple sites of interest.

Depending on the pathology, each device can be equipped for a specific mission — delivering therapeutic molecules, implanting electrodes, or collecting cellular and live data via embedded sensors.

This modular architecture allows a single platform to be adapted across a wide range of clinical applications, from tissue sampling and targeted drug delivery to electrode implantation and real-time data collection from deep within the brain — opening new possibilities for both treatment and understanding of complex neuropathologies.

From extreme environments to the human brain

Robeauté’s founder, Bertrand Duplat, spent more than 30 years working in robotics, including at McGill University and the European Space Agency, specialising in robots designed for extreme environments. Earlier in his career, he also founded 3D software company Virtools, which Dassault Systèmes later acquired.

After decades working on undersea, nuclear, space, and archaeological robotics, Duplat decided to apply his expertise to medicine — a decision catalysed by his mother’s diagnosis with glioblastoma. He went on to found Robeauté with co-founder Joana Cartocci, an operations specialist.

“He had been doing robotics for 30 years in extreme environments and decided to put that experience to good use,” Cartocci said.

“That’s around the time we met. He asked me to join and lead operations — everything that’s not scientific — and we’ve been at it since 2017.”

Why most academic microrobots never leave the lab 

Crucially, Cartocci comes from an operational background. She notes that a lot of founders spin out of labs and struggle with the transition to entrepreneurship.

"It’s a completely different skill set. Time management, identity, and team alignment — those things make or break a company. My role is to take that weight off my co-founder so he can focus on what he does best: the technology. And get it to patients as fast as possible.”

Designing for control, not magnets

According to Cartocci, in most academic labs today, microrobots remain largely passive tools — probes or magnetic particles set in motion by very large external electromagnetic coils. “It’s extremely hard to scale,” Cartocci explained, “and it doesn’t give surgeons much confidence when it comes to control.”

Robeauté takes a fundamentally different approach. Its system is built around a tiny, active device composed of two parts: a carrier and an extension.

“The carrier contains our core technology,” Cartocci said.

“It’s what allows the robot to move non-linearly through the brain, while being continuously tracked with ultrasound, so we always know exactly where it is.”

The extension, meanwhile, is what defines the medical task itself. “That’s where you specify the pathology or the intervention,” she explained.

“It could be a biopsy tool, an electrode, a drug-delivery mechanism — whatever needs to be delivered locally.

It’s about the ability to reach multiple sites through non-linear routes and to do so with the right level of accuracy. That fundamentally changes how neurosurgery is performed today.”

The benefits of not being first 

Other companies that have tried to industrialise this academic approach— passive probes moved electromagnetically.

“They started before we did, so they had a first-mover advantage, but they’re struggling now, " shared Cartocci

“They’re not bankrupt, but they’re on the downfall. Our approach is entirely different, and we were always very clear about what we deliberately didn’t go for, because we knew it wasn’t the right solution.”

There are also companies doing microrobotics in vascular environments rather than directly in brain tissue. But none are at the maturity level of Robeauté, particularly in regulatory engagement and strong relationships with surgeons.

Why incremental innovation isn’t enough

Cartocci speaks of the urgency of the kind of medtech her company is developing against companies which add incremental value which are hard to mobilise real change around.

"Being less disruptive means adding even less value.  At Robeauté, we’re clearly in the transformational category. People are dying. 

The unmet need is enormous.

A billion people are affected by brain disease. The drug market alone is around $72 billion. Electrodes are now valued in the trillions if you include healthy patients.

The biopsy market is $4 billion. Brain–computer interfaces are now in the multiple billions. There’s so much that requires precision — having one tool that opens all doors — that unlocks a completely different way of performing brain surgery. We see tremendous excitement from the medical profession.”

Finding the right investors by letting go

In January 2025, Robeauté raised $28 million in funding. Cartocci describes the fundraising experience as “traumatic.”She admits.

“It’s brutal. But at some point, you reach a kind of surrender. You know your company, your value, your pitch — and the more you try to control things, the more frustrating it becomes. When you let go and surf the waves, you find the right people.

I spoke to another founder recently who described it as “conjuring destiny.” That resonated. You prepare, you show up, you’re relentless. You don’t take rejection personally. Everyone comes with their own constraints.”

Her biggest takeaway was learning to enjoy speaking to investors — understanding their language. “That wasn’t always the case.”

Following funding, Robeauté doubled its team, which, according to Cartocci, felt incredible:  “Before, it felt like we were constantly being held back. Now there’s space for iteration, greater experimentation, error, and proper building. 

Europe’s regulatory fragmentation problem

The company opened its US subsidiary, which is critical for go-to-market and clinical trials. The US will be its first market, which   Cartocci attributes primarily to regulations:

“In the US, you co-create your regulatory strategy with the FDA. In Europe, you complete everything, submit the file, and wait a year with no feedback.

For a platform technology like ours — moving from one indication to another — the US model makes a huge difference. Once you have FDA data, it becomes a blueprint for Europe.”

She also highlighted Europe’s fragmentation as problematic — “It would take as much time and money to open France as it would to open the entire US.”

That said, Robeaute is European, and as  Cartocci shared, “we care deeply about bringing this back here. Our investors believe in Europe and don’t want to just flip everything to the US. Healthcare values matter. If we don’t feed technology back into public systems, we push them toward the American model — and that’s a failure.”

For Cartocci, commercial brain microrobots are closer than they’ve ever been.

“We need collective effort — investors, developers, regulators — to make sure these solutions reach the market. We’re also hiring: scientific and technical roles, quality and regulatory roles, US go-to-market roles, and engineering roles. It’s a great time to watch — or join — the company.”

The company's next goal is first-in-human studies by the end of 2026.

Lead image: Robeauté. Photo: uncredited. 

Looking back at the evolution of healthtech, many revolutionary startups have been built, often completely shifting the way clinicians work. We have moved from digital record keeping to telehealth platforms, preventative healthcare devices, and now the latest wave of AI co-pilots. Decades of investing in digitisation have consequently turned healthcare into fertile ground for AI applications and automation.

Despite all that digitisation, the industry is still drowning in bureaucracy. Doctors spend almost as much time on admin as they do with patients. Filling forms, writing notes, juggling schedules, and navigating claims is inefficient, expensive, and a massive drain on the system’s capacity. AI can change that – not by replacing clinicians, but by taking the heavy administrative load off their shoulders.

According to the World Economic Forum, AI has the potential to bridge the gap for the 4.5 billion people who lack access to basic healthcare, and to help address the expected shortage of 11 million healthcare workers by 2030. Yet despite this potential, healthcare remains below average in its adoption of AI compared to other industries. That is especially striking when you consider that healthcare generates an estimated 30% of the world’s data, yet 97% of hospital data goes unused. The time is now to make use of that data by implementing AI and fully leveraging the opportunity.

Why remove admin from healthcare workers?

Healthcare services are, as we know, heavily dependent on a highly skilled workforce. Out of the €1.6 trillion spent on healthcare in Europe each year, roughly 50% goes to salaries. We do not think we can – or should – replace these individuals. They are doing astonishing work, and we are far from leaving our destinies in the hands of AI. But these healthcare professionals should be able to spend less time on admin and more time with patients. That is not just an efficiency play; it is also a staffing play. Burnout and workload are core drivers of shortages, and freeing clinicians from unnecessary admin is one of the most realistic ways to expand care capacity without needing millions of new hires who simply do not exist.

Why now?

Europe spends around €300 billion on healthcare administration every year. For the first time, we have the right technology to tackle this huge opportunity and, importantly, to generate meaningful societal impact at the same time. AI solutions are already automating everyday tasks like appointment scheduling, intake, documentation, and claims, with the potential to cut a meaningful share of healthcare spending. So how can AI transform the administrative workload of clinicians?

There are three areas that I believe have a high potential for AI to disrupt:

1. Patient scheduling: Out of administrative costs in European healthcare, about €90bn is spent on calls, booking and rescheduling appointments – where automated scheduling could make a major difference. Checking staff and patient availability, room capacity, back-and-forth around no-shows, and constant calendar updates is a fiddly task that consumes huge amounts of time.

AI can handle this end-to-end, and we see clear potential for verticalisation. Solutions like Roger specifically address dental practices; Wawa Fertility in the IVF space; and Vocca, which is moving to a cross-speciality offering. As AI voice continues to improve, we expect many more companies to verticalise in this space.

2. Patient record management: Maintaining and updating electronic health records. This market is estimated to account for roughly €65 billion annually across European healthcare practices. We have seen a boom in so-called AI scribe startups that record conversations between patients and doctors and fill out health records automatically, shifting a doctor’s attention away from the keyboard and onto the patient.

Companies like Abridge, Nabla and Tandem Health are early examples, while uncovr and Sonia are younger and, for now, more verticalised solutions. We believe the next wave of companies in this space will go beyond documentation, addressing more complex tasks and managing multimodal, end-to-end workflows for clinicians. The most influential healthcare AI companies will not just transform a single segment; they will reshape multiple interconnected workflows across the broader healthcare ecosystem.

3. Medical billing and claims: A lot of time is spent by healthcare and back-office professionals entering the correct codes. There is an estimated €50 billion to be freed up here, where AI can process large volumes of data quickly, likely with greater precision and fewer errors. Platforms like Nelly and Phare Health are already shaking up the billing process, and this category feels overdue for automation.

Long-term, new start-ups will likely push forward along the lines of a few different themes. Firstly, companies that build solutions on top of data that has historically been unused, like the 97% of hospital data mentioned earlier. Second, end-to-end multimodal technologies that integrate several data types into coherent workflows and deeply embed into clinical workstreams. And thirdly, domain-specific solutions that target particular verticals rather than staying broad and high-level. That is where we expect both the biggest efficiency gains and the deepest, most durable impact.

An evolving ecosystem

There are still real constraints today: nuance, edge cases, and trust in clinical settings cannot be solved overnight. But the technology is catching up fast. If we make healthcare more efficient, the wider world works better. Few markets touch all of us, and as populations age and disease burdens rise, this is one of the most important arenas for AI to deliver impact without increasing cost: a rare and timely proposition for an overextended system.

The post AI for healthcare admin: Meet the startups that are providing the right tech at the right time appeared first on EU-Startups.

Europe’s DeepTech ecosystem is approaching an important crossroads. Across the continent, universities and research institutes continue to generate breakthrough science, supported by one of the world’s strongest pipelines of technical talent. The ingredients for globally relevant DeepTech companies are already present.

Japan’s recent commitment of more than €33 billion to European DeepTech and AI highlights this potential from the outside. It is a reminder that Europe has the scientific strength to shape future industries but has yet to fully prove that this research can consistently translate into scaled companies. Whether the region can build true technological sovereignty will depend on how effectively founders, investors, and institutions adapt to the long horizons and structural demands unique to DeepTech innovation.

The paradox facing European DeepTech

Forecasts suggest that European DeepTech could generate around one trillion dollars in enterprise value by 2030. While DeepTech as a whole is attracting a record share of European funding, early-stage rounds are down by 30% since their peak in 2021.

This also happens to be the stage where founders face some of their toughest hurdles. Long development cycles, highly technical milestones, and limited commercial traction make traditional venture investors hesitant. Many VCs acknowledge the importance of DeepTech but still default to evaluating companies through software-style metrics.

Japan’s current strategy highlights this gap. Its investment vehicles target Europe’s scientific innovation while offering industrial scale, manufacturing experience, and long-term capital. With the right investment structures, Europe can pair its research excellence with pathways that support long-cycle technologies. 

Europe’s structural advantages

Europe enters this moment with real strengths. Technical education is one of them. Across the EU, about one quarter of all Master’s degrees are awarded in STEM subjects. In countries such as Germany, more than one-third of tertiary graduates hold a STEM qualification. These numbers reflect a strong pipeline of technical talent.

Funding support is another advantage. Europe offers substantial grant programmes for early DeepTech work. The European Innovation Council provides grants of up to €2.5 million, along with potential equity financing. The EIC Pre-Accelerator supports smaller DeepTech organisations through grants of €300,000 to €500,000. These resources exist in meaningful volume. The challenge is not their availability, but how effectively founders integrate them into an investment strategy.

What founders can do to succeed

Founders can take several concrete steps to make an early-stage DeepTech fundable.

1. Teams need to speak both technical and commercial languages: When a company frames progress only through research milestones, investors struggle to understand how development connects to market outcomes.
2. Grants should be approached as essential capital: Competitive grant wins validate technical merit and extend runway, and they work best when paired with early venture support. For VCs, non-dilutive funding reduces risk perception and strengthens conviction around long-cycle technologies.
3. Create intermediate milestones that are understandable beyond specialist audiences: These may include paid pilots, environmental tests, regulatory-aligned prototypes, or limited-scope deployments with industry partners. They demonstrate motion, even if they are not final products.
4. Early commercial relationships matter: Many corporates in Europe sponsor pilots or co-development projects. Several major European companies, including Atos and Siemens, already operate innovation programmes that support early technical validation. These partnerships give founders something tangible to show investors.
5. Translate research timelines into business timelines: Industry expects predictable communication and steady progress. The scientific world allows long, exploratory cycles. Commercial environments do not.

Where European VCs can strengthen the early-stage environment

Europe’s venture ecosystem has an opportunity to evolve in parallel with the region’s scientific strengths. Generalist funds already play an important role at the earliest stages, yet specialised DeepTech VCs often benefit from investors who can draw on technical networks, industrial partners, or sector-specific knowledge. Expanding this pool of expertise, especially around seed and Series A, would create clearer roads for founders working on long development timelines.

Patient capital is also part of this shift. Many DeepTech companies progress through stages that resemble life sciences, where well-defined milestones and longer horizons are standard. European funds are already beginning to adopt elements of this approach as hard technology becomes more central to the economy. Life sciences investors in the United States have adapted to similar challenges, and their experience offers a useful reference point for European DeepTech venture capital as it develops its own investment frameworks. 

Europe’s sovereignty opportunity

Europe’s position in DeepTech relates directly to technological sovereignty. Japan’s investment programme shows that global actors see European research as a foundation worth building on at an industrial scale. This external confidence should motivate European institutions and investors to improve the early-stage environment rather than assume that technical excellence will translate into commercial leadership on its own.

The ingredients already exist. Europe has strong universities, a large population of STEM graduates, and grants that actually deploy capital. What is missing is alignment between founders who build long-cycle technologies and investors who can support them with expertise, patient capital, and structured pathways to commercial validation.

Final thoughts

DeepTech momentum in Europe depends on two shifts happening at the same time. Founders can increase their chances of success by building teams that translate science into market progress by using grants strategically and forming early partnerships. Investors can support this transition by expanding their technical networks, adopting long-term horizons, and developing evaluation frameworks suited to DeepTech trajectories. 

If both sides move in this direction, Europe can turn its scientific strengths into companies that scale and play a defining role in the next generation of global technology.

The post Can Europe become the global centre of gravity for DeepTech? appeared first on EU-Startups.

Alvotech, a Reykjavík-based BioTech company specialising in the development and manufacture of biosimilar medicines, has secured a €100 million senior term loan facility to strengthen liquidity and support the execution of its strategic priorities in 2026.

The term loan facility bears an interest rate of 12.50%, payable monthly in cash, and has a maturity date of 2 years. The transaction, led by GoldenTree Asset Management, replaces the company’s previously disclosed working capital facility (ABL) and provides Alvotech with access to the full €100 million for the duration of the loan term. The structure offers enhanced operational flexibility, Alvotech claims. 

“This €100 million financing underscores the long-term commitment of our financing partners at GoldenTree and their alignment with Alvotech’s strategy. Their support strengthens our ability to execute on our growth plans, invest in R&D, and deliver high-quality biosimilars to patients worldwide,” said Robert Wessman, Chairman and CEO of Alvotech.

Founded in 2013 by Wessman, Alvotech aims to be a global leader in the biosimilar domain. “A biosimilar is a biologic medicine that is highly similar to and has no clinically meaningful differences from an existing approved biologic medicine, or reference product. Biosimilars (like reference products) are produced in living systems,” Alvotech explained on its website. 

According to the Icelandic company, five biosimilars are already approved and marketed in multiple global markets, including biosimilars to Humira® (adalimumab), Stelara® (ustekinumab), Simponi® (golimumab), Eylea® (aflibercept) and Prolia®/Xgeva® (denosumab).

Its current development pipeline includes nine disclosed biosimilar candidates aimed at treating autoimmune disorders, eye disorders, osteoporosis, respiratory disease, and cancer.

In December 2025, Alvotech announced the successful placement of €91.9 million ($108 million) senior unsecured convertible bonds due 2030. Additionally, in June 2025, the company announced the repricing of its existing facility to an interest rate of SOFR plus 6.0% per annum, equivalent to approximately 9.8% based on the 30-day average SOFR rate of ~3.8%. In June 2024, it announced the successful arrangement of a strategic refinancing transaction maturing in June 2029, also led by GoldenTree Asset Management. 

Alvotech stated that its R&D pipeline currently includes 30 products in development. The company is expanding its production capacity and strengthening its supply chain to support four global product launches planned through 2026.

The company claims to have established a network of commercial partnerships to support market access in regions including the United States, Europe, Japan, China, other Asian countries, and parts of South America, Africa, and the Middle East.

The post Icelandic BioTech Alvotech secures €100 million term loan facility to bolster liquidity and R&D investment appeared first on EU-Startups.

Pragmatech, a Spanish HealthTech startup focused on AI solutions in pharmacology, microbiology, and infectious diseases, has raised €650k in funding to boost the commercial deployment of iAST®, its antibiotic prescribing software that obtained CE marking in July 2025. 

The round was led by First Drop, which invested €300k, and included Urriellu Ventures with €175k. The round also comprised an ENISA loan of €180k and the conversion into equity of €162,500 in convertible notes from a previous round.

Pragmatech is an Oviedo-based startup founded in 2021. According to the company, its flagship product, iAST®, is the first CE-marked AI-powered software designed to support clinical decision-making in antibiotic prescription. It has been clinically proven to reduce prescribing errors and recommend treatments with lower resistance potential, it claims. 

“With this funding round, we are strengthening our ability to bring iAST® to the hospitals and healthcare professionals who need it most. The support of investors like First Drop and Urriellu Ventures reinforces confidence in our technology and the positive impact it is having on clinical practice,” said Javier Fernández, co-CEO of Pragmatech.

With a B2B model, the company aims to lead the digital transformation of infectious disease management, enhancing patient outcomes and healthcare system sustainability.

Pablo Valledor, co-CEO and CTO, said, “This funding not only enables us to roll out iAST® at a commercial level, but also provides us with the resources necessary to continue refining our products and to ensure that artificial intelligence supports clinical decision-making in a safe and effective manner.”

The post Spanish HealthTech Pragmatech raises €650k to roll out its CE-marked AI antibiotic prescribing software appeared first on EU-Startups.