The adoption of Artificial Intelligence (AI) has accelerated substantially, yet industry gains might not be substantial due to adaptation needs. With this ‘productivity paradox’ and the risk of an AI investment bubble looming, companies around the world must decide whether and how to invest in AI relative to the cost of change – be it financial, infrastructural, or human.
Technopolis Group has been speaking with companies across the EU1 about their AI uptake journey, in terms of opportunities and challenges, and some of the emerging AI technology trends.
Based on interviews and insights, it is apparent that directors, CTOs, change management officers, IT managers, etc., are already implementing the latest AI systems and processes. While it is one thing to know what technology is trending, what makes a company ‘AI ready’?
Deployment realities in the EU
AI adoption today encompasses very different levels of engagement across organisations, from employees independently using ChatGPT or Claude to companies deploying purpose-built AI systems integrated into core business processes. This variation means that no single survey captures the full picture. What is clear, however, is that the use of general-purpose Generative AI (GenAI) tools accelerated sharply in 2023 and 2024.
Globally, McKinsey’s 2025 survey found that 88% of organisations report using AI in at least one business function, up from 78% in 2024. In 2025, 20% of EU enterprises with 10 or more employees Incorporated AI models and processes in their work (an increase in use rate of 6.47 percentage points compared with 2024). In 2025, 27% of EU companies surveyed across all industrial ecosystems adopted AI either by consuming prebuilt AI tools or creating AI systems in-house. It is worth noting that this rate is higher for large businesses, more than 3x the rate of small businesses. This discrepancy can be explained, in part, by the fact that large companies are able to absorb the upfront investment, handle the implementation and data complexity, and capture outsized returns through economies of scale. These survey results also indicate that the most-often adopted type of AI technology across all industries is GenAI and machine learning, followed by chat bots, text responses, and other natural language processing (NLP) forms. Autonomous systems (e.g., autonomous drones, self-driving cars, autonomous warehouse robots), computer vision or predictive analytics have been much less mentioned. A more recent trend of agentic AI is growing quickly. In particular, virtual coworkers are being used to support business operations, such as scheduling, finance, or human resources (e.g., recruitment, onboarding), but are not yet widespread in industrial applications.
AI uptake rates also differ strongly depending on the sector. In the EU, AI adoption in automative, gaming, and manufacturing appear to enhance efficiency, innovation, and competitiveness. In the automotive industry for example, machine learning and computer vision enable autonomous driving, advanced driver-assistance systems, and real-time quality control in manufacturing. Predictive maintenance powered by AI reduces downtime, while digital twin models optimise production and logistics. Automotive firms also leverage AI for supply chain forecasting and energy efficiency, making operations more resilient and sustainable. AI adoption is also high in the professional services and ICT sectors, with more than 60% AI users in 2025, up from less than 50% in 2024. According to the Media Industry Outlook 2025, AI adoption is more pronounced in the video game sector as a driver of realism, personalisation, and adaptive gameplay. However, AI tools and applications are less prominent in the audiovisual and news media sector more broadly.
Some questions about returns remain
When it comes to AI adoption for EU companies, key questions remain: What is the impact? What are the implementation costs? Where is AI delivering the most value for EU companies?
The question of returns on investment deserves particular attention. While the promise of AI-driven productivity gains is well documented, the reality for many companies, particularly those making more substantial AI investments, is that immediate financial returns remain elusive. The above-mentioned McKinsey survey found that only 39% of organisations reported any measurable effect on enterprise-level EBIT from AI in 2025; among those, the majority attribute less than 5% of EBIT to AI use. Only around 6% of surveyed organisations qualify as high performers capturing enterprise-wide value. This suggests that, for the moment, the more serious AI investments remain an investment without immediate gains, and companies must weigh long-term strategic positioning against short-term cost pressures.
On the one hand, industry data suggest that up to 95% of identified use cases are not yet yielding consistent or scalable results, highlighting a significant gap between experimental success and real-world deployment. This challenge is particularly evident in applications requiring high precision, reliability, or regulatory compliance (e.g., pharmaceuticals, medical devices, automotive industry). In these sectors, workflows cannot be automated without strict verification, validation, and quality-control procedures. The cost of errors (product recalls, regulatory sanctions, patient safety incidents) means that companies must invest heavily in testing infrastructure and compliance processes before any AI-driven automation can be deployed at scale. On the other hand, as observed above, companies across professional service industry and advanced manufacturing have claimed productivity gains from AI solutions. While EU-level funding instruments are accelerating AI experimentation, the transition from pilot to scaled deployment remains the principal bottleneck.
Significant deployment challenges remain
Regardless of AI’s potential as a game-changer for business, there are significant obstacles to effective and profitable uptake:
A shortage of financing opportunities, through European instruments and the European Investment Bank, also threatens to limit the uptake of AI. Based on interviews with business representatives and members of European industry, this is a particular challenge for SMEs. Smaller companies do not necessarily have access to capital that larger companies have, let alone the agility to integrate the technology to swiftly launch new processes and products.
Access to finance
A shortage of financing opportunities, through European instruments and the European Investment Bank, also threatens to limit the uptake of AI. Based on interviews with business representatives and members of European industry, this is a particular challenge for SMEs. Smaller companies do not necessarily have access to capital that larger companies have, let alone the agility to integrate the technology to swiftly launch new processes and products.
Cost-effectiveness and reliability
Integrating AI into enterprise applications presents major opportunities, but many organisations struggle to turn experiments into reliable, cost-effective production systems. Even experienced, well-funded teams face challenges with scaling, ensuring model safety, and achieving consistent return on investment.
Limited transparency and explainability of AI systems
Most AI models, especially deep learning systems, are ‘black boxes’ with decision processes that are difficult to interpret. This limited transparency means users and regulators often ‘cannot find out why an AI system has made a decision or prediction’, complicating efforts to ensure fairness or accountability.
Risks related to data and information security
Deploying AI at scale introduces significant data privacy and security challenges. AI systems often ingest and generate sensitive data, creating risk of leaks or misuse if not properly secured.
Compliance cannot appear to be complex
The EU AI Act represents a pioneering effort to establish a risk-based regulatory framework for AI systems, and the EU’s leadership in this area serves European interests by building public trust and setting global standards for responsible AI deployment. At the same time, the Act’s requirements introduce costs and uncertainties that can slow down AI deployment. Recognising this, the European Commission has indicated it will phase the implementation timeline and provide targeted support to protect start-ups and SMEs from disproportionate compliance burdens. Consulted experts, nonetheless, remain wary of the Act’s complexity and its potential effect on product development and uptake timelines.
A shortage of top-tier AI engineers with advanced IT and data analytics expertise
The EU lacks this high-level expertise, particularly in frontier areas such as deep learning, reinforcement learning, foundation models, and physical AI. Another gap is in interdisciplinary expertise that fuses technical skills with legal, ethical, and domain-specific knowledge. Consulted experts have noted several technical execution skills gaps.
The business forecast: 5 deployment trends to monitor
In light of deployment challenges, European companies need to be strategic when navigating future developments. The following trends were identified while implementing a study for the European Commission on the EU’s critical digital capacities deployment beyond 2027.
1. The rise of agentic AI
In the short term, AI adoption is expected to continue to increase as developed solutions improve. Particularly, executives expect an increased usage of agentic AI with a continued growth of GenAI use. Executives also anticipate a rapid increase in the use of agentic AI, systems based on foundation models that can plan, reason, and execute multi-step tasks autonomously across business functions. According to Gartner, 40% of enterprise applications are expected to embed task-specific AI agents by 2026, while McKinsey’s 2025 global survey found that 62% of organisations are at least experimenting with AI agents and 23% are scaling them in at least one function. Early use cases include the automation of customer service operations (e.g. refunds), finance (invoicing, forecasting, expense auditing), and security (anomaly detection, policy enforcement). The transition from single-purpose agents to multi-agent systems, where specialised agents collaborate under central coordination, is expected to accelerate in the following year.
2. From large language models to world models
Looking at general AI trends, a shift is emerging from large language models towards world models. The latter are potentially transformative because they could enable sample-efficient learning (learning from fewer real-world interactions by practicing in imagination), better generalisation to new situations, and more robust planning and reasoning. In practical terms, world models allow an AI system to ‘simulate’ different scenarios internally (for instance., predicting what will happen if a robot takes a particular action, or how a supply chain will respond to a disruption). This is different from relying purely on pattern recognition based on historical data.
3. AI safety and cybersecurity as a deployment imperative
As AI becomes embedded deeper into core business operations, the safety and security dimensions of AI deployment are moving from a compliance afterthought to a strategic priority. In the EU, this trend operates on 2 levels: first, companies must secure their own AI systems against adversarial threats; second, they can use AI to strengthen their cybersecurity posture. Both dimensions will demand increased investment and new capabilities from EU companies in the coming years.
As an example, AI is amplifying the speed, scale, and sophistication of cyberattacks. AI-generated phishing, deepfake impersonation of executives, and automated vulnerability discovery are all growing in prevalence. IBM’s 2026 cybersecurity predictions warn that shadow AI (unapproved AI tools deployed by employees without oversight) will be a leading cause of intellectual property compromises, given AI’s handling of proprietary algorithms and strategic decision-making.
4. Purpose-built AI for industrial processes
Some European companies report that the next phase of AI adoption will be driven by purpose-built solutions, optimised for specific industrial processes and systems, but also capable of learning and retaining knowledge over time.
These targeted, continuously improving systems are likely to generate more sustainable value and higher operational impact than general-purpose tools. This trend is reinforced by the broader movement towards smaller, domain-specific models; rather than deploying one general-purpose model for all tasks, companies are fine-tuning efficient models for particular use cases in legal, manufacturing, healthcare, or supply chain management.
5. AI and robotics integration (physical AI)
Following current technological developments, another trend related to AI uptake in businesses is the integration of AI and robotics, otherwise known as ‘physical AI’. Physical AI refers to AI systems that operate in the physical world through robotic or autonomous hardware, using AI-driven perception, reasoning, and planning to interact with real environments. This is relevant because it extends AI’s impact beyond software and data processing into manufacturing, logistics, healthcare, and other physical domains where the EU has established industrial strengths. In practice, near-term physical AI deployment is likely to take the form of increasingly autonomous collaborative robots (‘cobots’) on factory floors, AI-guided warehouse logistics systems, and predictive maintenance robots, rather than the humanoid robots of popular imagination. That said, the humanoid robotics field is also expanding. For EU companies, physical AI represents both an opportunity and a challenge. The EU’s manufacturing base and robotics expertise provide a strong foundation for further deployment; however, the capital intensity, the need for new interdisciplinary skills at the intersection of AI and mechanical engineering, and the growing importance of the Asian market all point to the need for sustained investment and policy support.
Taking stock
This article explores the critical challenges and emerging technology trends shaping the industrial and competitive playing field in Europe. Drawing on insights from the companies, themselves, there are real concerns about return of investment given the deployment challenges.
Technopolis Group’s foresight analysis on AI deployment points to emerging technology trends with direct implications for European companies. Our work with the European Commission’s Directorate-General for Communications Networks, Content and Technology2 further confirms that while EU-level funding instruments are accelerating AI experimentation, the transition from pilot to scaled deployment remains the principal bottleneck.
Many of the challenges raised in this piece are also pertinent for the public sector, which has a growing demand for AI.
Want to find out Technopolis Group’s approach to developing digital capacity and integrating AI tools in our work?
[1] Disclaimer: While this article focuses on companies based in the European Union, this discussion acknowledges that US, Chinese and international companies may face different challenges.
[1] Two studies: Study on the EU’s digital capacities deployment beyond 2027 and Study on the research, development and innovation of strategic digital technologies beyond 2027.