AI or Obsolescence: Why Every Business Needs an AI Strategy Now

By Dirk Röthig | CEO, VERDANTIS Impact Capital | 07. March 2026

91 percent of business leaders say AI is business-critical. Yet adoption gaps between the US, China, and Europe are widening every quarter — and the consequences are becoming irreversible.

The Question Is No Longer Whether — It Is How Fast

Three years ago, artificial intelligence was a subject for technology conferences and academic journals. Today it is the central strategic question facing every organisation of every size in every sector. The shift happened faster than almost anyone predicted, and it has left many executives in a genuinely uncomfortable position: they know they need an AI strategy, they are not sure what that strategy should look like, and they are acutely aware that the cost of getting it wrong — either by moving too slowly or by adopting the wrong approach — could be existential.

This is not hyperbole. KPMG's most recent survey of senior business leaders found that 91 percent now describe AI as business-critical for their organisation (KPMG, 2025). Not as a useful enhancement. Not as an optional add-on. As critical. That number, and what it represents in terms of competitive urgency, should concentrate minds.

The challenge is that urgency does not automatically translate into clarity. And in the absence of clarity, many organisations do one of two things: they either delay indefinitely while waiting for the perfect moment — which never arrives — or they invest in AI tools in an undisciplined, unstrategic way that produces costs without proportionate returns. Both responses are understandable. Neither is acceptable.

Dirk Röthig, CEO of VERDANTIS Impact Capital, has observed this dynamic across multiple European markets and consistently emphasises that the strategic work must precede technology selection. "The organisations that are succeeding," Dirk Roethig notes, "are those that have done the unglamorous work of capability assessment and use-case prioritisation before they touch a single AI tool."

A Global Landscape Defined by Divergence

To understand the stakes, it helps to map the global AI landscape with some precision. The picture that emerges is one of sharp divergence — divergence between regions, between sectors, and between companies within the same sector.

The United States leads in AI model development, infrastructure investment, and commercialisation. The major American technology companies — Microsoft, Google, Amazon, Meta, OpenAI — have committed hundreds of billions of dollars to AI capability building over the next several years. The US venture capital ecosystem channels tens of billions annually into AI startups. American research universities, federal agencies like DARPA, and private laboratories are collectively generating a substantial majority of the world's most cited AI research.

China has adopted a different but equally ambitious approach: state-directed AI development, with national champions like Baidu, Alibaba, Tencent, and emerging players like DeepSeek receiving both strategic guidance and financial support from a government that views AI leadership as a geopolitical imperative. China's approach prioritises mass deployment in manufacturing, surveillance, and public services — creating large-scale real-world datasets that are, in their own way, a source of competitive advantage.

Europe — and Germany in particular — occupies a different position. The German economy has 687 AI startups according to the appliedAI initiative's most recent analysis (appliedAI, 2024). The country has world-class engineering universities, a sophisticated industrial base, and a tradition of methodical, high-quality applied research. But it lacks the venture capital depth of the US, the state-directed ambition of China, and — critically — the cultural appetite for the kind of rapid, fail-fast iteration that AI development seems to require at its best.

The result, as documented by Bitkom's 2025 survey, is that only 36 percent of German companies currently use AI in any meaningful way (Bitkom, 2025). That is a remarkable gap given the 91 percent who describe it as business-critical — a tension that Dirk Röthig has explored extensively in his analysis of AI adoption patterns in the German Mittelstand.

What Productivity Data Actually Shows

Arguments about competitive positioning can sometimes feel abstract. The productivity data is not abstract.

PwC's 2025 Global Workforce Hopes and Fears Survey found that in sectors where AI has been substantively adopted, productivity gains average 27 percent (PwC, 2025). To put that in concrete terms: a team of 100 people in an AI-enabled workflow produces the equivalent output of 127 people operating in a conventional workflow. Over a year, compounded across thousands of workers and multiple business units, that differential becomes a structural competitive advantage — one that manifests in lower unit costs, faster product cycles, quicker customer response times, and higher margins.

The talent market tells the same story from a different angle. Workers with demonstrated AI competence now command salaries 56 percent above those of comparably qualified workers without such competence (PwC, 2025). This premium exists because the market is pricing in the productivity differential directly. An employee who can use AI tools effectively is, quite literally, doing the work of 1.27 employees. The market is beginning to compensate them accordingly.

For employers, this creates an imperative that goes beyond competitive strategy. It becomes a talent retention issue. Workers who develop AI competence within an organisation — or who come to an organisation with AI skills — will increasingly compare their compensation against what AI-competent workers earn elsewhere. Organisations that do not have coherent plans to develop AI competence internally will find themselves in a persistent talent disadvantage.

In conversations with business leaders across Europe, Dirk Roethig frequently encounters the complaint that they cannot afford to lose their technically skilled staff to competitors with more mature AI strategies. "The answer," says Dirk Röthig, "is not to pay more. It is to build an organisational culture where people develop cutting-edge skills and use them in meaningful ways. That is actually what people want — not just higher salaries, but the chance to work on problems that matter using tools that are genuinely at the frontier."

The EU AI Act: Navigating the Regulatory Landscape

Any serious discussion of AI strategy in a European context must address the EU AI Act — the world's first comprehensive regulatory framework for artificial intelligence, which entered its implementation phases in 2024 and 2025.

The AI Act operates on a risk-based principle. AI systems are classified into risk categories — unacceptable risk (prohibited), high risk (strictly regulated), limited risk (transparency requirements), and minimal risk (largely unregulated). The practical implication for most businesses is that the majority of AI applications they are likely to deploy fall into the limited-risk or minimal-risk categories — chatbots, recommendation systems, productivity tools, content generation — and are therefore subject to relatively modest requirements.

The genuinely high-risk applications — AI in credit scoring, AI-driven hiring decisions, AI in safety-critical systems — face much more stringent requirements around transparency, explainability, human oversight, and technical documentation. For companies in financial services, healthcare, or recruitment, these requirements are substantive and require careful compliance planning.

From a strategic perspective, the EU AI Act is probably best understood as a competitive variable that cuts both ways. On one hand, it imposes compliance costs that American and Chinese competitors do not face in their home markets. On the other hand, it creates a regulatory advantage in global markets where AI governance is increasingly demanded — particularly in B2B enterprise contexts, where procurement processes increasingly include AI governance due diligence. A company that can credibly demonstrate EU AI Act compliance is, in many enterprise sales contexts, better positioned than a competitor that cannot.

Dirk Röthig has been vocal in arguing that European companies should view the regulatory clarity of the EU AI Act not as a burden but as a strategic asset. "In uncertain times, regulatory clarity is actually a competitive advantage," Dirk Roethig explains. "Companies that understand the rules can move confidently. Companies in jurisdictions without clear rules are either moving recklessly or not moving at all. That is not a subtle difference."

What a Real AI Strategy Looks Like

Given the pressures from all sides — competitive urgency, talent markets, regulatory requirements — what does a coherent AI strategy actually look like for a company that is not a technology giant?

It does not look like a comprehensive AI transformation programme. It does not start with a platform decision or a thousand-day roadmap. It starts with much more fundamental work.

Step one is honest capability assessment. What data does the organisation have? How well-structured is it? What are the highest-cost, most time-consuming processes? Where are human capabilities being systematically underutilised because people are doing things that machines could do? This step requires genuine intellectual honesty. Many organisations, when asked what data they have, discover through this exercise that they have vastly more data than they realised — and vastly more disorder in how it is managed.

Step two is prioritised use-case selection. Not every process benefits equally from AI. The highest ROI use cases tend to share common characteristics: they are data-rich, they are repetitive and rule-based in significant portions, they are currently resource-intensive, and the quality of the output is measurable. Document processing, customer inquiry handling, inventory forecasting, quality control inspection — these are areas where AI can deliver measurable results within months rather than years. Dirk Röthig emphasises that this prioritisation step is where most strategic value is created. "If you get the use cases right," Dirk Roethig observes, "the execution becomes relatively straightforward. If you get them wrong, you can have the best technology in the world and still fail."

Step three is build-versus-buy discipline. In 2026, the range of off-the-shelf AI tools available is enormous. For the vast majority of use cases, buying or licensing existing solutions is faster, cheaper, and less risky than building from scratch. Custom development should be reserved for genuinely differentiated capabilities — areas where a proprietary AI model would constitute a sustainable competitive advantage. The temptation to build internally is often strong — particularly in engineering-driven cultures — but it is frequently misplaced. Dirk Roethig counsels organisations to ask themselves a hard question: "Is this AI capability something that will define how we compete, or is it something that we need to do our work better? If it is the latter, buy it."

Step four is workforce strategy. The 56-percent salary premium for AI-competent workers (PwC, 2025) is not a reason to hire only externally. It is a reason to invest aggressively in developing AI competence within the existing workforce. Internal AI literacy programmes, access to training resources, and structured time for experimentation are the most sustainable ways to build the competence base an organisation needs. Organisations that have invested heavily in internal AI training programmes are consistently reporting higher adoption rates, better use-case outcomes, and significantly lower staff turnover among technical employees.

Step five is governance. Every organisation deploying AI needs clear policies on data use, model oversight, human review requirements, and error accountability. This is not optional from a regulatory standpoint — and it is a basic requirement for responsible deployment. Governance should not be viewed as a constraint on innovation. Rather, it is what makes sustainable innovation possible. Organisations with clear governance frameworks can move faster, not slower, because they are not constantly renegotiating the boundaries of acceptable risk.

Sectors Under Maximum Pressure

Certain sectors face AI disruption with particular intensity, and their leaders need strategies calibrated to that intensity.

Financial services: AI is already transforming credit assessment, fraud detection, customer onboarding, and investment research. Institutions that are not actively deploying AI in these areas are conceding ground to competitors who are. The EU AI Act's requirements for high-risk AI in credit scoring are real — but they are manageable. The bigger risk is falling behind on capability. Banks and insurance companies that have deployed AI-driven credit models in the past 18 months are already seeing measurable improvements in default prediction accuracy and customer onboarding friction. Those that have not are increasingly difficult to compete with.

Manufacturing: Germany's industrial strength is precisely the domain where AI delivers some of its most dramatic productivity gains. Computer vision for quality control, predictive maintenance, generative design for engineering — these are not future applications. They are deployed today by leading manufacturers. The 27 percent productivity gain figure (PwC, 2025) has already been achieved in leading manufacturing operations. German Mittelstand companies that have integrated AI into their quality control processes are reporting not only reduced defect rates but also substantially faster product iteration cycles. The combination of these two benefits compounds into a significant competitive advantage.

Professional services: Law firms, consulting firms, accounting practices — all face AI disruption with a particular character. AI does not replace the strategic judgment, client relationships, and professional expertise that define the core value proposition. But it dramatically reshapes the economics of the underlying work — research, document analysis, first-draft production, data analysis. Firms that integrate AI into these workflows will be able to deliver more for less, or charge more for genuinely higher-value work. Partners in firms that have embraced AI-enabled workflows are reporting that they have substantially more time for client-facing work and strategic thinking — precisely the work that differentiates high-value professional service providers.

Public sector: The McKinsey analysis of 165,000 potentially AI-replaceable public sector positions in Germany is a signal that even the most traditionally change-resistant sector is not immune (McKinsey, 2024). Government agencies that proactively design AI-enabled workflows will be better positioned to manage administrative efficiency. Dirk Röthig has been particularly engaged with thinking about how public sector institutions can deploy AI responsibly — an area where the governance clarity of the EU AI Act becomes genuinely valuable.

The Cost of Waiting

There is a specific economic logic to AI adoption timing that is worth articulating explicitly. AI systems improve through use. Models trained on more data, deployed in more contexts, and refined through more feedback loops are systematically better than models at an earlier stage. This means that early adopters develop a compounding advantage: they get better results faster, because their systems are learning on more real-world data.

The German companies that adopted AI three years ago are not merely three years ahead of companies that start today. They are structurally ahead — in model quality, in organisational learning, in staff competence, and in customer expectation calibration. That gap does not close simply by adopting the same technology today. It requires years of real-world learning to close.

This is the real cost of waiting. It is not merely a cost measured in missed quarterly productivity gains. It is a cost measured in compounding capability gaps that become progressively harder to close. A manufacturing company that deployed computer vision systems two years ago has already worked through the integration challenges, trained its workforce, and achieved measurable quality improvements. A competitor starting today faces the same learning curve — but faces it from a position of relative competitive disadvantage.

Dirk Röthig frequently emphasises this temporal dimension in his advisory work. "There is no such thing as catching up quickly in AI," says Dirk Roethig. "You can catch up on features. You can catch up on technology. But catching up on organisational learning and capability development takes time. That means the decision to delay is not a decision to delay implementing AI. It is a decision to accept permanent competitive disadvantage."

Conclusion: Strategy Before Technology

The 91 percent who say AI is business-critical are right (KPMG, 2025). But the right response to that recognition is not to purchase the most visible AI tool on the market and declare an AI strategy. It is to do the rigorous strategic work — capability assessment, use-case prioritisation, workforce development, governance — that turns AI from a source of cost and confusion into a source of sustainable competitive advantage.

The global divergence between AI leaders and laggards is real and widening. Europe has the talent, the research infrastructure, and the regulatory clarity to build a distinctive AI competence. What remains is the organisational will to act — quickly, strategically, and with the disciplined focus that has historically been one of European business's greatest strengths.

Dirk Röthig, through VERDANTIS Impact Capital, continues to engage with European organisations grappling with these strategic questions. The conversation with Dirk Roethig consistently returns to a single point: the moment for strategic clarity has arrived. Organisations that use the next six months to build coherent AI strategies will be the organisations that reap the benefits of AI deployment. Those that remain undecided will find that the window of opportunity — while not yet closed — is narrowing every quarter.

The window is open. Not indefinitely.