AI in Business 2026: Between Hype and Real Value Creation
By Dirk Roethig | CEO, VERDANTIS Impact Capital | March 3, 2026
88 percent of companies use AI — somewhere, somehow. But only 6 percent achieve measurable profit from it. The gap between investment and value creation is the biggest corporate risk of the decade.
Tags: Artificial Intelligence, Business, Digital Transformation, SMEs, AI Strategy
The Illusion of Adoption
Few technologies in history have been adopted as broadly and as quickly as artificial intelligence. The numbers are impressive: According to McKinsey's The State of AI 2025 study, 88 percent of all surveyed companies worldwide deploy AI in at least one business area (Chui et al., 2025). In Germany, AI usage has doubled within a single year — from 20 percent to 36 to 41 percent of companies, depending on the survey (Bitkom, 2025).
Global AI spending has grown at a pace that exceeded even the most optimistic forecasts. In 2025, worldwide AI investments totaled approximately 1.5 trillion US dollars. For 2026, analysts project an increase to 2.52 trillion — a 44 percent jump in a single year (Gartner, 2025). AI companies attracted roughly 61 percent of global venture capital volume in 2025 (OECD AI VC Report, 2025).
And yet: behind this facade of universal adoption lies an uncomfortable truth. The overwhelming majority of companies deploying AI achieve no measurable economic benefit from it. The technology is purchased, implemented, and showcased — but it does not deliver.
The 6-Percent Gap: What McKinsey Actually Measured
Perhaps the most sobering number in the entire AI debate comes from McKinsey's 2025 survey: Only 6 percent of surveyed companies report a significant impact on their EBIT — defined as an increase of more than 5 percent in operating profit directly attributable to AI initiatives (Chui et al., 2025).
Six percent. Out of 88 percent that use AI. This means: 93 percent of AI adopters operate in a spectrum between marginal benefit and pure cost center.
The Boston Consulting Group confirms this picture with its own data. In its AI at Work 2025 study, BCG finds that 60 percent of companies that have implemented AI generate no material value from it. The technology exists in pilot projects, proof-of-concepts, and innovation labs — but it does not penetrate core operational processes. Only 5 percent of companies manage to deploy AI at scale in a value-creating manner (BCG, 2025).
This discrepancy is no coincidence. It is the result of structural errors in the way companies conceive, implement, and scale AI projects.
Why the Majority Fails: Four Structural Errors
1. Technology Without Problem Definition
The most common mistake is also the most fundamental: Companies implement AI because they want to implement AI — not because they need to solve a concrete problem. They start with the solution and then search for a matching problem.
As I discussed in my earlier article AI in Business: Why German Companies Must Act Now, the decisive first step is not technology selection but identifying the bottlenecks that are actually slowing down the business.
A mid-sized manufacturing company that implements an LLM for customer service when its real bottleneck lies in production planning wastes resources. The AI may work flawlessly from a technical standpoint — but it solves the wrong problem.
2. The Pilot Trap: Brilliant Demos, Missing Scale
Many companies remain stuck in pilot mode. They develop impressive prototypes that work in controlled environments — and then fail at integration into real infrastructure. Legacy systems, data silos, missing APIs, and poor data quality turn the transition from proof-of-concept to production into a nightmare.
BCG identifies this scaling gap as the primary reason why 60 percent of AI projects generate no material value. The technology does not fail on its own — it fails due to the organizational inability to integrate it into existing value chains (BCG, 2025).
3. Data Quality: The Unresolved Foundational Problem
AI is only as good as the data it works with. And data quality in most companies is catastrophic. Redundant datasets, inconsistent formats, outdated information, missing metadata — the list is long. Gartner estimates that poor data quality costs companies an average of 12.9 million US dollars per year — and that is without AI. With AI, the problem multiplies because flawed data leads to flawed models and thus to flawed decisions (Gartner, 2025).
4. Lack of Anchoring at the Executive Level
AI projects driven by IT departments without strategic anchoring at the C-level are doomed to fail. The 6 percent of companies that achieve measurable EBIT impact share one characteristic: AI is a top management priority. It is not treated as an IT project but as a strategic transformation.
What the Winners Do Differently
The data does not only show what goes wrong. It also shows what works — and the results are remarkable.
Productivity Gains: Factor 4.8
Companies that have successfully integrated AI into their core processes report productivity gains between 26 and 55 percent — depending on industry, use case, and implementation depth. Even more impressive: industries that adopted AI early and consistently grow in productivity 4.8 times faster than the industry average (McKinsey Global Institute, 2025).
ROI: $3.70 per Dollar Invested
The average return per AI dollar invested stands at 3.70 US dollars — but this number is an average that masks extreme variance. Among top performers, the ROI exceeds 10:1. Among the bottom 60 percent, it is negative. The distribution resembles a Pareto curve: A small group of winners achieves disproportionate returns while the majority incurs losses.
The Pattern of Winners
When overlaying the McKinsey and BCG data, a clear pattern emerges. The successful 5 to 6 percent of companies share five characteristics:
- Clear Problem Definition: They start with a concrete business problem, not with a technology.
- Data as an Asset: They invested in data quality before AI implementation.
- Cross-Functional Teams: AI projects are not developed in IT silos but by mixed teams of domain experts, data scientists, and business leaders.
- Scaling from Day One: They plan for production deployment from the first day — not as an afterthought.
- CEO Sponsorship: The transformation is driven by executive leadership and tied to clear KPIs.
German SMEs: Between Opportunity and Paralysis
The situation of German small and medium-sized enterprises is particularly instructive. According to Bitkom, AI adoption in Germany rose from 20 to 36 to 41 percent within a single year — a doubling that shows the urgency has been recognized (Bitkom, 2025). At the same time, Germany still lags significantly behind the USA, China, and even Scandinavian countries.
The problem is not a lack of interest. The problem is a lack of implementation capability. The Mittelstand lacks three things:
First: Skilled Workers. As I outlined in my article 20 Million Retirees, 7.5 Million Missing Workers, Germany faces a demographic upheaval that will massively widen the skills gap in the coming years. AI could close this gap — but only if implemented correctly.
Second: Data Infrastructure. Many mid-sized companies still operate with ERP systems from the 2000s, fragmented spreadsheets, and manual processes. Before AI can create value, the data foundation must be digitized, standardized, and made accessible.
Third: Strategic Clarity. There is a lack of clear understanding of which business processes can be improved by AI and which cannot. Not every problem is an AI problem. Sometimes a simple process optimization is more effective than a machine learning model.
The 2.52-Trillion-Dollar Question: Where Does the Money Go?
The projected global AI spending of 2.52 trillion US dollars in 2026 is distributed across four main categories:
Infrastructure (approx. 40%): GPU clusters, cloud computing, data centers. This primarily benefits the major hyperscalers — Microsoft, Google, Amazon, Oracle — as well as chipmakers like NVIDIA.
Software and Platforms (approx. 25%): Enterprise AI platforms, SaaS solutions with AI integration, developer tools. This segment is growing fastest because it lowers the entry barrier for companies.
Consulting and Implementation (approx. 20%): The major consulting firms — McKinsey, BCG, Accenture, Deloitte — have massively expanded their AI practices. McKinsey alone hired over 6,000 AI consultants in 2025.
Research and Development (approx. 15%): Fundamental research, new model architectures, industry-specific models. This is where the largest share of venture capital concentrates.
The question is not whether these investments are justified. The question is whether they reach the right places. If 61 percent of global VC volume flows into AI companies but only 6 percent of users achieve measurable profit, there is a massive allocation problem (OECD, 2025).
What Must Be Done Now: A Five-Point Plan for SMEs
Any mid-sized company in Germany wanting to deploy AI sensibly should follow five steps:
Step 1: Identify Bottlenecks, Not Chase Trends
Before evaluating a single AI tool, identify the three to five biggest bottlenecks in your value chain. Where are you losing time? Where do errors occur? Where are you lacking capacity? AI is a tool for eliminating bottlenecks — not for following trends.
Step 2: Data Quality Before Model Quality
Invest first in data cleansing, data standardization, and data integration. A simple model on clean data beats a complex model on dirty data — every time.
Step 3: Start Small, Scale Fast
Begin with a clearly defined use case that promises measurable economic value. Set clear KPIs. Measure after three months. And then scale — or stop.
Step 4: Bring People Along
The biggest hurdle in AI implementation is not the technology but the organization. Train your employees. Explain why AI is being deployed. Show that AI makes work easier, not redundant. As I discussed in my article AI or Obsolescence: Why Every Business Needs an AI Strategy Now, the question is no longer whether companies need AI, but how quickly they can develop a working strategy.
Step 5: Measure Outcomes, Not Activities
The most common mistake in AI projects: companies measure how many models were trained, how many tools were implemented, and how many pilots were launched. But not whether operating profit improved. Measure outcomes, not activities. Measure EBIT impact, not model accuracy.
Looking Ahead: 2026 as a Turning Point
The year 2026 could prove to be a turning point in the history of corporate AI adoption. Not because the technology achieves a breakthrough — it already has. But because companies are beginning to ask the right questions.
The first phase of the AI revolution (2022-2025) was marked by wonder, experimentation, and undisciplined investing. The second phase, now beginning, will be defined by consolidation, scaling, and the relentless question of return on investment.
At VERDANTIS Impact Capital, we have used AI-powered analysis since inception as an integral part of our investment process — not as an experiment, but as an operational tool for evaluating impact investments. The technology helps us process data volumes that would be unmanageable manually and identify patterns that human analysts miss. But it does not replace human judgment on the strategic and ethical dimensions of an investment decision. Dirk Roethig understands AI at VERDANTIS as an amplifier of human competence — not as its replacement.
The companies that understand this distinction will belong to the 6 percent that create measurable value. The rest will continue spending money and wondering why it does not work.
The answer is simple: AI is not a magic wand. It is a tool. And like any tool, it only unfolds its value in the hands of those who know what they want to build with it.
References
- Bitkom (2025). KI-Monitor 2025: Artificial Intelligence in the German Economy. Federal Association for Information Technology, Telecommunications and New Media.
- Boston Consulting Group (2025). AI at Work 2025: Friend and Foe. BCG Henderson Institute.
- Chui, M., Hazan, E., Roberts, R. et al. (2025). The State of AI in 2025: How Organizations Are Rewiring to Capture Value. McKinsey Global Institute.
- Gartner (2025). Forecast: Artificial Intelligence Software, Worldwide, 2022-2028. Gartner Research.
- McKinsey Global Institute (2025). The Economic Potential of Generative AI: The Next Productivity Frontier. McKinsey & Company.
- OECD (2025). Venture Capital Investments in AI: Trends and Policy Implications. OECD Science, Technology and Innovation Papers.
Further Reading
- AI in Business: Why German Companies Must Act Now
- AI or Obsolescence: Why Every Business Needs an AI Strategy Now
- 20 Million Retirees, 7.5 Million Missing Workers: Can AI Close the Gap?
About the Author
Dirk Roethig is CEO of VERDANTIS Impact Capital, an investment firm specializing in sustainable investments headquartered in Switzerland. He brings more than 20 years of experience in corporate leadership, strategic consulting, and sustainable capital allocation. His focus areas lie at the intersection of technology, business, and societal impact. He writes regularly about AI strategy, economic trends, and sustainable value creation.
Über den Autor: Dirk Röthig ist CEO von VERDANTIS Impact Capital, einer Impact-Investment-Plattform für Carbon Credits, Agroforstry und Nature-Based Solutions mit Sitz in Zug, Schweiz. Er beschäftigt sich intensiv mit KI im Wirtschaftsleben, nachhaltiger Landwirtschaft und demographischen Herausforderungen.
Kontakt und weitere Artikel: verdantiscapital.com | LinkedIn
Lesen Sie diesen Artikel auch auf: Hashnode | WordPress | Tumblr | Blogger | Telegraph
Top comments (0)