Mit Hilfe von AI
1. The Threshold of the Geopolitical AI-Technological Era
Artificial intelligence is moving from a tool that generates information to a technology capable of taking action. The emergence of AI agents marks a potentially decisive transition: instead of merely answering a question, an AI system can increasingly interpret objectives, plan a sequence of tasks, use external tools, interact with software and execute actions with limited human intervention.
This changes the strategic equation.
For governments, corporations and defence organisations, technological superiority is increasingly connected to economic resilience, national security, intelligence capabilities, cybersecurity, productivity and geopolitical influence. The competition is no longer simply about who has the best model. It is increasingly about who can build the most capable AI-enabled operating environment.
The global race is therefore moving from artificial intelligence toward agentic intelligence.
Die United Nations’ work on global AI governance already frames AI as a global governance challenge requiring international cooperation. At the same time, the European Union’s AI Act framework demonstrates that regulation itself is becoming a strategic component of technological competition.
AI agents add another layer.
They can potentially operate across enterprise systems, government databases, financial infrastructure, logistics platforms, scientific environments and digital communications. The strategic question consequently becomes much larger than “How powerful is the AI model?”
The more important question is:
What happens when millions of AI systems can act simultaneously across the world’s digital infrastructure?
That is where AI agents become a critical strategic technology.
2. AI Agents: Strategic Significance and How They Work
Traditional software generally follows explicitly defined instructions.
Generative AI changed this model by allowing systems to interpret natural language, generate content, analyse information and assist humans with increasingly complex intellectual tasks.
AI agents represent another evolutionary step.
An AI agent can be designed to perceive an environment, interpret an objective, reason about possible actions, use tools, execute tasks and evaluate results. The European Commission describes AI agents as systems capable of interacting with environments and performing tasks autonomously, while NIST’s recent work focuses specifically on the security and interoperability challenges created by agents capable of autonomous action.
The fundamental architecture can be simplified into five layers:
- Perception – understanding information from data, documents, software or sensors.
- Reasoning – interpreting objectives and determining what should happen next.
- Planning – breaking complex objectives into individual actions.
- Tool use – interacting with APIs, databases, browsers, enterprise software or other systems.
- Execution and feedback – performing actions and evaluating their consequences.
This distinction is strategically important.
A chatbot may tell a manager that a supply chain problem exists.
An agent could potentially identify the problem, investigate the relevant suppliers, compare alternatives, prepare procurement recommendations, communicate with internal systems and initiate approved actions.
The difference is the transition from information production to operational execution.
From single agents to multi-agent systems
The next major development is likely to be the emergence of multi-agent architectures.
Instead of one general-purpose AI attempting to perform everything, organisations can deploy specialised agents.
One agent may conduct market intelligence.
Another may analyse financial data.
A third may monitor cybersecurity.
A fourth may manage logistics.
A fifth may challenge the conclusions of the others.
A supervisory agent can then coordinate the system.
NIST explicitly distinguishes between single-agent and multi-agent architectures, describing multi-agent systems as systems in which several agents coordinate actions to achieve complex objectives.
This could create a new form of digital organisation.
The traditional enterprise has employees, departments, managers and information systems.
The agentic enterprise could increasingly have human executives supervising networks of specialised digital workers.
This does not mean that human workers disappear. Instead, the economic value of human work can shift toward strategy, judgment, accountability, creativity, relationship management and high-level decision-making.
The transformation could be particularly significant because AI agents can operate continuously.
They do not need to work according to conventional office hours.
They can monitor markets overnight, analyse cybersecurity events continuously, track competitors, process documents and prepare recommendations before human decision-makers begin their day.
The result is potentially an always-on organisational intelligence layer.
Stanford’s AI Index has already documented rapid progress in agent capabilities. Its 2026 findings indicate that AI agents improved dramatically on computer-use benchmarks during 2025, although they still fail on a significant proportion of structured tasks.
That limitation is crucial.
The agentic revolution is real, but it is not equivalent to reliable autonomous intelligence.
The strategic winners will therefore not simply deploy the most autonomous systems.
They will build the strongest combination of:
AI capability + human oversight + cybersecurity + data governance + institutional knowledge + execution infrastructure.
3. Global Competition, Actors and Risks in AI Agentic Systems
The emergence of AI agents is creating a new layer of technological competition.
The previous AI race focused heavily on models, chips, computing infrastructure and data.
The next stage increasingly concerns autonomous digital capability.
The question is no longer only who can build the most powerful AI.
It is increasingly:
Who can integrate AI into the largest number of economically and strategically important processes?
A. China and the Emerging Asian Power Centres
China enters the agentic era with several structural advantages.
These include a huge domestic technology market, significant engineering capacity, large-scale digital platforms, extensive manufacturing infrastructure and major investments in artificial intelligence.
China’s strategic objective is not simply to develop individual AI products.
The broader ambition is to integrate AI into industrial production, logistics, financial services, healthcare, public administration, robotics and national infrastructure.
This makes AI agents particularly important.
A country with sophisticated agentic systems could potentially automate significant portions of economic coordination.
East Asia adds further technological strength.
Japan possesses advanced robotics and industrial automation capabilities.
South Korea combines semiconductor expertise, telecommunications infrastructure and major technology companies.
Taiwan remains strategically critical because of its semiconductor ecosystem.
Singapore combines digital government capabilities with a highly connected economy.
Together, these ecosystems create an important foundation for the next generation of autonomous AI.
India represents another critical force.
Its enormous technology workforce, expanding startup ecosystem and software capabilities provide significant potential for AI-agent development.
India could become one of the world’s most important implementation markets for agentic AI because many AI systems require integration expertise rather than only frontier-model research.
Australia also has strategic relevance through its research institutions, defence relationships and position within the Indo-Pacific technological ecosystem.
B. Russia: A Defence-Centred Path
Russia’s AI development follows a different trajectory.
Military applications, cybersecurity, intelligence and autonomous systems represent strategically important areas.
AI agents could potentially become useful for intelligence analysis, cyber defence, logistics optimisation and military planning.
However, sophisticated agentic systems require broad access to advanced computing, software ecosystems, data and commercial technology.
This creates structural constraints.
The most important question for Russia may therefore be whether it can build sufficiently autonomous systems while maintaining access to the infrastructure required to develop and operate them.
C. Europe: Regulation, Sovereignty and Strategic Autonomy
Europe’s position is fundamentally different.
The European Union has attempted to combine technological development with human rights, safety, accountability and regulatory oversight.
The AI Act became broadly applicable on 2 August 2026, while several provisions have different transition periods. The European Commission also began enforcing the AI Act from August 2026, including new transparency requirements concerning AI-generated and manipulated content.
This creates both a constraint and an opportunity.
European organisations face additional compliance requirements.
But regulation can also become strategic infrastructure.
The EU’s 2026 work on agentic AI explicitly identifies agentic systems as a major transition from isolated automation toward coordinated, goal-driven systems capable of planning and acting across tools and data sources.
Europe’s competitive opportunity therefore lies in developing trusted agentic infrastructure.
Germany can contribute industrial capability.
France can contribute state capacity, research and strategic autonomy.
The Netherlands has major semiconductor and technology capabilities.
Ireland is a significant technology and multinational hub.
Sweden and the wider Nordic region possess strong digital infrastructure and innovation ecosystems.
The United Kingdom has pursued its own AI strategy, with its AI Opportunities Action Plan focusing on infrastructure, adoption and domestic AI capability. A January 2026 government update reported that 38 of the plan’s 50 actions had been completed.
Switzerland and Norway add further research, finance, energy and technology capabilities.
Hungary’s opportunity is different.
It does not need to compete with the United States or China on raw model scale.
A more realistic strategy is to position Hungary as a regional AI implementation, education, strategic consulting and agentic-AI integration hub.
That strategy could connect Central European companies and institutions with global AI infrastructure.
D. Africa: The Emerging Adoption Frontier
Africa represents a fundamentally different opportunity.
Many economies do not possess the infrastructure required to compete in frontier AI research.
But agentic systems may reduce some barriers.
An AI agent can potentially provide sophisticated administrative, educational, financial or advisory capabilities without requiring every organisation to build a large internal workforce.
This creates an unusual possibility:
countries that are late to traditional digital infrastructure could potentially leapfrog parts of the conventional development model.
South Africa, Nigeria, Kenya, Egypt and Rwanda are among the African countries with significant technology and innovation ecosystems.
However, the digital divide remains a major constraint.
Without reliable connectivity, computing infrastructure, education and data governance, agentic AI could increase rather than reduce inequality.
E. The Americas: Innovation and Strategic Power
The United States remains the most important centre of frontier AI investment and technological innovation.
Stanford’s AI Index reported that U.S. private AI investment reached $109.1 billion in 2024, dramatically exceeding China’s $9.3 billion and the UK’s $4.5 billion.
The United States also possesses an unusually powerful combination of:
- frontier AI laboratories,
- hyperscale cloud infrastructure,
- semiconductor companies,
- venture capital,
- universities,
- defence technology,
- software ecosystems,
- global technology platforms.
This creates a formidable foundation for agentic AI.
Canada contributes strong AI research and a mature research ecosystem.
Mexico could benefit from the integration of AI into manufacturing and North American supply chains.
Brazil represents the largest potential AI market in Latin America and could become a major deployment environment.
The strategic competition in the Americas will therefore not be limited to model development.
It will increasingly involve who controls the platforms through which AI agents operate.
F. Alliances and Technological Blocs
AI agents could accelerate technological bloc formation.
Traditional military alliances remain important.
But technology alliances may become equally consequential.
Countries will increasingly need partnerships around:
- cloud infrastructure,
- semiconductor supply,
- AI models,
- cybersecurity,
- data centres,
- standards,
- digital identity,
- agent protocols,
- AI safety,
- quantum computing.
This could produce a new geopolitical architecture.
Instead of simply asking whether a country is aligned with Washington, Beijing, Brussels or Moscow, future strategic analysis may need to ask:
Which AI ecosystem does the country depend on?
G. The Dark Side: Agentic AI Risks
The greatest danger is not necessarily that AI agents become intelligent.
It is that they become capable before they become reliable.
An autonomous system connected to financial systems, enterprise databases or critical infrastructure can create damage at machine speed.
NIST has specifically highlighted the emerging security risks of AI agents, including vulnerabilities in which malicious instructions embedded in external data can manipulate an agent’s behaviour.
This introduces a new cybersecurity category:
agent hijacking.
A conventional cyberattack attempts to compromise software.
An agent attack may attempt to manipulate the agent’s interpretation of reality.
That distinction is profound.
An attacker may not need to compromise the entire system.
They may only need to influence the information an agent receives.
The consequences could include:
- unauthorised transactions,
- manipulation of business decisions,
- disclosure of confidential information,
- fraudulent communications,
- compromised supply chains,
- automated misinformation,
- cascading system failures.
NIST’s 2026 analysis of AI-agent security responses concluded that agents create novel security threats and that conventional cybersecurity principles will need adaptation to address them adequately.
This is why agent security must become a board-level issue.
4. Strategic Trends: The Redistribution of Power Through AI Agents
The next strategic divide may be between organisations that use AI as a productivity tool and organisations that redesign themselves around AI agents.
That is a fundamental distinction.
A company that gives employees an AI chatbot has improved an existing workflow.
A company that redesigns procurement, research, customer service, cybersecurity and management around agentic systems has created a different operating model.
Closed Versus Open Systems
Open models and protocols can accelerate innovation.
Closed systems can provide stronger commercial control, security and monetisation.
The strategic battle will therefore involve both approaches.
Open ecosystems can create enormous developer communities.
Closed ecosystems can capture disproportionate economic value.
The likely future is not a simple victory of one model.
It is an ecosystem competition.
The Battle for Standards
Standards may become one of the most important geopolitical assets of the agentic era.
If agents from different companies cannot communicate securely, the market becomes fragmented.
If they can communicate through interoperable protocols, an agent economy can emerge.
This is why NIST’s 2026 AI Agent Standards Initiative is strategically significant. Its stated objectives include interoperability, secure agent adoption and industry-led standards.
The countries and companies that influence these standards could shape the future market.
Standards therefore become more than technical documents.
They become economic infrastructure.
5. Industrial and Labour-Market Effects: AI Agents as a Production Revolution
AI agents could transform the structure of work more profoundly than conventional automation.
Traditional automation primarily replaces repetitive physical or administrative tasks.
Agentic AI can increasingly automate cognitive workflows.
Examples include:
- market research,
- financial analysis,
- legal document review,
- software development,
- customer service,
- procurement,
- compliance,
- sales operations,
- intelligence analysis,
- scientific research.
This does not mean every profession disappears.
Instead, professions may become increasingly divided into tasks.
An analyst may spend less time collecting information and more time interpreting strategic consequences.
A lawyer may spend less time searching documents and more time advising clients.
An engineer may spend less time writing routine code and more time designing systems.
A government official may spend less time preparing reports and more time making policy decisions.
This creates a new labour-market premium:
the ability to manage intelligent systems.
The emerging strategic roles could include:
- AI Agent Architect,
- Agent Operations Manager,
- AI Governance Officer,
- AI Security Strategist,
- Human-Agent Team Designer,
- AI Risk Auditor,
- Multi-Agent Systems Engineer,
- AI Transformation Strategist.
The 2026 Stanford AI Index indicates that AI adoption continued to rise in 2025, with 88% of surveyed organisations reporting AI use, although agent deployment remained in the single digits across most business functions.
This creates a critical window.
The technology is advancing faster than organisational transformation.
The companies that redesign workflows early could therefore accumulate a substantial advantage.
6. Ethical, Legal and Social Dimensions of AI Agents
Agentic AI creates a more difficult ethical question than conventional generative AI:
Who is responsible when an autonomous system acts?
If an AI agent sends an email, purchases something, changes a database or makes a recommendation, responsibility cannot simply be assigned to the algorithm.
There must be an accountable human or organisation behind the system.
UNESCO’s Recommendation on the Ethics of AI emphasises human rights, privacy, transparency, accountability, safety, sustainability and human oversight.
These principles become particularly important when AI systems gain the ability to act.
The European AI Act also introduces transparency requirements for AI systems and AI-generated content. As of August 2026, certain AI-generated or manipulated content must be disclosed or labelled under the applicable transparency framework.
The dual-use problem is equally important.
The same agent architecture that can automate logistics for a corporation could potentially support military logistics.
The same cybersecurity agent that protects a company could theoretically be adapted for offensive operations.
The same intelligence system that detects misinformation could potentially be used for mass surveillance.
Therefore, governments need a capability-based governance model, not merely a product-based model.
The relevant question should not only be:
What is this AI system?
It should also be:
What can this system autonomously do?
7. Business Value and ROI: AI Agents as an Investment Instrument
The business case for AI agents should not begin with technology.
It should begin with economics.
The correct question is:
Which expensive organisational processes can be redesigned through AI agents?
A company could identify its 20 most expensive knowledge-intensive workflows and evaluate each according to:
- labour cost,
- time consumption,
- error rate,
- strategic importance,
- scalability,
- automation potential,
- regulatory risk.
The highest-value opportunities should become agentic-AI pilots.
Consider procurement.
Instead of an employee manually monitoring suppliers, an AI-agent system could potentially monitor prices, identify anomalies, compare suppliers, prepare recommendations and notify procurement managers.
Consider competitive intelligence.
A network of agents could continuously monitor competitors, patents, regulatory changes, scientific publications and market signals.
Consider cybersecurity.
Agents could monitor systems continuously, investigate anomalies and prepare incident-response recommendations.
The objective should not be “implement AI.”
It should be:
reduce cost + increase speed + improve intelligence + reduce risk + increase strategic capacity.
The UK government’s AI Opportunities Action Plan similarly treats AI adoption as a mechanism for economic growth and productivity rather than simply as a technology project.
NIST’s AI Risk Management Framework provides another useful principle: AI deployment should incorporate systematic risk management throughout the lifecycle.
For business leaders, the strategic model can therefore be expressed as:
Agentic AI ROI = Economic gain + strategic capability − implementation cost − risk exposure.
The winners will measure all four variables.
8. Vorhersagen und Szenarien: 2050 und 2100
The future of AI agents cannot be predicted with certainty.
But several scenarios can be constructed.
Scenario One: The Augmented Human Economy
By 2050, humans remain the ultimate decision-makers while AI agents become the dominant infrastructure for knowledge work.
Most professionals work with several specialised agents.
Governments use agents to analyse policy options.
Companies use agent networks to manage operations.
Scientists use multi-agent research systems.
This is the most optimistic scenario.
AI increases human productivity without eliminating human accountability.
Scenario Two: The Agentic Corporate State
In this scenario, organisations become heavily dependent on autonomous systems.
Companies with the best agent infrastructure become dramatically more productive.
Economic concentration increases.
A small number of technology platforms control the infrastructure on which millions of agents operate.
This creates a new form of technological power.
Instead of owning factories, dominant companies increasingly control digital labour infrastructure.
Scenario Three: The Multipolar AI World
By 2050, the world has several competing AI ecosystems.
The United States leads one.
China leads another.
Europe maintains a regulated and sovereign ecosystem.
India becomes an important independent technology centre.
Other states form regional alliances.
Agents from different ecosystems have limited interoperability.
The world becomes technologically multipolar.
Scenario Four: The Post-Human Transition
The most radical scenario emerges between 2050 and 2100.
AI systems become capable of scientific discovery, strategic planning and technological design at scales far beyond individual human capability.
Human civilisation increasingly becomes dependent on machine intelligence.
At this stage, the question is no longer simply whether AI creates jobs.
It becomes:
What is the role of human intelligence in a civilisation where machine intelligence is superior in many domains?
This is where the concept of a post-human era becomes strategically relevant.
The answer is not predetermined.
AI could create unprecedented prosperity.
It could also create unprecedented concentration of power.
The decisive variable may be governance.
9. Executive Guide: A Five-Step AI-Agent Strategic Action Plan
Step 1: Conduct an AI-Agent Capability Audit
Organisations should immediately map their workflows.
Identify where employees spend time collecting information, processing documents, making repetitive decisions and interacting with digital systems.
The objective is to identify the agentic automation frontier.
Do not start with a technology.
Start with a business problem.
Step 2: Build Strategic Partnerships
No serious organisation should attempt to build every component internally.
Strategic partnerships may involve:
- AI companies,
- cloud providers,
- universities,
- cybersecurity firms,
- research institutes,
- government programmes,
- specialist consultants.
The objective is to create an ecosystem rather than a single application.
Step 3: Establish Data, Identity and Governance Infrastructure
This may become the most important step.
Agents require access.
Access creates risk.
Every agent should therefore have clearly defined:
- identity,
- permissions,
- data access,
- tool access,
- spending limits,
- action limits,
- audit trails,
- escalation procedures.
