Artificial intelligence is taking a giant leap forward: whereas previous AI systems mainly responded to queries, so-called AI agents can think, plan, and act independently. A recent study by Lufthansa Industry Solutions (LHIND) shows how agentic AI is taking automation to a whole new level and what benefits companies can gain from it.

From reaction to action: What makes Agentic AI so special

Agentic AI is more than just another AI hype. It is the evolutionary leap from reactive to proactive systems. While traditional AI tools such as ChatGPT wait for input and then respond, AI agents independently develop strategies and perform complex tasks autonomously. Agents “think” (plan), ‘act’ (execute steps), and “learn” (improve through feedback loops). NVIDIA describes Agentic AI as systems with reasonable reasoning and iterative planning that autonomously solve complex, multi-step tasks ranging from customer service to software development.

Julian Staub, Business Director at LHIND, explains:

“A holistic AI strategy helps companies to sensibly distribute responsibilities for development, implementation, and operation.”

Konkrete Anwendung von Agentic AI

Eine Grafik des LHIND-Papiers visualisiert zentrale Bausteine eines Agenten-Ökosystems – von Autonomie, adaptivem Lernen und Sprachverständnis bis hin zu Multi-Agent-Konversation und Workflow-Optimierung – und macht deutlich, dass Agentic AI über reine Textgenerierung weit hinausgeht.

The white paper lists specific use cases across the value chain: customer service (responding to bookings, cancellations, inquiries), sales (lead qualification, CRM maintenance), coding assistance (writing code, finding errors, searching documentation), data analytics (queries, analyses, generating dashboards), and IT ops/incident response (fault analysis, self-healing tendencies). How can companies benefit from this in concrete terms?

• Automation of variable processes (e.g., cases with many exceptions) instead of just rigid workflows.
• 24/7 scaling while reducing the burden of routine tasks—a lever against the shortage of skilled workers.
• Democratization of expert knowledge: Agents incorporate rules, playbooks, and policy knowledge.
• Continuous improvement via feedback/telemetry.

A tangible example from the Lufthansa study: An AI agent for travel bookings receives the instruction “I have to travel to London on business from July 14 to 17. Please find an inexpensive flight and comfortable, reasonably priced accommodation.” The agent not only searches for suitable options, but also weighs up various alternatives, makes independent decisions, and completes all bookings without further human intervention.

Sascha Poggemann, co-founder and COO of Cognigy, emphasizes:

“It is important to note that AI will not replace personal relationships with customers or between employees, but will instead support them in a different way.”

Agentic AI from the consumer’s perspective

In addition to the corporate perspective, agentic AI is also increasingly changing the everyday lives of consumers. This opens up new potential in terms of convenience, while at the same time raising the bar in terms of trust, transparency, and control. Recent studies, including the Zendesk CX Report 2025 and a survey by Bain & Company on agentic AI, paint a nuanced picture of attitudes toward autonomous AI agents.

In the future, digital assistants could not only manage appointments or suggest purchases, but also independently book flights, compare insurance policies, or control health apps that recommend preventive measures. A tangible example from everyday life: when the household is running low on laundry detergent, an agent recognizes this, compares offers, uses loyalty programs, orders the product, and pays for it—without human intervention. Interaction with technology is shifting from passive use to active delegation.

Trust, error rate, and the role of empathy

The integration of agentic AI into the purchasing process is currently met with a mixture of curiosity and reluctance. According to a survey of 2,000 Americans conducted by Bain & Company, 72% have already used AI tools, but only 10% have actually made purchases using such systems. Most of these purchases were low-priced items such as groceries or household goods.

A key obstacle is trust: only 24% feel comfortable today with AI agents making completely autonomous purchasing decisions. The biggest concerns relate to the security of payment information, the protection of personal data, possible misinterpretations of individual needs, and the risk that fraud attempts will go undetected.

The following features can increase trust in Agentic AI:

1. Ensuring data security and privacy: 39% feel more comfortable when AI is linked to a trusted payment provider such as Apple Pay, PayPal, or Google Pay.
2. Systems behave like humans: 64% of consumers tend to trust AI agents more when they are empathetic (Zendesk 2024).

Potential and limitations

Despite existing concerns, there is a high level of interest: according to the Bain & Company survey, 64% have already used AI for transactions or could imagine doing so. The Zendesk Report also shows a clear trend: 53% of respondents say they will prefer interacting with AI-powered agents in the coming years. The reason for this is that AI is less prone to errors than humans.

Systems such as OpenAI’s Agent or Google’s AI Mode are already demonstrating how consumers will find, compare, and purchase products in the future without having to navigate through stores themselves. Agent-based commerce has great potential, especially for purchases with low emotional attachment and clear objectives. However, when it comes to purchases involving personal style, higher costs, or emotional value, AI faces three hurdles: trust, taste, and context.

Do not underestimate challenges and risks

Despite all the potential, experts warn against exaggerated expectations. IBM experts emphasize the need to have realistic discussions about AI agents and to break through the hype. The following aspects are particularly critical for debt collection companies:

Regulatory compliance: Agentic AI must comply with strict data protection and debt collection regulations. The autonomous decisions made by the systems must be traceable and legally sound at all times.

Menschliche Kontrolle: Trotz Autonomie müssen wichtige Entscheidungen, etwa bei komplexen Verhandlungen oder rechtlichen Fragen, weiterhin unter menschlicher Kontrolle bleiben.

Implementation costs: Building an agentic AI infrastructure requires significant investment in technology, data quality, and staff training.

For the successful introduction of Agentic AI, the Lufthansa study has therefore defined ten critical success factors:

1. Clear goal definition: Precise task definition for focused systems
2. Security right from the start: Cybersecurity as an integral part
3. Transparency and traceability: Decision-making processes must be documentable.
4. Ethical guidelines: Fairness and data protection as fundamental principles
5. Focus on functionality: practicality over technical perfection
6. Clear governance: Define clear responsibilities
7. Data quality: Clean, up-to-date, and structured data
8. User-centered design: Focus on intuitive usability
9. System integration: Compatibility with existing systems
10. Pilot projects: Gradual introduction and optimization

It’s time to let AI take action

Current market forecasts underscore the relevance of agentic AI: Deloitte predicts that 25 percent of companies using generative AI will launch agentic AI pilot projects as early as 2025 – and that figure is expected to rise to 50 percent by 2027. Gartner predicts that by 2029, AI agents will independently solve 80 percent of common customer service issues, resulting in a 30 percent reduction in operating costs.

Agentic AI marks the transition from assisting to executing. Those who now set up competencies, data access, and governance cleanly can automate complex, varied processes, reduce costs, increase quality, and assign employees to more complex tasks in a targeted manner. Or, as the LHIND paper summarizes: Strategy, infrastructure, governance, and use case discipline are the levers that enable agents to operate securely, efficiently, and scalably.

“AI technology has left the testing phase and is now ready for mass adoption … The goal must be to make interactions with AI agents intuitive, helpful, natural, and, ideally, entertaining and enriching,” says Dr. Lars Schwabe, CTO at LHIND.

Cover image: © Wanan

Sources:

Bain & Company.Bain & Company. (2024). Agentic AI Commerce Hinges on Consumer Trust. Available at: https://www.bain.com/insights/agentic-ai-commerce-hinges-on-consumer-trust/ [As of September 2025].

Gartner, Inc. (2025, 5. März). Gartner Predicts Agentic AI Will Autonomously Resolve 80% of Common Customer Service Issues Without Human Intervention by 2029. Press release. Stamford, CT. Available at: https://www.gartner.com/en/newsroom/press-releases/2025-03-05-gartner-predicts-agentic-ai-will-autonomously-resolve-80-percent-of-common-customer-service-issues-without-human-intervention-by-20290 [As of September 2025].

Lufthansa Industry Solutions GmbH. (2025). AGENTIC AI: The next level of automation. White paper. Norderstedt: LHIND Marketing & Communications. Available at: https://www.lufthansa-industry-solutions.com/de-de/studien/whitepaper-agentic-ai-2025 [As of September 2025].

NVIDIA Corporation. (2024). What Is Agentic AI? NVIDIA Developer Blog. Available at: https://blogs.nvidia.com/blog/what-is-agentic-ai/. [As of September 2025]

Zendesk. (2024). How AI Adoption in CX is Driving Revenue Gains. Available at: https://technologymagazine.com/articles/zendesk-how-ai-adoption-in-cx-is-driving-revenue-gains [As of September 2025].

Zendesk. (2025). CX Trends 2025. Surge ahead with human-centric AI. Available at: https://cxtrends.zendesk.com/?_gl=11nedhbl_gcl_auMTM0NDUwNDY1My4xNzU3OTM1NTgxLjc4ODEzNTU0MS4xNzU3OTM1NjA2LjE3NTc5MzU2MjI._gaNjM3MTkzMTE0LjE3NTc5MzU2MDQ._ga_FBP7C61M6Z*czE3NTg2Mzc2NDIkbzMkZzEkdDE3NTg2Mzc4OTgkajQ0JGwwJGgw [As of September 2025].