AI Agents Take Over Complex Workflows — But One in Five German Firms Sees Staff as Easily Replaceable
Veröffentlicht: 29.06.2026 um 10:56 Uhr, Redaktion boerse-global.de
A new vocational course in central Switzerland is teaching marketing professionals how to deploy AI agents for sales tasks while keeping data protection front of mind. The program, launched today at KV Luzern, reflects a growing international push to train workers alongside autonomous software — even as fresh data shows that many employers see the technology as a direct threat to human jobs.
The shift toward agentic AI has accelerated fast. According to an OpenAI study published June 26, 85 percent of all generated output tokens now come from AI agents. Seven in ten users hand off jobs that take more than an hour to complete. The company's new GPT-5.6 family — built around models named Sol, Terra and Luna — offers dedicated reasoning modes for different tasks. Sol scored 91.91 percent on a terminal benchmark.
Large consultancies and financial institutions are moving quickly. KPMG is rolling out Microsoft Agent 365 across its 276,000 employees. Deloitte has embedded AI agents into its Omnia audit platform for 85,000 auditors. Stripe reports a 26 percent cut in compliance processing time using agents on AWS Bedrock. In Asia, the Shinhan Financial Group kicked off its SCoRE AI project today.
Yet the workforce picture is more nuanced than a simple automation scare. A monitoring program that started in California on June 27 found no evidence of mass layoffs driven by AI as of spring 2026. Still, a study by the Ifo Institute shows that 19.2 percent of German companies consider their skilled workers potentially easy to replace with AI. University graduates and employees aged 25 to 35 are seen as the most vulnerable.
The Raise Us initiative, founded by former U.S. Commerce Secretary Gina Raimondo, aims to counter that trend. It plans to raise one billion U.S. dollars for retraining programs; half a billion has already been pledged by Amazon, Anthropic, Microsoft and OpenAI.
Markus Behrens of Germany’s Federal Employment Agency stresses that IT professions still have a future despite the technology. The agency itself already uses 23 AI tools in its daily operations.
On the performance side, AI agents are being stress-tested in simulations. Princeton University’s CEO-Bench ran agents for 500 days as startup founders. Models such as Claude Fable 5 and GPT-5.5 generated revenues in the tens of millions; the strongest topped 47 million U.S. dollars. Google yesterday reported cost savings of up to 60 percent from its Gemini 3.1 Flash-Lite model.
But experts also flag operational dangers. Without clear accountability, AI agents could erode governance in small and medium-sized enterprises. A Google employee noted that while agents handle feature requests well, they stumble on complex summaries or deciphering abbreviations.
Nations are taking divergent approaches to integration. Estonia plans to issue digital identities for AI agents. Austria is pushing for agentic companions that process data locally and has founded centers for digital humanism. Switzerland, with its new course at KV Luzern, is betting on practical, privacy-conscious adoption.
