The GenAI paradox
Two years after ChatGPT appeared on the scene, disillusionment reigns. 80 percent of organizations use GenAI, but see no impact on business. Employees use private accounts because initiatives are stuck. While individual productivity increases, the ROI remains invisible.
GenAI is going in the wrong direction. We delegate creative tasks to AI that we like to do ourselves. Knowledge workers spend 60 percent of their time on routines: copying data, checking AI content, implementing recommendations.
The bottom-up era shows: Individual efficiency gains do not scale. Horizontal tools distribute productivity so poorly that they remain invisible.
The solution: Agentic AI – autonomous systems that understand goals, take initiative, and adapt procedures. Agents transfer data, correct errors, implement – supervised, without manual steps.
The potential: Redesigned processes show 20–40 percent time savings and 30–50 percent less backlog instead of 5–10 percent individual efficiency gains.
CIOs need to think strategically: Which processes need to be redesigned? The question should be «What to build differently?» instead of «Where to use AI?».
It's about meaningful work, not job cuts. The shift: Instead of comprehensive tools, CIOs should select a few areas and design outcome-oriented systems in which people work in a targeted manner instead of processing workflows.
The more autonomous the systems, the more critical the governance. Agents with data access and communication capabilities are a high risk. An e-mail agent could exfiltrate data by manipulating instructions. Clear accountability without paralyzing control is required.
Success depends on people. When agents take over routine tasks, time is freed up. How to make the best possible use of it?
Together with their teams, managers must find out where their strengths lie and how they can use them to generate value for the company and clients.
Example: Customer service that does not involve standard inquiries could intensify customer relationships, develop solutions to problems or derive product improvements from customer feedback.
The next 12−18 months will be crucial. Early adopters will build production environments.
CIOs must act: fewer pilot projects, but radical top-down thinking. Technology is developing rapidly, with a lot of innovation taking place via US-based SaaS solutions. This is a challenge for Switzerland. CIOs should partner with Swiss IT companies in order not to lose their sovereignty.
The GenAI paradox can be resolved. With courage, an appropriate strategy, and a focus on human strengths.
This article was originally published in the December 2025 edition of Netzwoche.