Malte Kosub
Co-Founder & CEO, Parloa
I recently asked a room of enterprise and business leaders why they believe AI agents stall after launch. It became clear that after the hype has subsided, the road ahead feels more uncertain.
The evolution from generative to agentic AI is redefining how enterprises interpret data, make decisions and deliver customer experience. As evidence of ROI grows, appetite for agentic AI is undeniable, so why have so few moved past the pilot stage?
Customer experience AI divide
I don’t believe it’s a technology problem. The technology is here: agile, proven and accessible. It’s enterprise maturity (in IT architecture and governance) holding back meaningful progress. Real deployment needs to work across legacy systems, multiple languages and channels and strict compliance requirements — all with zero tolerance for getting a customer-facing interaction wrong. Projects stall here because of the gap between what AI can do in a controlled environment and what it delivers in the real world, against real customers, systems and complexity. We call this the CX AI Divide.
It’s enterprise maturity (in IT architecture and governance) holding back meaningful progress.
What ‘enterprise-ready’ really means
As part of our recent ‘State of Agentic CX’ study*, we tested this gap directly, evaluating 10,000 enterprise websites and conducting 4,000 live chat and voice interactions. The results were stark, with 92.5% of the chatbots we could classify still running on inflexible rule-based systems, not real AI. Just 8.9% of chat conversations resolved the customer’s issue.
This is the problem AI agents are meant to solve. Yet, too often, behind the ambition sits a decades-old infrastructure that’s unable to support the potential of agentic AI. Enterprise-ready AI needs a firm foundation with governance, regular testing and safety controls built in from day one. The practical fix is in better operating discipline. An AI agent isn’t software you configure once. It needs continuous testing, monitoring and optimisation as products, policies and customer expectations shift. Sustained improvement, and ultimately ROI, will come from treating every deployment as an ongoing, evolutionary requirement. The appetite to reinvent customer experience is there. Closing the gap means building a strong foundation – in the tech stack, in the data and in the teams who make it work.
[1] Parloa ‘State of Agentic CX’ Study, 2026
