UNZER pilots agentic transactions and forecasts 25% adoption

UNZER, a Berlin-based payments company, is conducting two pilot projects with German mid-sized retailers to execute the first agentic transactions. Pascal Beesch, Chief Commercial Officer at UNZER, stated that the company is building infrastructure including "agentic flagging" protocols required by card schemes like Mastercard. Beesch predicts that the first large-scale agentic transaction with a major agent will occur next year, following the launch of Meta's new agent "Muse" in the US. He forecasts that within the next four to five years, 20 to 25 percent of online purchases could be executed by AI agents, a slower adoption curve than initially hoped.

Beesch distinguishes between "conversational commerce," where agents assist with research, and fully autonomous agentic commerce, where agents are mandated to pay. Consumer surveys indicate that while up to 40% of respondents are willing to mandate an agent for purchases, more than 60% of those still want a final human confirmation step. He notes that standardized items like business travel bookings are prime candidates for agent-driven purchasing, whereas experiential items are less likely to be automated.

A key technical challenge is "GEO" (Generative Engine Optimization), the equivalent of SEO for AI models. Beesch explains that AI agents do not respond to traditional marketing materials but require structured data to match user prompts effectively. He warns that the reliability of community data is a security concern, as a significant portion of online traffic may be bot-generated, potentially misleading agents. Beesch argues that companies must shift their marketing strategies from targeting humans to targeting machines, creating data structures that are readable and trustworthy for AI agents.