AI Agents That Hit a Wall Will Recommend Your Competitor Instead, Encore AI Founder Warns

Businesses that block AI shopping agents from interacting with their sites may be training those same agents to send customers to competitors instead, according to Dvir Ginzburg, founder and chief executive of Encore AI, a company that builds AI agents for revenue generating interactions. Ginzburg, who previously built AI recommendation systems at Facebook and Microsoft and holds a PhD in deep learning, argues that an agent turned away from a business does not get frustrated the way a person would. It simply records that the attempt failed and moves on.

Encore AI's own Agent Readiness Index, drawn from more than 1,500 organizations across 25 industries, found that 94% could not accept a connection from an AI agent at all, and 26% actively block agents that identify themselves as such. Only 6% of the organizations surveyed had published any public position on how they handle AI agents, and just 1% let customers approve an agent for a narrow, limited task.

When an agent hits a wall, Ginzburg argues it affects more than that one interaction. The task gets diverted to a competitor on the spot, the agent builds a personalized memory of which sites actually work for that kind of request, and the blocked business risks being excluded from the data agents are trained on in the first place over time.

Ginzburg's prescription is for businesses to publish clear, machine readable criteria for what agents are and are not allowed to do on their site, build direct connections for agents rather than forcing them through interfaces designed for humans, and let customers delegate narrow, time limited tasks to agents using open standards rather than blocking the interaction outright. He frames the gap as a strategic blind spot: most companies by now have a mobile strategy and a social media strategy, he says, but very few have anything resembling an agent strategy.