If you block an AI agent, you are teaching it to recommend someone else

Dvir Ginzburg is the founder and CEO of Encore AI, which builds AI agents for revenue-generating interactions. Earlier in his career, Ginzburg built AI recommendation systems at Facebook and Microsoft. He holds a PhD in deep learning. Encore publishes the Agent Readiness Index cited in this piece.

For about 20 years, the smart engineering decision was to keep automated traffic off your website. CAPTCHAs, bot walls, rate limits, and phone trees were all built on a sound assumption: that software arriving without a person behind it was probably there to scrape, defraud, or overload you. For the most part, this assumption was correct, but it’s the wrong approach for today.

Both technology and the people using it have changed. A growing share of the automated traffic hitting a business now belongs to their own customers, who have started sending personal AI assistants to book flights, pay bills, compare policies, and sit through hold music on their behalf. Traditional website defenses cannot tell the difference, and end up treating the customer as the threat.

Most of the discussion about this has centered around access, which is the wrong frame. The better question is not whether an agent gets in today, but what the agent learns when it does not.

An agent does not experience a block the same way a person experiences a locked door. There is no moment of indignation and no complaint to customer service. What the agent records is a failed trajectory, or a sequence of attempted steps that did not reach its goal. That record does not disappear when the session ends, and it compounds in three places:

The first impact is........

© Fortune