Imagine a customer asks their AI agent to rebook a trip. They’ve already chosen you. But if your service blocks automated traffic, the agent may never reach the booking.
On 5 November, Cloudflare outlined its work on Trusted Agent Delegation. It raises a question for any business whose customers might use AI agents online. How can a service verify who’s behind the agent and establish who’s accountable for its actions? Cloudflare describes ongoing work on open protocols and collaboration across the industry. Read Cloudflare’s post
The same agent service can be used by legitimate customers and by people trying to abuse your service. Block the traffic and you risk turning both away. Let the agent through and you still need to know what the customer has authorised. Asking an agent to find a better departure is one thing. Giving it permission to change the ticket and accept an extra fee is a separate decision.
Where should the customer take over?
Start with a task where an agent could save the customer time. In a rebooking flow, it could find alternative departures and show what they cost. If switching comes with an extra fee, the service can ask the customer to approve it before changing the booking. The software needs to verify who the agent represents, enforce the limits of its permissions and keep a record of what happened.
HiQ’s AI Autonomy Playbook gives you tools to decide how much freedom to give an agent, assess the cost of a mistake and identify where human oversight is needed.
Download the AI Autonomy Playbook
