We’re seeing a lot more agentic assistant tools contacting our organization and navigating our virtual agents’ menus, speaking with agents, and generally just wasting time. Often, the user execution of these assistants is often incomplete and leaves our agents unable to actually complete any work. I’ve been investigating the personal agent protocol and how it’s being implemented, but it seemingly has no answer for voice yet.

Today Cloudflare announced that it has the opportunity to detect which virtual assistant is being utilized when a page is being accessed, and several institutions have identified themselves as deploying PAPI but no published metrics yet.

What do you all think? Should the PAP protocol be extended to cover voice channel communications, or should another solution be identified for voice and in general?

I can see the next six months being very assistant-heavy if Meta decreases their bar of entry from $20, or another major player enters the space.