Table of Contents
What you'll learn
Consumer agent traffic grew 10x in September. Weekly volume rose from ~40 to ~400 chats and has held there. It’s small enough to hide in a dashboard and growing fast enough to matter.
Fewer than 3 in 10 say they’re AI. Most call themselves an “assistant,” and four in five never name the product behind them. Scanning chats for “AI” misses most of them.
They get the price first and commit later. At the airlines, agents ask for the exact total and wait for their person’s approval about 7x as often as comparably detailed human customers.
They come to negotiate. At one telecom provider, 56% of agents wanted a lower bill or retention offer, 4.3x the rate of comparable customers. Many arrive with a competitor’s price in hand.
They bring the whole story. The median opening message is 366 characters, vs. 35 for a typical customer, with account details, amounts and clear limits.
Over 60 days, we found about 2,300 customer-service chats at four ASAPP customers (two major airlines and two telecom providers) where the “customer” was an AI agent acting for a person. The Index is built from that real customer data, not a survey, and we’ll release a new edition every month.
Inside the report
The full findings by industry, the AI products behind the agents, and five moves CX leaders can make now, from measuring agent traffic to building delegated authorization.
See how agent traffic shows up in your own queue. [Book a working session]
Vol. 1 covers Aug 29 – Oct 28, 2026, across four US consumer brands. Each conversation was read by an LLM, which identified AI agents and labeled their goals and tactics.
