The Verified Index
What AI support agents actually do, measured the same way for every account in it, published by each account's own choice, and stamped so a figure cannot be edited after the fact.
Every vendor in this category defines “resolved” itself, measures it itself, and is paid on it. This is the same question asked once, the same way, of everyone willing to answer in public.
The cohorts
| Agent and volume | Accounts | Answered without a source | Confirmed success | Handed to a person |
|---|---|---|---|---|
| Nothing published yet. | ||||
Median, with the middle half beneath it. A cohort under 5 accounts shows a dash: below that a “typical” figure describes the accounts in it rather than the way this kind of agent behaves. Neither the best nor the worst account is ever shown — a minimum or a maximum is one account's own figure.
What this is not
- It is not a sample of the market. Only accounts
that chose to publish are in it, and an account that publishes is
more likely to be one that is pleased with its numbers. Expect this
index to read better than the industry it is drawn from. That
limitation cannot be edited out of the code: the type a cohort is
returned in carries
selfSelected: trueand will not compile without it. - It is not causal. Nothing here says a practice produced an outcome. It says what these accounts reported, over 30 days, measured identically.
- It is not audited. Each figure is stamped, which proves it came from this system unedited. It does not prove the records behind it are complete. That is a seal, not an audit, and calling it more would be the thing this page exists to argue against.