
SEPTEMBER 2026
The agentic
Experience index
When your customer sends their AI Agents
Can they find you, understand you,
and get the job done?
So we put it to
the test
3 key
Questions
Can agents find you?
190 of 255 companies are reachable
Nearly a quarter of the market is invisible to agents
Can agents read you?
129 of 239 look-up tasks were successful
Half of all look-up tasks can be answered in full
Can agents act?
14 of 360 act tasks completed end-to-end
Hardly any transactional task can be completed by agents
what answers
did our agents get?
3 categories
of agentic experiences
of 245
Invisibles
An agent cannot reach or read them, they are just not part of the conversation.
Silent
No response (an error, a timeout ...). Hard to tell if it’s policy or neglect.
Fortress
Answer an agent with a refusal. A choice, not an accident.
of 245
Informers
Agents can get in and answer a question, but they cannot complete a job.
Showroom
Everything on display but no till. Usually by accident.
Gatekeeper
The task can only be completed behind a login, identity app, or in a dedicated app. Deliberate, every time.
of 245
Actors
Agents can complete at least one job end to end.
Always a decision to point outwards.
What is
stopping
agents?
Technical accidents
of all failures
Nobody in these companies decided to exclude agents. The front end simply happens to be invisible to anything that does not run a browser. These are build choices, and they are the cheapest things on this page to fix.
Strategic choices
of all failures
These are decisions: block non-browser traffic, put the service behind a login or an identity app, send people to the app. Each was rational on its own terms.








