Case studies
Two stories, told in full.
These are not our customers: they are the two that best show the extremes. A large company with two audiences to keep apart, and a small team that cannot be on the phone. Everything here is in production and verifiable.
Two audiences, one inbox: how Transfeero talks to passengers and drivers without mixing them up
Transfeero runs private transfers worldwide and has two audiences with nothing in common: the passenger who booked, and the partner who drives. Each gets its own help center, its own knowledge and its own AI agent.
Transfeero · Private transfers and mobility
Read the caseSmall businessA small team answering like a big one: Go-Etna
Go-Etna takes people up Mount Etna. It is a small team, and during the day it is on the volcano rather than in front of an inbox. The AI agent covers the first line, in several languages, including when the office is closed.
Go-Etna · Mount Etna tours and experiences
Read the caseWhat they have in common
A note on what is not here
We do not publish deflection rates, response times or cost savings. Those numbers belong to the customers and we do not circulate them without their explicit approval — and an invented one would destroy the only thing that makes a case study worth reading. What you get instead is a description of how each implementation is configured, with links to the public parts you can check.
Why only two? Because a case study is worth reading when it is detailed. We would rather publish two implementations explained all the way down than twelve cards claiming "+40% efficiency". The other projects are not here because publishing a customer’s name is their decision, not ours.
- 01Both are in production, not demos
- The help centers linked here are the real ones, public, maintained by the customers’ own teams. You can open them right now and judge for yourself the quality of the material the AI agent works from.
- 02Both started from the content
- Neither installed the software first and went looking for content afterwards. In both cases the real work was writing down the answers the team was already giving out loud — which is what decides whether an AI agent is worth anything.
- 03The main channel is WhatsApp
- True for a transfer company operating worldwide and for a handful of mountain guides alike: around the Mediterranean, that is where the customer writes from. Email and the widget come after, and they have to land in the same thread.
- 04The AI is the first turn, not the product
- In both cases the agent handles what is already written down and hands everything else to a person. The difference between them is how much "everything else" there is: a lot for Transfeero, little for Go-Etna.