Three ways teams actually use it
These are anonymised deployment patterns, not case studies with logos and numbers. No customer names, no performance percentages. What you get instead is the configuration, so you can judge whether it applies to you.
Why no names
What this page deliberately leaves out
Customer names and "improved by N%" figures require authorisation we do not have for every deployment, and a percentage from someone else's site tells you very little about yours. So each pattern below states what we would configure first, and how you can check it yourself. Authorised case studies will replace these when we have them.
Industrial materials · 20–200 people
The pattern: enquiries arrive as "does this grade meet spec X" plus a drawing or a specification sheet. Most of them are answerable from material that already exists internally.
Configure it this way: build knowledge entries as model × document type, and route any message containing batch number, customisation, or contract straight to a person.
B2B software · 10–100 people
The pattern: trial users repeatedly ask about pricing structure, whether it integrates with their stack, and where the data lives. One person ends up answering all of it.
Configure it this way: write pricing, deployment, and integration into entries; route the words quote, contract, and on-premise to a human, and carry the conversation over so the visitor does not repeat themselves.
E-commerce · peak season
The pattern: on promotion days, shipping times, return rules, and invoices make up most of the traffic. The questions are nearly identical; the volume is not.
Configure it this way: connect order lookup through the open API so the agent sees the order number; route refund, complaint, and delivery-exception to a human.
Before you look for case studies
You can test it yourself in one evening
Human chat is free forever and registration is instant. Running it on your own site for a week beats reading anyone else's numbers.