Constraint
A website assistant should not present an unsupported answer as if it came from approved business knowledge.
A website assistant built and operated by SwitchCase Studios, grounded in selected knowledge and connected to a clear route to a person when help is needed.

Visitors ask questions at the moment of intent. The experience should help them find a relevant answer and a useful next step without hiding what the answer is grounded in.
Companion brings source curation, cited answers, appearance, and review into one workflow so teams can meet visitors with useful knowledge and a clear path to contact when a person is needed.
A website assistant should not present an unsupported answer as if it came from approved business knowledge.
Answers are checked against retrieved passages and citations; unsupported output is replaced before it reaches the visitor.
A strict knowledge boundary can leave some questions unanswered, where a less constrained assistant might attempt a broader reply.
The visitor experience can make evidence and escalation part of the interaction instead of hiding uncertainty behind a fluent response.
Select the pages and documents the assistant is allowed to use.
Show citations with supported answers so visitors can inspect where they came from.
Use the configured support path when the available knowledge does not support an answer.
Yes. Companion is an owned SwitchCase Studios product, built and operated as a working example of source-grounded product design, engineering, and ongoing improvement.
No. It complements existing content and contact paths by helping visitors find selected knowledge, then giving them a configured route to your team when they need a person.
No. Citations and answer checks support review, while response rules can make uncertainty and contact escalation explicit when the selected knowledge does not support an answer.
The next move
SwitchCase can help shape the knowledge boundary, interaction model, and production system behind an AI experience that stays clear about what it knows and what people should do next.