AI chat can make a website more useful when it helps visitors describe their needs, understand service categories and prepare better enquiries. It becomes risky when it tries to quote, negotiate, approve scope or discuss confidential information.
A safe public AI layer needs four parts: a public knowledge base, a knowledge policy, human verification, and server-side API key handling. It should also escalate quotes, contracts, payments, invoices, complaints, private data and formal commitments to a human.
ONEXUS treats AI as a routing and summary layer. Human judgement remains responsible for commercial decisions.
