An answer without provenance becomes an argument. The customer remembers one sentence, the supervisor sees another in the transcript, and somebody still has to decide whether the information was approved. An answer with provenance becomes a lookup: open the conversation, follow the recorded source, and compare the response with the knowledge entry behind it.
The chain starts with approved knowledge
SayTrue can use approved Q&A entries, searchable documents, and dynamic entries that expire. Each Q&A entry also carries an autonomy class: answer directly, create a ticket, or transfer to a person. That class controls what the agent may do with the knowledge, not merely what words it can repeat.
When a matching approved entry supports the customer’s question, the agent builds its response from that material. The conversation records a reference to the source used for the lookup. Dynamic facts stop being available after expiry, so an old seasonal notice cannot continue presenting itself as current knowledge.
What the supervisor sees
The conversation page keeps the transcript and the source log together. A supervisor can read the customer’s wording, the answer that followed, and the knowledge reference associated with it. For a live tool call, the transcript also records the read-only endpoint call and its result, keeping that lookup in the same review path.
This matters because the answer is not reviewed in isolation. The team can see what was asked, which route the system took, and whether the approved material actually covered that wording. If an entry is unclear, the fix belongs in the knowledge base, where it updates every enabled channel that uses it.
Disputes become traceable
Suppose a customer disputes an opening time or an instruction. The supervisor does not need to reconstruct the event from memory. They can inspect the transcript, open the cited entry, and decide whether the answer followed the approved fact. If the fact was wrong, the entry can be corrected. If the entry was right but the response was unclear, that is a different issue and should be reviewed as such.
No source stays visible
When approved knowledge does not support an answer, the system records the no-source outcome. It may escalate to a person or record a request, depending on the configured rule and working hours. The unanswered question can also enter the knowledge-gap report with its frequency.
That blank is useful evidence. Covering it with a plausible sentence would remove the very signal the team needs. The source log therefore does two jobs: it shows why an answer was given and preserves the places where an answer could not be justified.