Texa AI → Quality checks every conversation, Texa AI’s and your team’s, on every channel. When a check finds something, the people you choose hear about it within minutes, with a link to the conversation.
Monitors
Choose New monitor, start from an example or write your own, and set:
- Name and Severity.
- How: by words, or in plain English.
- Words: whole words, any case. End a word with
*for any ending:refund*finds “refunds” and “refunded”. - What to look for, in plain English: say it the way you would tell a supervisor, such as “The customer says they will report us to the regulator”. Write one sentence. It’s what the alert says, and what the judge reads.
- Words: whole words, any case. End a word with
- Who it watches: Anyone in the conversation, The customer, Texa AI or An agent.
- Scoring (optional): points for each thing it finds, so findings are ranked.
- Alerts: who hears about it, and how: email, SMS, Slack, WhatsApp or a webhook. Choose Send a test alert to check it. At most one alert an hour is sent per conversation. Later findings are still listed.
Examples: “Regulator threat” (CBN, regulator*, report you), or “Customer says they are leaving”.
Findings
Findings lists what the monitors found, with the words that were said. Open a conversation to read it in full, Mark as seen, or Draft a procedure from it when the finding shows something Texa AI should handle better.
Reviews
Reviews lists closed Texa AI conversations for your team to score. Read each one the way the customer did, then say for each line whether Texa AI got it right. Texa’s judge scores the same conversations. Your reviews are how it learns to agree with you, and its scores stay hidden until it does.