Set up a support assistant with 60-day memory

AT&T used Gemini to route support requests and handle routine calls and chats with recent customer history in view.

Done with Geminiยท Advanced

What they did
AT&T moved away from fixed call-flow rules and used BigQuery to look at customer history before deciding where a caller should go. If the system sent someone to a virtual assistant, Gemini Enterprise for Customer Experience handled the conversation, with access to up to 60 days of activity across web, retail, and phone. AT&T said the assistants were kept secure and closely watched.
How it went
AT&T says the change led to about 20% better containment at the start of calls, 20% more inquiries fully automated, and roughly $150 million in yearly savings. It also reported a better Net Promoter Score. The source does not break the results out by channel.
Tips
Use a memory window that matches your support needs, and make sure the assistant can hand off cleanly to a specialist when it should. If you copy this pattern, define what history is fair game and review the test conversations before launch.

Try it yourself with Gemini

Iโ€™m building a voice and chat support assistant for [business type]. Use the policies and FAQs below: [paste materials]. First, decide what the assistant should remember from the last [time window] and what it should ignore. Then write the assistant instructions, the types of requests it can resolve on its own, and the cases that must go to a human agent. Before any irreversible action such as sending, deleting, or submitting, stop and ask me to approve it. Finish with 10 test conversations that include common questions, tricky edge cases, and at least one handoff to a person.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on cloud.google.com โ†—

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.