Call center ai · Production

Allianz Direct improves contact center agent-assist accuracy by 10–15% with Databricks Mosaic AI

The problem

Allianz Direct's contact center agents were spending too much time on mundane back-office tasks, leaving less time for direct customer engagement. Policy questions required agents to search multiple systems and second-guess answers rather than focusing on building personal customer relationships.

First attempt

Allianz Direct had an existing GenAI-powered tool for contact center agents, but it was less accurate and required a more complex development and implementation process than the Databricks-based solution.

Workflow diagram · grounded in source
1
Customer policy question received
trigger
“customers had questions about policy terms and conditions. These were common policy questions such as "Can my brother drive my car?"”
2
RAG retrieves T&C information
ai_action
“The team built a proof of concept for a RAG-based agent-assist tool. This application incorporated all of Allianz Direct products' terms and conditions, intelligently delivering the right information to the agent at the right time.”
3
Human agent reviews AI answer
human_review
“the feedback went to a human agent rather than directly to the customer, which provided further confidence”
4
Agent delivers accurate answer
output
“the GenAI agent assist tool quickly delivered accurate information to them”
Reported outcome

The RAG-based agent-assist tool demonstrated a 10 to 15% accuracy uplift over the previous solution, earned agent trust, increased agent usage, and is being rolled out across all Allianz Direct contact centers.

Reported metrics
Accuracy uplift vs previous solution10 to 15%
Agent tool adoptionuse it more often
Agent time with customersspend more time talking to customers
Reported stack
Databricks Mosaic AIDatabricks Data Intelligence PlatformDatabricks NotebookAgent Bricks Custom AgentsUnity CatalogRAGSlack
Source
https://www.databricks.com/customers/allianz-direct
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

The RAG-based agent-assist tool demonstrated a 10 to 15% accuracy uplift over the previous solution, earned agent trust, increased agent usage, and is being rolled out across all Allianz Direct contact centers.

What tools did this team use?

Databricks Mosaic AI, Databricks Data Intelligence Platform, Databricks Notebook, Agent Bricks Custom Agents, Unity Catalog, RAG, Slack.

What results were reported?

Accuracy uplift vs previous solution: 10 to 15%; Agent tool adoption: use it more often; Agent time with customers: spend more time talking to customers (source-reported, not independently verified).

What failed first in this deployment?

Allianz Direct had an existing GenAI-powered tool for contact center agents, but it was less accurate and required a more complex development and implementation process than the Databricks-based solution.

How is this call center ai AI workflow structured?

Customer policy question received → RAG retrieves T&C information → Human agent reviews AI answer → Agent delivers accurate answer.