Handle customer complaints with a team of specialised agents
Indonesian e-wallet DANA runs DIANA, a complaint-handling assistant built on Azure OpenAI in which separate agents handle dialogue, context, knowledge lookup and insight analysis.
What they did
DIANA is a task-oriented dialogue system for user complaints. One agent handles the conversation and empathy, another keeps track of context, a third retrieves answers from the knowledge base, and a fourth analyses interactions to turn them into insights for improving service.
What happened
The company says that in 2025 the setup lifted operational productivity by about 57%, customer satisfaction by 13% and resolution speed by about 10%.
Keep in mind
The figures come from a vendor article with no baseline or measurement method. It does not say which complaints still go to human staff.
Try it with your agent
Act as the complaint desk for <business name>. Use <knowledge base> as your only source of answers. For each complaint, acknowledge the customer's problem in one sentence, check the conversation history for context, give the fix or next step from the knowledge base, and escalate to <team> if no answer exists. At the end of the week, summarise the top complaint themes and suggest one service fix for each.
Replace the <angle brackets> with your own details.