Júlia was applied to Unicall's voice channel in the Sistema Unimed Paraná. The initial scope is concrete: sending payment slips and checking the authorization status of a healthcare authorization request. The operation combines member identification, system lookups, and transfers with context when a person needs to take over.
The problem
Many calls went directly to customer service representatives. Queues and repeated information increased the effort required from members and the operation.
What was put into practice
Júlia was integrated into the interactive voice response system (IVR, known as URA in Portuguese) to identify the member and handle defined workflows, such as authorization inquiries and duplicate payment slips. Requests are routed according to the need.
- 01Voice request
- 02Member identification
- 03Lookup or sending within the covered workflow
- 04Transfer with context when needed
From a voice request to two defined workflows
The account describes an operation in which many calls went directly to human customer service. Queues and repeated context increased the effort required from members and representatives. The solution introduced an agent integrated into the interactive voice response system (IVR, known as URA in Portuguese) to recognize and route requests.
Before transactions, the source describes validation of the member's identity. For the covered tasks, Júlia uses the APIs for payment slips and healthcare authorization requests. Checking the status of an authorization is the published scope; deciding whether to authorize care is not assigned to the agent.
The person receives the point where the conversation stopped
When a transfer occurs, the representative receives the identification and the information collected. The case describes routing to the appropriate queue, seeking to reduce transfers between departments and repeated work. The destination queue and the information that reaches the representative are part of the operation's design.
The source cites interactive voice response (IVR/URA) and telephony, identity validation, ElevenLabs speech synthesis, Google Gemini, and Oracle Cloud infrastructure. It also describes recording, call categorization, and monitoring dashboards. Together, these connect the voice conversation to the systems needed for the two workflows.
Monitoring and results within the scope
The case reports reduced handling time in the automated workflows, resolution of some calls, and better routing to queues. Monitoring separates the time spent on each interaction from the share of calls resolved by the agent, allowing both speed and the solution's reach to be examined.
The production rollout took place in phases, interspersed with adjustments and joint learning. Call categorization and dashboards help identify recurring requests and opportunities for improvement. Reimbursement and contract verification appear as next steps, subject to new workflows and integrations.
The reported result and its scope
The source reports faster service in the automated workflows and less repeated work in call routing.
The results belong to the channel and workflows described. The case does not demonstrate control over actions by agents built outside this solution.
The lesson learned
When a situation needs a person, the data and history are passed along with the handoff. The experience is designed as a shared operation between AI and the team.
What questions does this case help you ask?
- Which requests have a sufficiently clear beginning and end for an initial scope?
- How does the representative receive the identification and the unresolved issue?
- What record distinguishes a completed lookup from an attempt without confirmation?
Source of the account
Summary of the case published by Starya. Results and scope belong to this operation; they are not a guarantee for other contexts.
Read the original account ↗