Leumit Health Services: Building a GenAI-Based IVR Message Creation Platform for Emergency and Operational Use

‍

Overview

Leumit Health Services is a healthcare organization operating in Israel. In a sector where communications must be clear, timely, and accessible, the ability to update voice messaging across multiple channels is operationally significant, particularly during periods of disruption or emergency. In this case, the organization needed a more resilient and responsive way to produce IVR content without depending on a traditional recording workflow.

‍

Challenge

The project was triggered by the discovery that IVR voice messages were still being produced through a recording studio process. While workable under normal conditions, that model introduced a dependency that became problematic in urgent situations. The organization identified that the process was not flexible or fast enough for emergency response needs, where messages may need to be created, updated, and published in very short timeframes.

Three core challenges shaped the initiative.

First, the output had to deliver accurate pronunciation, including terminology and phrasing relevant to the medical domain. In healthcare communications, even small pronunciation errors can reduce clarity and trust.

Second, the solution had to preserve translation accuracy across non-Hebrew languages, specifically English, Russian, Arabic, and Amharic. This was not only a language issue, but also a consistency and quality challenge, as messages needed to remain faithful to the intended meaning across audiences.

Third, the organization required a meaningful improvement in operational speed and flexibility. The new process needed to reduce reliance on manual studio production and enable internal teams to produce messages more quickly, in a format suitable for IVR systems.

From a strategic standpoint, success meant enabling the production of GenAI-based IVR messages in multiple languages, with appropriate voice gender options and the ability to generate messages with or without background audio elements, according to business need.

‍

Solution

CodeValue Group designed and delivered a dedicated platform tailored to Leumit Health Services’ requirements for GenAI-based IVR content creation. The scope included both the underlying infrastructure and the application layer required to generate voice messages in a format aligned with IVR operational needs.

The engagement followed an Agile delivery model and covered the full product development lifecycle. This included infrastructure setup, architecture design, product definition, backend and frontend development, UI/UX work, and production deployment. Rather than delivering a narrow technical component, the program addressed the end-to-end operating need: enabling authorized users to generate IVR-ready messages in a more responsive and controlled way.

At a high level, the project progressed through four major phases: establishing a test environment, developing the system itself, standing up the production environment, and completing user testing. This sequencing supported controlled validation before operational rollout and reduced risk during transition from development to live use.

The operating model also reflected cross-functional coordination. The work included a project manager, a delivery team of developers, and infrastructure specialists from across the group. Weekly status meetings were held with the client, supporting decision-making, visibility, and ongoing alignment throughout delivery.

‍

Technology Stack

The solution was built in the cloud and incorporated LLM and text-to-speech capabilities as core components of the architecture. These capabilities were applied to support multilingual message generation and audio output creation for IVR use cases.

A clear separation was maintained between test and production environments. This was an important architectural and operational choice, allowing the client to validate output quality, user flows, and readiness before live deployment. Given the sensitivity of healthcare communications and the emergency-response context, this separation supported more disciplined rollout and reduced the risk of introducing unverified changes into operational use.

While the technical design was intentionally aligned to Leumit’s specific requirements, the broader architecture also reflected standard enterprise priorities: controlled environments, fit-for-purpose integration with IVR workflows, and an application layer designed around practical operational use rather than experimental AI adoption.

‍

Results

No verified quantitative metrics were provided for publication. However, the qualitative outcome was clear: the platform was put into operational use during an emergency scenario, demonstrating that the organization had moved from a studio-dependent process to a more agile and actionable capability.

Beyond the immediate use case, the project established a foundation for broader adoption. The client expressed interest in expanding the system from an emergency-focused capability into a wider operational platform that could support additional Leumit channels over time.

In practical terms, the initiative delivered more than a technology implementation. It created a structured capability for faster multilingual IVR communication, improved operational flexibility, and a more sustainable model for future voice-message production. For a healthcare organization operating in a high-responsibility environment, that shift represents a meaningful step toward more resilient digital operations

‍