[Image: Marina Risher (right) and Liat Malchi (left). Courtesy of AudioCodes]

Many organizations in sensitive sectors—such as defense, healthcare, and finance—struggle to adopt AI-based solutions for improving internal workflows. The reason? Most of these solutions rely on public cloud infrastructure, which these organizations either avoid or are legally restricted from using due to data security and regulatory concerns.

AudioCodes has recently introduced a suite of generative AI solutions designed to transcribe, analyze, and extract insights from business meetings held on platforms like Zoom and Microsoft Teams. These solutions, marketed under the Meeting Insights brand, are typically offered as cloud-based SaaS services. However, to enable use in closed environments with no internet access, AudioCodes has now launched a dedicated version: Meeting Insights On-Prem. This on-premises variant performs the same tasks entirely within the organization’s internal network—without relying on cloud or internet access.

To power this functionality, AudioCodes developed a small language model (SLM) that can be efficiently deployed using the organization’s existing compute resources. The model can be fine-tuned using the organization’s own data to adapt to its specific terminology and linguistic codes. Over time, accuracy improves as users provide feedback, with AudioCodes reporting up to 95% precision in understanding domain-specific terms and context.

The solution has already been adopted by several leading Israeli defense organizations. In an interview with Techtime, Marina Risher, Senior Product Manager for Secure Regulatory Solutions, explained: “While there’s a broader trend toward cloud migration, some organizations are barred from moving data off-premises. In sectors like defense and government, communication systems still rely on legacy infrastructure. Thanks to our long-standing experience in traditional enterprise telephony, we were able to bridge AI with legacy systems. We initially targeted Israeli defense needs arising during wartime, but soon realized demand for on-prem solutions is global and spans multiple industries.”

AudioCodes specializes in voice and data communication solutions for IP networks. The company provides a broad portfolio of products and services for enterprises, telecom service providers, and global businesses. Its core competencies include VoIP systems, media gateways, analog telephony adapters (ATAs), unified communications platforms, and network monitoring, management, and security services.

Founded in the 1990s, AudioCodes’ legacy in enterprise communication provides a strong foundation—both technically and commercially—for leveraging the power of generative AI. This is particularly true in fields like natural language processing (NLP) and language model development (LLM).

Liat Malchi, Senior Marketing Manager at AudioCodes, commented: “GenAI is fundamentally transforming our voice solutions. It enables us to build intelligent meeting tools suited for all types of organizational discussions—from traditional meeting rooms to hybrid setups and video calls. Our deep expertise gives us a major competitive edge. Every word spoken in a business meeting is a strategic asset, and AI turns those words into actionable business insights.”

Cloud Repatriation on the Rise

While the on-prem solution was initially seen as a niche complement to cloud-based offerings, AudioCodes discovered a much larger opportunity. Contrary to expectations, the market for on-premises solutions is not only sizable—it’s growing. Alongside the massive migration to the cloud, many organizations are now reversing course through “cloud repatriation,” moving workloads and sensitive data back to local servers or hybrid environments. This trend is especially visible in government, healthcare, finance, and defense sectors, largely due to security concerns and regulatory requirements.

Cost is another factor. For some, the long-term expense of public cloud usage surpasses the investment needed to deploy local infrastructure. Risher concluded: “At first, we questioned whether investing in an on-prem solution made sense in a cloud-first world. But we identified a significant countertrend: organizations pulling data and workloads back from the cloud to their own servers. It’s driven mainly by security and cost considerations, and a desire to reduce dependence on external vendors.”