No Cloud, No Leaks: This Is What the Private AI Revolution Looks Like

As organizations grow increasingly concerned about data leakage, lack of control, and the false sense of security offered by public cloud models, AGAT's Pragatix platform delivers a holistic Private AI and AI Firewall framework. This enables the adoption of advanced artificial intelligence without compromising data privacy, while providing comprehensive risk management and control over AI usage across the organization.

For three years, while the world was captivated by the capabilities of ChatGPT, Gemini, cloud-based tools, and other public generative AI solutions, recognizing the security implications, Yoav Crombie and his team at Jerusalem-based AGAT Software worked quietly behind the scenes. They recognized early on that the future of AI, particularly for organizations handling sensitive information, did not lie in the public cloud but within on-premises environments.

Crombie did not wait for regulation to slow the AI revolution. Instead, he founded Pragatix, an AGAT spin-off that enables security, financial, and government organizations to run advanced large language models (LLMs) in an environment completely disconnected from the internet. "It's a security-first AI platform, designed for organizations that simply cannot afford information leakage," says Crombie.

All AI services under one secure roof

Pragatix's flagship offering is the AI Private Suite, a comprehensive collection of advanced AI services that operate directly on organizational databases. Unlike public models, the AI here accesses information entirely in-house. "We bring the model to the data, not the data to the model," Crombie explains.
The platform includes, among other capabilities, an organizational chatbot and an intelligent search engine with unique mechanisms designed to prevent hallucinations and deliver reliable answers based on organizational information. It also features AI agents that perform tasks such as document translation, image analysis and generation, and audio content creation, as well as engines that enable automations using these agents and tools.

The system includes a unique agent that supports data analysis and business insights without requiring the AI to be exposed to all table data. This enables Text-to-SQL querying to extract insights without exposing information externally. The platform integrates built-in information security components through an AI Firewall, with policies that can be defined according to information sensitivity, group affiliation, and purpose of use. It also includes a protection layer that detects anomalies in user behavior and monitors AI-based cyber threats in real time.

All information within the system is synchronized with the organization's existing authorization framework. As a result, an employee will never receive responses derived from information they are not authorized to access. For example, if an employee attempts to retrieve salary-related data without proper authorization, the system will not provide an answer, even if the model technically has access to that information.

"Employees get all the tools they are familiar with from the consumer world, while the CIO can rest assured that sensitive information remains locked in a digital vault," Crombie emphasizes. "Organizations no longer need to invest millions in infrastructure. We deploy the most up-to-date model directly on the customer's premises, delivering cloud-level performance on-site. Moreover, because the model is trained on the organization's specific data, it is often more accurate and relevant than generic public models."

Risk management for public AI as well

AGAT also offers a solution for organizations that choose to use public AI tools but require control and transparency over how employees use them. This is where the AI Firewall plays a critical role, providing IT teams with full visibility and control over interactions between employees and cloud-based AI services.
"The AI Firewall understands not only who is using AI, but also the purpose of use and the sensitivity of the information involved, enabling accurate risk assessment," Crombie explains. The AI Firewall provides three layers of protection: Usage, Agent, and Model. The Usage layer is a real-time proxy that governs access to public AI services like ChatGPT, Copilot, and Gemini. The Agent layer controls AI agents via MCP servers, enforcing policy on access and behavior. The Model layer includes a Risk Engine that evaluates AI models using source intelligence and vulnerability assessments.

Leading the way

Crombie is a serial entrepreneur with over three decades of experience in the technology industry, including a successful exit from Symantec. In 2013, he founded AGAT Software, which has built a strong reputation for delivering security and compliance solutions for leading communication and collaboration platforms such as Skype for Business, Microsoft Teams, Zoom, and Slack. Today, the company serves hundreds of global customers, including 25 Fortune 500 organizations.

AGAT operates out of the Har Hotzvim hi-tech park and employs approximately 25 people. The company is committed to occupational diversity across sectors, genders, and geographies. While most of its clients are international, AGAT is increasingly expanding into the Israeli market, where several pilot deployments are already underway. The company maintains a broad international presence across the United States, Europe, and Australia, and recently opened a branch in South Africa. After operating as a bootstrapped company for 12 years, AGAT recently welcomed a private investor.

"Beyond the business aspect, I believe we are contributing something meaningful," says Crombie. "We are enabling people to use artificial intelligence without unnecessary risk, and without being exposed to misleading AI-generated outputs, thanks to built-in validation mechanisms for every response."
You have undergone a fascinating evolution, from a company focused on monitoring and filtering communications to a leading player in the Private AI space. What vision continues to drive you forward?

"The vision is for Pragatix to become the leading solution for the safe implementation of artificial intelligence within organizations. Today, most companies focus either on AI infrastructure or on AI security infrastructure. We are pioneering a new approach that delivers a holistic, one-stop solution combining both capabilities on a single platform. Our goal is to remove fear from the equation and enable even the most sensitive organizations to embrace the AI revolution with complete confidence, ensuring that not a single piece of information ever leaks."

To Website>>>
In cooperation with AGAT Software