S12 Bonus: Intercepting Risky Prompts and Deploying Local LLMs to Neutralize Data Leakage via Pragatix with Yoav Crombie, Co-Founder & CEO of AGAT Software

Episode Summary

Executive Summary: Joav Crombie, CEO of Agat, explains how his company evolved from an early meeting-summarization AI product into Progodix/PragatiX, an enterprise platform for secure AI use. Drawing on decades in cybersecurity, he focuses on on-prem AI, AI firewalls, model risk checks, and behavior intelligence for regulated customers who need control, privacy, and visibility.

Main Topics: From early AI assistant to secure AI platform (Priority: 5/5): Agat first built a meeting-analysis assistant with manually labeled data, but ChatGPT commoditized the use case, forcing a pivot toward enterprise secure AI. Cybersecurity roots shaping product strategy (Priority: 5/5): Joav’s background in email security and application firewalls for regulated customers informed the company’s approach to AI security and on-prem deployment. Secure AI for regulated and isolated enterprises (Priority: 5/5): The product targets customers who cannot expose data to public models or even access the internet, offering an offline AI suite connected to internal data sources. AI firewall and model risk controls (Priority: 4/5): Beyond AI services, the company provides visibility, policy enforcement, and model analysis to monitor employee AI use and reduce risks like prompt injection and jailbreaking. Customer-led iteration and rapid prototyping (Priority: 4/5): Joav emphasizes short feedback loops, prototype-first development, and direct customer conversations to shape the roadmap and validate needs. Team culture, hiring, and leadership philosophy (Priority: 4/5): He values long-tenured staff and hires for the 4H framework: humble, honest, hungry, and happy, while also knowing when a fit is not right. Belief, focus, and scaling with investor support (Priority: 4/5): After 12 years bootstrapped, the company brought in a private investor to accelerate growth and pursue leadership in secure AI, guided by a 'think big' mindset.

Key Arguments: Public AI tools create unacceptable data exposure for regulated, military, and highly secure organizations, so enterprises need private, controllable AI alternatives. A combined offering of secure AI services plus AI security/firewall capabilities is more valuable than focusing on only one side of the problem. The company’s prior cybersecurity and proxy/firewall experience gives it a strong technical foundation for AI visibility and control. Customer feedback should drive product direction; prototypes and short iterations are the fastest way to validate market needs. AI adoption is not only a security issue but also a behavior and productivity issue, so visibility can support training and better usage. Hiring for cultural fit and character is essential; technical skill alone is not enough. Bootstrapping built discipline, but outside capital became necessary to move faster in a rapidly changing AI market.

Data Points: Years in tech: 35 years - Joav says he has been in the tech industry for 35 years. Army service: 6 years - He served in the Israeli Army as an engineer for six years. Company size: ~25 employees - He says the company is small, around 25 people, while serving Fortune 500 customers. Fortune 500 customers: 25 - He says the application firewall line sold to 25 Fortune 500 companies. Bootstrapped period: 12 years - He says the company was bootstrapped for 12 years before taking investment. Core developer tenure: 5 to 10 years - Most core developers have stayed with him for five to ten years. Investor experience: 2 exits / nine-digit deals - He describes the private investor as having completed two very large exits in the past.

Pivotal Quotes: "If you think big, you'll become big." — Joav Crombie: His advice on mindset, ambition, and how belief shapes outcomes. "The 4H stands for humble, honest, hungry, and happy." — Joav Crombie: His hiring framework for evaluating candidates and team fit. "We enable AI for companies that have problems or restrictions from using AI." — Joav Crombie: His concise description of the company’s mission and market position.

Implications: The episode shows secure AI becoming a major enterprise category, especially for regulated industries. It also suggests that AI vendors will need to combine privacy, governance, and adoption analytics—not just model access—to win trust and scale.