“How David VonThenen Designs Production-Ready AI Systems and Agent Architectures”

“How Keith Belanger Builds AI-Ready Data Platforms Without Sacrificing Governance or Trust”

Episode Summary Everyone wants faster automation and better AI outcomes, but few organizations stop to examine the data foundations underneath them. In this episode of Automate or Die Trying, host Will talks with Keith Belanger about why acceleration alone is not a strategy—and why AI will amplify both the strengths and weaknesses already present in your data ecosystem. Drawing on three decades of experience in data architecture, including leading a 300+ terabyte migration to Snowflake in a highly regulated insurance environment, Keith explains why many enterprise data failures are not dramatic breaches but quiet, cumulative problems that build over time. He shares lessons from large-scale modernization efforts, the importance of data modeling and architectural discipline, and why organizations often create technical debt by prioritizing speed over design. The conversation explores how AI agents interact with enterprise data, why governance and testing must become continuous practices, and how organizations can detect schema drift, data quality issues, and hidden risks before they impact business decisions. Keith also discusses the growing role of DataOps, the importance of embedding architectural intent into AI-assisted development, and why human oversight remains essential even as automation becomes more powerful. If you're responsible for data engineering, AI initiatives, security, governance, or enterprise architecture, this episode offers practical insights into building data platforms that can scale safely while remaining trustworthy and AI-ready. Key Topics Discussed * Why “acceleration is neutral” in data and AI * Lessons from a 300+ TB Snowflake migration * The hidden cost of cumulative data failures * Data contracts, schema drift, and governance * AI-ready data and enterprise trust * Continuous testing and observability * DataOps automation and CI/CD for data platforms * Human oversight in AI-assisted engineering * Balancing speed, security, and architectural discipline * Building resilient foundations for enterprise AI Guest Information * Keith Belanger – Field CTO, https://www.dataops.live?utm_source=chatgpt.comDataOps.live [https://www.dataops.live?utm_source=chatgpt.com] * LinkedIn: https://www.linkedin.com/in/krbelanger?utm_source=chatgpt.comKeith Belanger LinkedIn Profile [https://www.linkedin.com/in/krbelanger?utm_source=chatgpt.com] Listen Now & Subscribe: Apple Podcasts [https://podcasts.apple.com/us/podcast/automate-or-die-trying/id1858617259], Spotify [https://open.spotify.com/show/7JJN8GliYhaECVEduFcvmW?si=99700a8227cb47eb], Amazon Music [https://music.amazon.com/podcasts/e7e24a5f-61e0-4c7f-8af1-be7b87cbdc02/automate-or-die-trying], or find us wherever you get your podcasts. Powered by cybersecurityprivacy.com [https://cybersecurityprivacy.com/] Automate or Die Trying empowers leaders to adopt AI securely, move faster with confidence, and transform the future of work.

“How Christopher Moravec Is Building Autonomous AI Agents for Government, GIS, and the Physical World”

Episode Summary: In this episode of Automate or Die Trying, host Wil Ramos sits down with Christopher Moravec, CTO and founder of Dymatic, to explore what it actually looks like to deploy AI agents into real-world government, GIS, and public sector workflows. Christopher shares how Dymatic helps cities and agencies modernize permitting, infrastructure management, zoning feedback, and operational processes using AI-powered mapping and automation systems. From analyzing traffic plans to processing thousands of public comments, he explains how location-aware AI is transforming how governments operate at scale. The conversation also dives into Christopher’s personal AI ecosystem, including his autonomous AI agent named “Jaws” — an assistant running from a Mac Mini in his workshop that manages tasks, reviews emails, writes newsletters, controls environmental systems, and is now learning to operate inside a physical robot body it helped code itself. Christopher breaks down the practical realities of deploying AI into production environments, including why most organizations are giving agents too much access, how prompt injection attacks actually work, and why layered security and restricted permissions matter more than flashy demos. He also explains how AI guardrails fail in practice, how poison-pill data attacks could shape future disinformation campaigns, and why AI systems need multiple layers of validation, auditing, and monitoring. They also explore the future of operating systems, autonomous workflows, physical AI agents, and how AI may eventually replace traditional applications with dynamic, generated interfaces tailored to each user in real time. For leaders experimenting with agentic AI, Christopher emphasizes the importance of starting small, focusing on measurable workflow improvements, and limiting AI access to only the systems and permissions truly required. Key Takeaways: * AI agents should start with narrowly scoped tasks before expanding autonomy * Most AI security failures come from giving agents unnecessary permissions * Prompt injection attacks remain one of the biggest risks in production AI systems * Layered validation and monitoring are essential for safe AI deployment * AI systems consuming public data can become vulnerable to disinformation poisoning * Autonomous agents require immutable logging and continuous auditing * The future operating system may revolve around AI-generated interfaces instead of apps * AI-assisted automation can dramatically reduce repetitive operational workloads * Physical AI agents and embodied systems are becoming increasingly practical * The best AI deployments solve boring operational problems that save real time Connect with Christopher Moravec: * LinkedIn: https://www.linkedin.com/in/christophermoravec [https://www.linkedin.com/in/christophermoravec] Websites: * https://moravec.net [https://moravec.net/] * https://moraveclabs.com [https://moraveclabs.com/] Listen Now & Subscribe: Apple Podcasts [https://podcasts.apple.com/us/podcast/automate-or-die-trying/id1858617259], Spotify [https://open.spotify.com/show/7JJN8GliYhaECVEduFcvmW?si=99700a8227cb47eb], Amazon Music [https://music.amazon.com/podcasts/e7e24a5f-61e0-4c7f-8af1-be7b87cbdc02/automate-or-die-trying], or find us wherever you get your podcasts. Powered by cybersecurityprivacy.com [https://cybersecurityprivacy.com/] Automate or Die Trying empowers leaders to adopt AI securely, move faster with confidence, and transform the future of work.

“How Lawrence Pingree Is Redefining AI Security, Agentic Defense, and Runtime Trust”

Episode Summary: In this episode of Automate or Die Trying, host Wil Ramos sits down with Lawrence Pingree, Head of Security and AI Research at SACR (Software Analyst Cyber Research), to explore how AI agents are fundamentally reshaping cybersecurity, governance, and enterprise defense strategies. Lawrence shares insights from a career spent defining major cybersecurity categories including EDR, content disarm and reconstruction, cloud incident response automation, and generative AI runtime defense. Drawing from his experience at Gartner, SonicWall, and SACR, he explains how industry terminology becomes standard practice and why language itself plays a critical role in advancing cybersecurity innovation. The conversation dives deep into the rise of agentic AI and why traditional security architectures are struggling to keep pace with machine-speed automation. Lawrence introduces concepts like identity dark matter, Unified Agentic Defense Platforms (UADP), and Just-in-Time Trust, describing how non-human identities and autonomous AI agents are creating entirely new operational and governance challenges for modern enterprises. They also unpack how runtime security is evolving beyond traditional EDR models, why organizations must rethink prevention versus detection, and how AI systems introduce additive risks on top of existing human vulnerabilities. Lawrence explains why future security strategies will require real-time behavioral scoring, federated trust systems, and continuous runtime inspection across AI workflows and agent communications. For leaders navigating the next wave of AI adoption, Lawrence emphasizes the importance of practical security design, runtime visibility, identity management, and building defenses that can adapt as fast as the technology itself evolves. Key Takeaways: * AI agents are introducing new forms of identity risk and “identity dark matter” * Traditional zero trust models are struggling to keep pace with agentic AI systems * Runtime security and real-time prevention are becoming critical for AI defense * Unified Agentic Defense Platforms (UADP) aim to standardize AI runtime protection * AI vulnerabilities are additive to existing human and infrastructure security risks * Non-human identities now vastly outnumber human identities in enterprise environments * Behavioral scoring and federated trust models may become essential for AI governance * Security automation has always existed, but AI dramatically increases its scale and speed * Prompt injection, credential leakage, and agent misalignment are emerging enterprise risks * Future security strategies must prioritize runtime visibility and adaptive prevention Connect with Lawrence Pingree: * LinkedIn: https://www.linkedin.com/in/lawrencepingree [https://www.linkedin.com/in/lawrencepingree] Websites: * https://lawrencepingree.com [https://lawrencepingree.com] * https://geek-guy.com [https://geek-guy.com] * https://iuniverse.com [https://iuniverse.com] Listen Now & Subscribe: Apple Podcasts [https://podcasts.apple.com/us/podcast/automate-or-die-trying/id1858617259], Spotify [https://open.spotify.com/show/7JJN8GliYhaECVEduFcvmW?si=99700a8227cb47eb], Amazon Music [https://music.amazon.com/podcasts/e7e24a5f-61e0-4c7f-8af1-be7b87cbdc02/automate-or-die-trying], or find us wherever you get your podcasts. Powered by cybersecurityprivacy.com [https://cybersecurityprivacy.com/] Automate or Die Trying empowers leaders to adopt AI securely, move faster with confidence, and transform the future of work.

“How John Shende Is Redefining AI Security, Agentic Governance, and the New Insider Threat”

Episode Summary: In this episode of Automate or Die Trying, host Wil Ramos sits down with Jon G Shende, CTO of Data Security & AI and Global VP of SecOps at Sutherland, to explore what it really takes to secure AI at enterprise scale. Jon shares how he helped rebuild a modern AI-driven security stack in partnership with Google Security, retiring legacy systems and standing up a new, machine-learning-powered operation from the ground up. He also reflects on his early work coining “cloud forensics as a service,” a concept that anticipated many of today’s cloud security challenges. The conversation examines the rise of agentic AI and why traditional governance models are no longer sufficient. Jon explains how AI is becoming the new insider threat, not because it’s malicious, but because of the level of access, autonomy, and trust organizations are giving it. They also unpack real-world lessons from large-scale security transformations, including the challenges of aligning stakeholders, managing risk, and ensuring data integrity across complex systems. Jon emphasizes that while automation and AI create massive opportunities, they also require strong fundamentals in access control, policy enforcement, and auditability. For leaders building in today’s AI-driven world, Jon highlights the importance of collaboration, practical security design, and responsible adoption of emerging technologies. Key Takeaways: * AI is becoming the new insider threat due to its trusted access and speed at scale * Governance must evolve to operate at the workflow level, not just static control points * Overprivileged AI agents increase the risk of data exposure and misuse * Automation without strong fundamentals can amplify security failures * The ROI of security lies in preventing incidents, not generating revenue * Large-scale transformations require alignment across technical and business stakeholders * Great leadership is about enabling teams, not taking the spotlight Connect with Jon G Shende: * LinkedIn: https://www.linkedin.com/in/jonshende [https://www.linkedin.com/in/jonshende] * Website: https://cloudfraas.com [https://cloudfraas.com] * Blog: https://jonshende.blogspot.com/view/magazine [https://jonshende.blogspot.com/view/magazine] Listen Now & Subscribe: Apple Podcasts [https://podcasts.apple.com/us/podcast/automate-or-die-trying/id1858617259], Spotify [https://open.spotify.com/show/7JJN8GliYhaECVEduFcvmW?si=99700a8227cb47eb], Amazon Music [https://music.amazon.com/podcasts/e7e24a5f-61e0-4c7f-8af1-be7b87cbdc02/automate-or-die-trying], or find us wherever you get your podcasts. Powered by cybersecurityprivacy.com [https://cybersecurityprivacy.com/] Automate or Die Trying empowers leaders to adopt AI securely, move faster with confidence, and transform the future of work.

“How Ryan Donahue Secures Millions of Devices and Billions of Transactions at Comcast”

Episode Summary: In this episode of Automate or Die Trying, host Wil Ramos sits down with Ryan Donahue, Senior Director of Cybersecurity at Comcast. Ryan leads vulnerability management, asset inventory, compliance, and fraud detection across one of the largest and most complex networks in the world, supporting millions of internal assets and more than 100 million customer devices. Ryan shares how his background as a software engineer shaped his approach to security leadership. Instead of relying solely on vendor tools, he often builds internal systems to solve security challenges at massive scale, including credential management platforms handling billions of transactions per day and automation frameworks designed to protect telecom infrastructure. The conversation explores the realities of securing large-scale telecom systems, where pushing updates or configuration changes can impact millions of devices simultaneously. Ryan discusses the challenge of maintaining visibility across massive infrastructure and how his team works to identify unknown assets and reduce operational risk. Ryan also explains why automation must be implemented carefully in security environments. While automation enables scale, poorly designed systems can amplify vulnerabilities. His approach focuses on building secure defaults into platforms so engineering teams can move quickly while maintaining strong security standards. For security leaders and engineers, Ryan emphasizes the importance of strong debugging skills, clear security metrics, and designing platforms that make security scalable across large engineering organizations. Key Takeaways: 1. Security at telecom scale requires deep engineering and custom tooling 2. Asset visibility is the foundation of an effective security program 3. Automation must be designed carefully to avoid amplifying risk 4. Security platforms should simplify protection for developers 5. Security reporting should focus on patterns rather than raw vulnerability counts 6. Strong debugging and system-level thinking remain essential skills Connect with Ryan Donahue: 1. LinkedIn: https://www.linkedin.com/in/ryan-donahue-28324a5https://www.linkedin.com/in/ryan-donahue-28324a5 [https://www.linkedin.com/in/ryan-donahue-28324a5] 2. Website: https://rydonahue.comhttps://rydonahue.com [https://rydonahue.com] Listen Now & Subscribe: Apple Podcasts [https://podcasts.apple.com/us/podcast/automate-or-die-trying/id1858617259], Spotify [https://open.spotify.com/show/7JJN8GliYhaECVEduFcvmW?si=99700a8227cb47eb], Amazon Music [https://music.amazon.com/podcasts/e7e24a5f-61e0-4c7f-8af1-be7b87cbdc02/automate-or-die-trying], or find us wherever you get your podcasts. Powered by cybersecurityprivacy.com [https://cybersecurityprivacy.com/] Automate or Die Trying empowers leaders to adopt AI securely, move faster with confidence, and transform the future of work.