From API Economy to Agent Economy/Oren Michaels - Barndoor AI, OpenClaw, Agentic Governance ~ Spark of Ages Ep 60

Rajiv talks with Oren Michaels about why enterprise AI ROI stalls when tools live in chat and stay stuck in read-only mode. We break down how a control plane for AI agents enables safe write access across SaaS and internal systems while keeping governance tight.

• chat interfaces working for developers but not for most business roles
• connectivity and trust as the core blockers to enterprise AI adoption
• one control plane to manage many AI models and agent tools
• task-level guardrails that limit what each agent can do
• a concrete write-access workflow across Google Sheets, email, and Salesforce
• scaling protections like rate limiting, token spend controls, and activity logs
• MCP as a way to expose a safe subset of API capabilities to an LLM
• least privilege for agents with gradual permission expansion over time
• moving from identity-based access to intent and context-based policies
• Venn.ai as a single-user on-ramp to governed automation
• Broadway business history in the Spark Tank and why creative structure matters
• lessons from selling Mashery to Intel and operating inside a zero-failure culture
• founder advice on execution and choosing co-founders wisely

AI looks magical in a demo, then it hits the enterprise and the ROI mysteriously evaporates. The gap is not model quality, it is operations: most companies trap AI in chat, strip away connectivity, and keep it read-only because write access feels too risky. That is how you end up with polished summaries instead of real work getting done.

We sit down with Oren Michaels, founder of Barndoor AI and former CEO of API management pioneer Mashery (acquired by Intel), to unpack what it takes to run an agentic workforce safely. We dig into the idea of an enterprise AI control plane that lets teams pick the best agents and models while centralizing governance, visibility, and policy. Oren shares why “least privilege” matters more for probabilistic AI agents than it does for trusted humans, how Model Context Protocol (MCP) can expose safe tool capabilities, and why CISOs want governance outside of the AI vendor’s app.

You will also hear concrete examples of high-ROI write access workflows, like pulling context from Gmail and Google Sheets, checking Salesforce, updating records, and drafting follow-ups in minutes. We explore scaling issues such as rate limiting, LLM token spend, and audit logs, plus a crucial architectural shift from identity-based access (“who are you”) to intent-based access (“what are you trying to do”). Oren also explains why Venn.ai exists as a lower-risk way to build trust with agentic automation before the full enterprise rollout.

If you like practical enterprise AI, AI governance, security for AI agents, and real agentic workflows that move the business forward, subscribe, share this with a teammate, and leave a review so more builders can find the show.

Oren Michaels: https://www.linkedin.com/in/omichels/

Oren Michaels is the Co-founder and CEO of Barndoor AI, a company building the control plane for agentic AI to safely manage and govern how AI agents interact with corporate systems. Previously, Oren co-founded Mashery in 2006, serving as CEO until the company was acquired by Intel in 2013. A true multi-hyphenate with an Electrical Engineering degree from MIT, Oren is also a Tony-nominated Broadway and Off

Website: https://www.position2.com/podcast/

Rajiv Parikh: https://www.linkedin.com/in/rajivparikh/

Sandeep Parikh: https://www.instagram.com/sandeepparikh/

Email us with any feedback for the show: sparkofages.podcast@position2.com