Identity Decoded | The Identity Security Podcast

Why Security Follows What’s Worth Protecting: An Investor on Market Timing and the Agentic Future

In this episode of Identity Decoded, hosts Roy Akerman and Rob Ainscough sit down with Mark Kraynak, a founding partner at Acrew Capital who spent 13 years at Imperva before becoming a cybersecurity investor, and, as it turns out, a trained poet. Mark argues that security has never been built in from the start of a new technology wave, and that new technologies like MCP and agentic AI, are no different. The conversation covers why agents represent new security challenges, why he splits identity into three buckets (people, machines, and AI), and whether AI will become its own security "estate." It ends with practical advice for buyers, and a surprising haiku.

🎧 Episode Highlights

[0:00] Intro

[02:16] Mark's path from an engineering degree and poetry grad school, to teaching writing, to 13 years at Imperva and a career as a VC.

[08:15] What has changed in identity: why passkeys don't work well for agents, and cloud-native applications.

[11:46] Why security is never built in from the start: market forces punish the company that slows down for security, and Mark's reframe that "that's why you have a job."

[14:00] A friendly debate with Rob on whether agents and MCP should be treated as unique problems, and why nobody buys security for its own sake.

[17:36] Three kinds of identity: people, "dumb" machines, and smart machines (AI), and the case for a unifying layer that sees across all three.

[28:30] Mark's haiku about least privilege, Rob's poem about access reviews, and the story of the customer who got 25,000 pages to review.

🔑 Key Takeaways:

Takeaway 1: Security is almost never built in from the start, but that is how markets work. With every new wave, from cloud native to mobile to MCP, security teams hope this time it will be different, and it isn't, because the company that slows down for security loses. Mark's advice to practitioners is to stop asking why it wasn't done right and focus on letting these technologies deliver value without breaking them.

Takeaway 2: Passkeys solve the last problem, and agents are the next. Mark says passkeys can reduce phishing of humans, but acting on someone's behalf is "almost impossible" with agents, and he believes people, machines, and AI will each need different identity stacks, with a layer on top that can see across the boundaries.

Takeaway 3: Be practical and move fast. Mark advises practitioners to put controls at the tools an MCP server calls using existing tools, ask current vendors what they can do, and understand what's sensitive about what agents are doing. In his words, MCP took about a year to adopt, compared with two decades for TCP.

👤 Guest Spotlight:

Mark Kraynak, Founding Partner, Acrew Capital

Mark Kraynak is a founding partner at Acrew Capital, where he invests across cybersecurity, data, and enterprise technology, with investments including Cato Networks, Protect AI, and Silverfort. An engineer by training with a second degree in poetry, he spent 13 years at Imperva in application security through its IPO, then served as an entrepreneur in residence before deciding to invest full-time. He became a founder of Acrew when the firm was reconstituted in 2019. In this episode, he shares what two decades in security have taught him about timing, new technology, and where identity goes next.

Stay connected

https://www.silverfort.com

https://www.linkedin.com/in/rob-ainscough

https://www.linkedin.com/in/roy-ak