KBKAST

In this episode, KB sits down with Mark Thomas, IT governance and risk veteran, ISACA Hall of Famer and president of Escoute Consulting, to pull apart a problem a lot of boards haven’t clocked yet – the gap between owning an AI policy and being able to prove it controls anything.

They get into the Air Canada chatbot case and what it says about accountability, why the honest board test is “would anyone notice if this was violated,” and how the risk changes once agents move from recommending to executing. Mark makes the case that human in the loop only counts when the human has the expertise, the authority and the time to say no. He also explains why only a small fraction of organisations have ever tested their ability to shut a system down, and why accountability never transfers to the vendor.

A practical, occasionally uncomfortable conversation for anyone putting AI into production.

——

About Mark:
Mark Thomas is a globally recognised expert in governance, risk management, and digital trust, with more than two decades of experience advising organisations operating in complex, regulated, and rapidly evolving environments. His work sits at a critical intersection where strategy, governance, and execution meet.
He works directly with boards and executive leadership to:

Strengthen oversight and accountability

Improve confidence in decision-making

Navigate emerging technologies and digital risk

Align governance with real-world execution

Mark is known for his ability to translate complex issues into clear, practical insight, helping leaders move from uncertainty to informed, defensible decisions.

Keywords: AI governance, AI risk, board accountability, agentic AI, human in the loop, kill switch, digital trust, AI policy, ISACA, Mark Thomas, Escoute Consulting, KBKast, enterprise AI, AI compliance, EU AI Act, shadow AI, Air Canada chatbot

Most security vendors are shipping AI that treats capability as authority. Sean Duca thinks that’s the mistake customers are already paying for. Back on KBKast for a third time, now as co-founder and CEO of getsoteria.ai, Sean makes the case that confidence and permission are two separate gates, and that an autonomous SOC needs both before it touches anything.

He walks Karissa through governed autonomy: the machine acts on its own, but only inside a boundary it can’t set or cross. A bank teller and a self-driving car do the heavy lifting on the difference between a human in the loop, on the loop, and off it entirely. He gets specific about shadow mode, the 85% agreement bar his threat hunters have to keep clearing, and why his system fails closed and hands everything back to a person the moment it hits a wall.

About Sean:
Sean Duca has spent 25+ years in cybersecurity, most of it advising boards and security leaders across Asia Pacific. He’s now co-founder and CEO of getsoteria.ai, following senior roles as CTO for Customer Experience at Cisco (APJC), VP and Regional Chief Security Officer at Palo Alto Networks (APJ), and CTO for APAC at Intel Security. He’s a published author on cybersecurity and calls Singapore home.

Keywords: autonomous SOC, AI governance, governed autonomy, agentic AI security, confidence vs permission, human in the loop, AI access control, SOC automation, shadow mode, security operations, AI authority model, CISO, board risk, Sean Duca, getsoteria.ai, KBKast

The old security math is broken. Teams used to get roughly 60 days between a vulnerability going public and attackers using it. Harman Kaur, Tanium’s CTO, explains why that number has flipped to negative, and what it means when the patch you need does not exist yet.

This conversation is about the difference between automation and transformation. Running the same broken process faster is not progress. Harman makes the case for rebuilding security from first principles: changing org charts, retiring tools people have relied on for years, and moving humans out of triage and into judgment.

She also gets specific about trust. In her view, trust now sits in the underlying data, because a confident AI acting on stale or incomplete data becomes dangerous at scale. Harman & KB get into the boardroom scramble around Mythos, why there is no single tool that solves it, and why the vendors closest to the data will be the ones left standing.

About Harman:
As Chief Technology Officer, Harman Kaur leads Tanium’s technology strategy, product management, AI and automation roadmap, and strategic technology partnerships.

Harman brings more than a decade of combined experience across the United States Air Force and Tanium. She continues to serve as a Cyber Officer in the U.S. Air Force. At Tanium, Harman has held senior roles across the customer organization, R&D, and most recently led the company’s AI and Autonomous Endpoint Management strategy as head of AI before stepping into the CTO role. Harman received an MBA from the University of Southern California and a BS in Information Systems from Hawaii Pacific University.

Keywords: cybersecurity, AI security, vulnerability management, patch management, endpoint security, autonomous endpoint management, Tanium, Harman Kaur, security automation, AI transformation, data quality, zero day, CISO, board risk, Mythos, security operations, KBKast

Bradon Rogers, Chief Customer Officer at Island, the company that created the enterprise browser category in 2020, joins KB to unpack why the industry spent two decades bolting security around the browser instead of building it in. They dig into whether organisations are solving complexity or just relocating it, why VPNs and VDI survived so long when everybody hated them, and Bradon’s blunt read on SASE: legacy on-prem architecture shoved into the cloud. He explains why SSL inspection leaves blind spots the laws of physics won’t let you close, what zero trust misses after access is granted, and how data boundaries keep company information out of personal AI tenants without turning security into the say no police.

Bradon closes with his prediction for the next two years: non-human identities, agents working without human hands on the wheel, and agentic engineers who won’t just be developers but lawyers and doctors shepherding agents through their work.

About Bradon:
Bradon Rogers is the Chief Customer Officer at Island, where he directs the technical aspects of all customer interactions, leveraging his vast experience in cybersecurity, enterprise software, and cloud technology. Bradon’s career in cybersecurity spans over 25 years, during which he has played an executive leadership role for some of the largest firms in the industry.

Keywords: enterprise browser, browser security, Island, zero trust, VPN replacement, VDI, SASE, SSL inspection, DLP, shadow AI, data boundary, agentic AI, AI agents, non-human identity, CISO, cybersecurity podcast

KB is on the ground at SAP Sapphire 2026 in Orlando, where AI has moved beyond experimentation and into the center of enterprise decision making. In this KB On The Go episode, Maura Hameroff (CMO, Cloud ERP Private and RISE with SAP) makes the case that modernizing your core is a business transformation decision rather than a technical upgrade, and that companies still running on legacy systems carry more operational and security risk than most realize.

Then Ted Way, PhD (VP & Chief Product Officer, Business AI Product Engineering, SAP) walks through what responsible enterprise AI actually takes: a governance layer, real business process knowledge, and data you can trust. As he puts it, AI first without security first is just a faster way to a data breach. A grounded look at what it takes to move from AI pilots to AI at scale without cutting corners that come back to bite you.