What happens when your AI agents start making decisions faster than your security team can even see them?

In this episode, I sit down with Sunil Agrawal, Chief Information Security Officer at Glean, to unpack a shift already underway in enterprises. With predictions that 40 percent of enterprise applications will include autonomous AI agents by the end of 2026, we are moving from human-led workflows to machine-to-machine interactions at a scale most organizations are not fully prepared for.

Sunil brings a rare perspective, blending more than 25 years of cybersecurity experience with an inventor's mindset shaped by over 40 patents. What stood out to me in our conversation is how quickly the traditional security model is becoming outdated. As he explained, "autonomous agents break those assumptions because they operate across tools, varying permissions and data sources with alarming speed and autonomy." This creates what he calls the "autonomy gap," in which the CIO's drive for speed collides with the CISO's need for visibility and control.

We explore how that tension is playing out in real organizations today, and why so many are already falling behind. Nearly half of businesses still lack the AI-specific controls needed to prevent untraceable incidents, and the risks are not always what you might expect. Sunil argues that the first major rogue-agent incident is unlikely to be a malicious attack. Instead, it will come from confusion: a well-intentioned system taking the wrong action in the wrong context, with consequences that ripple across the business.

The conversation then turns practical. Sunil breaks down his AWARE framework, a structured way to introduce real-time guardrails that evaluate intent, context, and risk before an agent takes action. Rather than relying on static policies, this approach focuses on continuous runtime enforcement, where systems are constantly assessed based on behavior rather than assumptions.

What I found particularly valuable is how this moves beyond theory into something leaders can act on today. From starting with tightly scoped use cases to investing in full observability, this episode offers a clear roadmap for balancing innovation with accountability. As Sunil put it, organizations that succeed will not be the ones that move fastest, but the ones that prove trust at scale.

So how do you embrace the productivity gains of autonomous AI without opening the door to invisible risk, and are your current security models ready for a world where the "user" is no longer human?

Useful Links

• Connect with Sunil Agrawal on LinkedIn
• Learn more about Glean
• Follow Glean on LinkedIn

Visit the Tech Talks Network Sponsor NordLayer Browser

What does it actually take to move AI from experimentation into something a business can depend on every single day?

Recording live from the show floor at Qlik Connect in Florida, I sat down with Qlik CEO Mike Capone to cut through the noise and get to the reality behind enterprise AI in 2026. Because while the headlines are still dominated by rapid innovation and new capabilities, many organizations are quietly facing a different challenge. They are struggling to turn AI ambition into measurable outcomes.

In our conversation, Mike shares what he is hearing from customers around the world and why so many companies remain stuck in cycles of pilots and proof of concepts. We talk about the growing pressure from boards and leadership teams to move faster, and why that urgency is often leading to what he calls a "ready, fire, aim" approach that fails to deliver real business value.

We also explore one of the biggest themes emerging at Qlik Connect this year. The shift toward agentic AI. But rather than focusing on the hype, Mike breaks down what this actually means inside a real enterprise workflow, where insights are not just generated but turned into decisions and actions. He also explains why getting the data foundation right is no longer optional, and how poor data quality can quickly turn AI from an opportunity into a risk.

From data trust and governance to the challenges of operating across increasingly complex regulatory environments, this episode offers a clear view of what it takes to build AI systems that are reliable, scalable, and grounded in real business context.

So as organizations look ahead to the next 12 to 24 months, what will separate those that successfully operationalize AI from those that remain stuck in pilot mode? And are we focusing too much on building more AI, rather than building better AI?

Join me for a candid conversation from the heart of Qlik Connect, and let me know where you stand on this shift. Are you seeing real progress, or are the same challenges holding things back?

How are brands supposed to deliver AI-powered customer experiences when their data is scattered across systems that were never designed to work together?

In this episode, I sit down with Peter Bell, VP EMEA Marketing at Twilio, to unpack one of the most important AI topics that still does not get enough attention outside technical circles, Model Context Protocol, or MCP. While many conversations about AI remain stuck on model hype, chatbots, and the latest product launch, Peter brings the discussion back to something far more practical. If businesses want AI to deliver real outcomes in customer service, marketing, and brand engagement, they first need a reliable way to connect large language models to the right data, in the right systems, with the right controls in place.

That is why this conversation matters. Peter explains how MCP could become one of the biggest unlocks for enterprise AI by creating a standard way for LLMs to access information across fragmented tools like CRM platforms, marketing systems, and other business applications. Instead of forcing every company to build custom integrations from scratch, MCP creates a more consistent path for connecting models to the context they need. For me, that is where this episode really earns its place, because it moves the AI conversation away from vague ambition and toward the plumbing that actually makes useful AI possible.

We also talk about why first-party data remains so important, especially as businesses try to create customer experiences that feel seamless, personal, and trustworthy. Peter makes the point that public models may be useful for general knowledge, but brands cannot rely on generic internet-trained systems to solve precise business problems. If you want AI to support travel bookings, customer service, or commerce journeys, you need specific data, strong governance, and a much clearer understanding of the problem you are trying to solve. That sounds obvious, but it is still where many AI projects fall apart.

Another part of our conversation focuses on trust, which feels especially relevant right now. From scams and impersonation to consumer fatigue and poor automation, brands are under pressure to move faster without losing credibility. Peter shares how Twilio is thinking about branded calling, RCS, conversational AI, and voice experiences that feel modern without becoming intrusive or robotic. We also discuss why too many companies still automate too broadly, too quickly, without defining the actual use case first.

What I enjoyed most here was Peter's balanced view. He is optimistic about where AI is heading, but he is also realistic about the work still required to get there. This is not a conversation about AI magic. It is about data access, governance, trust, brand experience, and the standards that may quietly shape the next phase of AI adoption far more than the flashy headlines.

So if you have been hearing more people mention MCP and wondering why it matters, or if you are trying to understand what needs to happen before enterprise AI can move from promise to practical value, this episode will give you plenty to think about. Is Model Context Protocol the missing layer that finally helps AI connect with the real world of business data?