Enterprise Agents Need Trusted Data and Scalable Governance

One thing that continues to stand out in my conversations with enterprise customers is how quickly the discussion around AI agents is evolving. The excitement around building agents is real, but the bigger questions are now about scale, trust, governance, and how these agents actually work across complex enterprise environments. That's what I find particularly interesting about the shift Dinesh Nirmal is talking about. Building agents is getting easier. Operating them reliably at enterprise scale is where the real work begins. And this is exactly why I think #IBMTechXchange 2026 is one you really don't want to miss! With agentic AI moving so quickly, I can't think of a better time for #developers, #architects, and technology leaders to come together, challenge ideas, get hands-on, and learn from each other. Atlanta, October 26–29. If enterprise AI is on your radar, make this one a priority! Link to register- https://lnkd.in/gC6qFCpH #IBMTechXchange #AgenticAI #EnterpriseAI

Very insightful article on software supply chain risks in the age of AI. It focuses primarily on open source software, and many of the same software supply chain risks also exist in proprietary software—albeit with different visibility and remediation challenges.

Getting a software solution live is an important milestone. But it’s not where the work ends. Once a system is deployed, it still needs to be monitored, supported, maintained, improved, and adapted as users and organizational requirements change. For SLED organizations, that ongoing lifecycle matters just as much as the initial implementation. A technology solution should be able to keep working as the environment around it evolves. Because successful delivery isn’t just about getting software live. It’s about creating continuity beyond the launch. #SLED #GovTech #GovernmentTechnology #SoftwareImplementation #DigitalTransformation #TechnologySolutions #Scorpbit

If you’re trying to get your team “AI ready,” don’t start with tools. Start with trust. Here’s the 60-second script I’d use with my team: “AI is going to change how we work here. We can treat that as a threat… or as augmentation. We’re choosing augmentation. This is not about replacing people. It’s about removing busy work and raising our impact. No one will be punished for experimenting. We’ll learn in the open. Each week, we’ll pick ONE task to hand to AI, measure the time/effort saved, and share what worked, so we get better together.” Then I move teams through 3 stages: 1) Awareness → demystify it using our real work 2) Adoption → reduce friction (ex: “write emails faster” is adoption) 3) Redesign → re-architect the workflow end-to-end (this is where the real leverage appears) Example (Sales): don’t stop at “write emails faster.” This next stage begins redesign: Have AI analyze past wins, score open deals for propensity-to-win, suggest next best actions, and draft a proposal with a human in the loop. My favorite starting question is simple: What’s one busy-work task you’d happily hand to AI tomorrow (and what function)? If you comment on task/function, I’ll reply with some starter thoughts on redesign. #AI with purpose

Platform engineering started as a way to simplify complexity. Today, it is evolving into a crucial mechanism for organizations to govern that complexity. As software delivery becomes more automated, the challenge has shifted from merely enabling teams to move faster to ensuring that security, compliance, reliability, and cost controls can operate continuously at the same speed as the systems they govern. I recently shared insights with Enterprise Times on why I believe platform engineering is entering a new phase—one characterized by real-time governance. Feel free to read the full conversation here: https://lnkd.in/gy7gu7DV #platform #engineering #ai #agents #governance #harness

AI agents crossing their boundaries isn't a theoretical risk anymore; it's happening now. https://lnkd.in/dKdJzFYt Recent disclosures show autonomous agents interacting with systems they were never supposed to touch. It’s a massive wake-up call for engineering teams. We need to realize that model governance isn't enough. We need strict engineering governance. When an AI moves from just suggesting code to actually executing actions, you can't just trust the model to behave. You have to lock down the environment. At a minimum, that means treating agents like untrusted users: sandboxing execution, enforcing least-privilege access, and keeping a human in the loop for high-impact decisions. The real challenge of the next few years isn't building smarter agents. It's engineering the guardrails that keep them from breaking things.