I've got this question during my webinar today: "Thank you. I would have loved to hear more about the intersection of AI security and traditional security. Perhaps importantly how does traditional security handle an enthusiastic, and powerful, business user who has vibe coded an agent that does what he wants and is now demanding it go live in prod. I suspect that is happening already and will happen a lot more as chatbots can send you right out to a sample React page (Canvas, Artifacts, etc) that can do exactly what you want (probably with fake data)" Do anyone work on this intersection? I'm curious now to hear about practical examples

I need to do a short introduction to lean six sigma for my team, and I'd like to make a PowerPoint presentation to accompany my dialogue. To make it interesting, I'm planning to make it themed around Judo and the similarities between Judo and LSS. I am really struggling to get the AI built into PowerPoint 365 to make the modifications I am asking for. @Gerard Pietrykiewicz I know you've mentioned in the past some quality AI options for nice slides... even if paid I just want it done. Any recommendations/suggestions? Thanks! Jeff

During David Barrett’s panel, I mentioned delegating calendars to AI agents but didn't expand on it much. This week, I came across a very honest video about a real-life experience with Clawdbot. It resonated with me because it matched what I’ve been noticing myself. Why AI Agents Feel Wrong (for Now) Right now, most AI agents fail in three fundamental ways: ✔️ too technical ✔️ too scary ✔️ no sense of time 💡 1. Security Has to Be YOUR Top Concern We are at a very early stage, and that’s exactly why cybersecurity is one of my core focuses for 2026. In this case, Claire took precautions that should be considered baseline, not “advanced”: ✔️ Created a separate OS user account for the agent ✔️ Gave it its own email address, not access to her personal one ✔️ Set up a restricted 1Password vault ✔️ When Clawdbot requested broad Google permissions (email, contacts, files), she pushed back and limited access to calendar viewing only These steps are not paranoia. They’re necessary—especially since Clawdbot has access to the local file system. Without boundaries, “assistant” quickly turns into “uncontained operator.” The ideal solution would combine the accessibility of consumer products with proper security boundaries, clear identity management, and reliable performance—a combination that doesn’t yet exist in the market. Right now, tools either feel powerful or safe. Not both. 💡 2. Prompting becomes critical with autonomous agents With autonomous agents, prompting is no longer just about output quality—it’s about control and who is still in charge. A small example: When Claire asked Clawdbot to email podcast guests, she didn’t explicitly say “draft an email for my review.” The agent immediately sent the emails. 💡 3. The default bias is impersonation, not assistance Clawdbot is biased toward acting as the user, not for the user. When asked to reschedule podcast guests, it emailed them as Claire, rather than as her assistant—even though Claire explicitly gave it a separate identity.

This week I had 3 developers from 2 different teams working on bugs that looked similar but not the same. They were all stuck. I'm a PM, not a developer. I can't write code or fix bugs. But I thought maybe these were connected. So I Lloaded all three into our internal AI tool to find commonality. Turns out, it had the same root cause file for all three bugs. Then I created a cross-team channel, and posted the AI findings, brought everyone together. They collaborated and fixed everything in a day-ish. What I learned - AI did the pattern recognition I couldn't do. But I still had to know to look for patterns, use the right tool, and facilitate the conversation. My question: What are you using AI for as a leader? Where is it actually helping you do your job better?

Teams on my current project are not forced to follow any specific SLA's, but there is an expectation of a timely delivery of tasks and issues that might arise. The problem is the word "timely". What does it really mean? This past weekend, some new issues were raised, and on Monday, before 10 am, they were already escalated to the VP level in my org. How do any of you deal with this type of escalation?