Twelve thousand people. A hundred countries. One question everybody in Las Vegas this week was trying to answer: how do we actually use AI safely and well, instead of just talking about it?
That’s the question Clear Winds went to find answers to at AI4 2026, and we brought a mic. Our CEO Stan Sargent, VP of Engineering Michael Gray, and Director of Development James Hammack spent the week moving between keynotes, security sessions, and the vendor hall — and host Jay Bradford caught up with all three live from the show floor. Here’s what stood out.
The Godfather of AI Sets the Tone
The biggest moment of the conference was Geoffrey Hinton’s keynote — the Nobel Prize-winning “Godfather of AI” whose foundational work on neural networks helped make the last decade of machine learning possible. Hinton didn’t play cheerleader or doomsayer. He came across as a realist: clear-eyed about AI’s potential and just as clear-eyed about where it can go wrong.
His central point echoed something the industry has heard before, just in a new context — he compared today’s lack of AI governance to the early, unregulated days of the tobacco and oil industries. The lesson wasn’t “slow down.” It was “build the guardrails now, before the mistakes get expensive.”
That framing showed up again and again throughout the conference: the goal isn’t to hard-stop AI out of fear, it’s to steer it deliberately.
Why This Technology Shift Feels Different
Michael Gray, who spent much of the conference in security-focused sessions, put his finger on why AI feels harder to plan for than past tech shifts. Even a disruptive shift like the VMware virtualization wave was still fundamentally deterministic — the same input reliably produced the same output. AI doesn’t work that way. It’s non-deterministic, which means the old playbooks for managing risk and change don’t fully apply.
That single distinction is why so many organizations are struggling to translate “we should use AI” into a real plan.
Treat Your AI Agents Like New Hires
One of the most practical takeaways from the week: stop thinking of AI agents as software and start thinking of them as identities.
Just like onboarding a new employee, an AI agent needs defined access, permissions, and logging — and it needs to be evaluated the same way you’d evaluate a person’s performance and judgment. Michael Gray put it simply: both AI and “Tom in the office” make mistakes. The fix isn’t different oversight, it’s the same oversight, applied consistently. Add data sovereignty questions on top — where is your information actually going once it hits an AI tool? — and it becomes clear why governance has to be a day-one conversation, not a bolt-on.
The Mistake That Sinks Most AI Projects
Stan Sargent didn’t mince words about where AI initiatives go wrong: companies rush. They see the potential, want a quick win, and start building before they’ve defined what they actually need — what workflow they’re automating, what data feeds it, where that data even lives.
Messy, undefined data is the single biggest reason AI projects fail to deliver value. It’s also exactly why Clear Winds starts every engagement with an AI readiness assessment: mapping data sources, workflows, and goals before writing a line of automation. Projects built on that foundation tend to stick around. Projects that skip it tend to produce something that technically works but isn’t actually useful.
Humans Are Still Very Much in the Loop
Six months ago, the conversation was full-automation hype. This year, nearly every session Clear Winds attended circled back to the human element — and for good reason. James Hammack described it as gatekeeping: give an agent a task, but require it to check in with a human before anything expensive, irreversible, or potentially unsafe. That means teaching the agent what “risky” even looks like, since it has no instinct for that on its own.
Used this way — as a tool wielded properly, not a replacement for judgment — AI agents and automation are already showing up in real deployments, from customer support to internal reporting to, as one Amazon session covered, global-scale agent rollouts.
The Bottom Line About AI4 2026
AI4 2026 reinforced something Clear Winds has been telling clients for a while: the winners in this next phase won’t be the companies that move fastest, they’ll be the ones that define their data, their workflows, and their governance clearly enough that AI can actually be trusted to run on top of them.
Want to know where your organization stands? Reach out for an AI readiness assessment and find out what it would take to bring AI into your workflows safely and effectively.
Want the full conversation? Listen to the live AI4 2026 recap episode of The IT Directors Podcast.
