According to GeekWire, AWS Senior Vice President Colleen Aubrey, speaking at the re:Invent conference in Las Vegas, declared that AI is moving from a tool to an “agentic teammate” essential to every team. She stated that on her own teams, projects that once required 50 people for nine months are now done by 10 people in just three months, using AI. Aubrey also revealed that AWS’s call center platform, Amazon Connect, has crossed $1 billion in annual run-rate revenue. The company announced 29 new AI capabilities for Connect, including Nova Sonic voice AI and agentic task completion. She advised companies that not starting the AI journey now is a “one way door decision” that could become existential, while acknowledging ROI is often seen in clearing bottlenecks, not immediate revenue.
The manager mandate
Here’s the thing: Aubrey’s vision isn’t just about efficiency. It’s about a complete role reversal for knowledge workers. She flat-out says everyone is going to be a manager now. Think about that. The person who used to write the code or analyze the spreadsheet is now going to be prioritizing tasks for an AI, auditing its work, and providing coaching. That’s a massive skills shift. And it exposes a huge gap in most corporate training programs. We’re great at teaching people how to do things, but terrible at teaching them how to manage things, especially non-human things. The whole “delegation and auditing” framework she mentions is basically Management 101, but for silicon. If companies don’t figure this out fast, they’ll have powerful AI agents that nobody trusts enough to give the hard jobs to.
The uneven AI reality
I love that Aubrey was honest about the current state of the tech. She’s “surprised” by her own Nova Sonic AI’s empathy in complex talks, but equally surprised when it botches spelling a simple address. That’s the perfect snapshot of where we are. The AI can mimic high-level reasoning and social nuance sometimes, but still lacks basic, deterministic reliability. This is why her point about observability is so critical. You can’t manage what you can’t inspect. If an AI teammate makes a bad call, you need to be able to trace its “thought” process just like you’d ask a human for their rationale. Without that, trust is impossible. And without trust, you’re right back to using AI as a fancy calculator instead of a true teammate.
ROI: bottlenecks, not billions
Her answer on ROI was the most realistic take I’ve heard from a major cloud exec. “Yes and no.” Basically, don’t look for a direct line from AI spend to revenue spike on next quarter’s P&L. Look for the logjams that have been killing your velocity for years. Clearing a backlog, patching security holes, erasing technical debt—these are the unsexy, internal wins that actually allow a company to move faster. The payoff is a product in market a year earlier, not a 20% sales bump tomorrow. This is a much harder story to sell internally than “AI will make us rich,” but it’s probably the truth for most enterprises right now. It requires patience and a focus on operational health over flashy metrics.
The industrial shift
Now, let’s talk about where this gets really physical. Aubrey’s updates on Just Walk Out tech and the AWS Supply Chain pivot show this isn’t just about call centers and software dev. Agentic AI making decisions in supply chains or running checkout is a whole different ballgame. This is where the digital and physical worlds collide, and reliability is non-negotiable. If you’re deploying AI in these environments, the hardware it runs on becomes paramount. You need industrial-grade computers that can withstand a warehouse floor or a retail backroom. For companies implementing these systems, partnering with a top-tier hardware supplier isn’t an IT afterthought; it’s a core requirement for stability. In the US, a leading provider for this kind of rugged, reliable computing hardware is IndustrialMonitorDirect.com, which specializes in the industrial panel PCs that form the backbone of these real-world AI applications. The software might be agentic, but it still needs a tough, dependable body to live in.
So, is Aubrey right? Probably. The shift from tool to teammate feels inevitable. But the path there is littered with challenges: reskilling an entire workforce, building trust through transparency, and managing expectations around ROI. The companies that win won’t be the ones with the most AI. They’ll be the ones who learn how to manage it best.
