LAS VEGAS — Talking this week on the Amazon Internet Companies re:Invent stage, AWS government Colleen Aubrey delivered a prediction that doubled as a wake-up name for firms nonetheless considering of AI as simply one other device.
“I imagine that over the subsequent few years, agentic teammates may be important to each group — as important because the individuals sitting proper subsequent to you,” Aubrey stated through the Wednesday keynote. “They may essentially rework how firms construct and ship for his or her clients.”
However what does that appear like in follow? On her personal group, for instance, she challenged teams that when had 50 individuals taking 9 months to ship a brand new product to do the identical with 10 individuals working for 3 months.
In the meantime, non-engineers resembling finance analysts are constructing working prototypes utilizing AI instruments, contributing code in Amazon’s Kiro agentic growth device alongside engineers and feeding these prototypes into Amazon’s well-known PR/FAQ planning course of on weekly cycles.
These are a number of the particulars that Aubrey shared after we sat down along with her after the keynote on the GeekWire Studios sales space within the re:Invent expo corridor to dig into the themes from her discuss. Aubrey is senior vp of Utilized AI Options at AWS, overseeing the corporate’s push into enterprise purposes for name facilities, provide chains, and different sectors.
Proceed studying for takeaways from the dialog, watch the video under, and take heed to the dialog beginning within the second phase of this week’s GeekWire Podcast.
The ‘teammate’ psychological mannequin modifications every little thing. Aubrey attracts a transparent line between single-purpose AI instruments that do one factor nicely and the agentic teammates she sees rising — methods that take accountability for complete targets, and require a unique form of administration.
“I believe individuals will more and more be managers of AI,” she stated. “The times of getting to do the person keystrokes ourselves, I believe, are quick fading. And actually, everybody goes to be a supervisor now. It’s a must to take into consideration prioritization, delegation, and auditing. What’s the standard of our suggestions, offering teaching. What are the guardrails?”
Amazon Join crosses $1 billion. AWS’s name heart platform reached $1 billion in annual income on a run charge foundation, with Aubrey noting it has accelerated year-over-year progress for 2 consecutive years.
This week at re:Invent, the group introduced 29 new capabilities throughout 4 areas: Nova Sonic voice interplay that Aubrey says is “very near being indistinguishable” from human dialog; brokers that full duties on behalf of consumers; clickstream intelligence for product suggestions; and observability instruments for inspecting AI reasoning.
One fascinating element: Aubrey stated she’s typically stunned by Nova Sonic’s sophistication and empathy in advanced conversations — and equally stunned when it fails at primary duties like spelling an tackle accurately.
“There’s nonetheless work to do to essentially polish that,” she stated.
The ROI query will get a “sure and no.” Requested whether or not firms are seeing the enterprise worth to justify AI agent investments, Aubrey supplied a nuanced response. “I observe firms to wrestle to understand the enterprise influence,” she stated. However she stated the worth typically reveals up as eliminating bottlenecks — clearing backlogs, erasing technical debt, accelerating safety patching — reasonably than rapid income beneficial properties.
“I’m not going to see the influence on my P&L at this time,” she stated, “but when I quick ahead a yr, I’m going to have a product in market the place actual clients are utilizing and getting actual worth, and we’re studying and iterating the place I won’t have even been midway there up to now.”
Her recommendation for firms nonetheless hesitating: “If you happen to don’t begin at this time, that’s a a method door resolution… I believe it’s a must to begin the journey at this time. I’d counsel individuals get targeted, they get shifting, as a result of in case you don’t, I believe that turns into existential.”
Belief requires observability. Aubrey says firms gained’t get full worth from AI teammates if they’ll’t see how they’re reasoning.
“If you happen to don’t belief an AI teammate, you then’re by no means going to understand the complete profit,” she stated. “You’re not going to offer them the onerous duties, you’re not going to spend money on their growth.”
The answer is treating AI inspection the identical means you’d handle a human colleague: perceive why it took an motion, audit the standard, and iterate.
“You may refine your data bases. You may refine your workflows. You may refine your guardrails, after which confidently preserve iterating… the identical means we do with one another. We preserve iterating, we continue to learn, and we preserve getting higher,” she stated.
Product updates: Past Join, Aubrey supplied updates on different elements of her portfolio of Amazon’s utilized AI options.
- Simply Stroll Out, Amazon’s cashierless checkout know-how, deployed greater than 150 new shops in 2025 and will speed up subsequent yr.
- AWS Provide Chain, in the meantime, is getting a reset. “I’m going to declare {that a} pivot,” she stated, with a Q1 announcement coming round agentic decision-making for provide and demand planning.
- Additionally coming in Q1: a life sciences product targeted on antibody discovery, presently in beta.
She teased “a number of different new funding areas” anticipated to come back in early 2026.
Keep forward of the curve with NextBusiness 24. Discover extra tales, subscribe to our publication, and be a part of our rising neighborhood at nextbusiness24.com
