AWS used its April 28 livestream, What's Next with AWS, to reposition Amazon Connect from a contact-center product into a four-pillar agentic AI suite — and to confirm that OpenAI's frontier models can now be operated as managed agents on AWS infrastructure. AWS CEO Matt Garman, SVP of Applied AI Solutions Colleen Aubrey, CMO Julia White, and OpenAI leaders shared the stage at 9 a.m. PT.
A Connect family, not a contact center
Amazon Connect is being split and rebranded into four agentic applications. Connect Customer retains the existing customer-experience workload, which Amazon says handles 20 million daily interactions for users including State Farm, Air Canada, U.S. Bank, and United Airlines. Connect Talent runs autonomous hiring workflows: it drafts requisitions, schedules candidates around the clock, conducts voice interviews, and scores responses against role-specific competencies. Constellation Research analyst Holger Mueller called it "the first offering that states agents will autonomously schedule, call and interview candidates."
Connect Decisions ports Amazon's own supply-chain stack to enterprise customers, drawing on 25+ specialized tools and foundation models built by Amazon's Supply Chain Optimization Technologies (SCOT) team, including the Chronos2 time-series model. Amazon says it can compress supply-chain incident resolution from "2+ weeks to minutes." Connect Health automates scheduling, documentation, and medical-coding tasks; Amazon cited usage across more than 1 million One Medical visits.
Amazon Quick goes external
The second flagship is Amazon Quick, an enterprise desktop AI assistant Amazon says it has already deployed to more than 500,000 internal users. Launch customers cited include AstraZeneca, the NFL, Jabil, 3M, BMW, and S&P Global. Quick wires into enterprise apps, data, and tools, and lets teams build custom agents, knowledge centers, and end-to-end workflows from a single surface.
Aubrey framed the design philosophy with a coined term — "humorphism" — arguing that "AI should work like a teammate, not a tool," and adding, "I need a teammate which is easy to work with. It's intuitive, it's trustworthy."
OpenAI lands inside Bedrock
The most strategically loaded announcement was Amazon Bedrock Managed Agents powered by OpenAI, now in preview. The service lets enterprises run OpenAI frontier models on AWS with a managed runtime that bundles memory, skills, identity, and compute, configurable in three steps: environment setup, runtime creation, and task submission.
Why it matters
The combined launch is AWS's clearest answer yet to Microsoft's Copilot stack and Google's Gemini Enterprise rollout. By bundling Amazon's own operational IP (SCOT, hiring, contact-center scale) with managed access to OpenAI models, AWS is selling agentic AI as packaged business outcomes rather than raw model APIs. As Garman put it on stage, "Customers are changing what is possible, iterating quickly and then inventing new processes."
