AWS expands Connect to agents

AWS just did something more ambitious than adding another chatbot feature. It took Amazon Connect — long known as a cloud contact-center product — and recast it as a broader family of business agents. The move matters because it shifts AWS up the stack. Instead of just selling infrastructure and models, AWS is now packaging domain-specific AI workers for actual business jobs. That change landed on April 28 at the company’s What’s Next with AWS event. ### What actually changed? Amazon Connect is no longer being pitched as one thing. AWS now describes it as a portfolio of four agentic AI solutions: Connect Decisions for supply chain, Connect Talent for hiring, Connect Customer for customer experience, and Connect Health for healthcare workflows. That is a real category expansion, not a branding tweak — three of those four go well beyond the old “contact center” box. (aws.amazon.com) ### Why is that a bigger deal than it sounds? Because AWS is borrowing Amazon’s internal operating playbook and turning it into software. The pitch is basically: Amazon spent decades building systems for large-scale hiring, customer operations, logistics, and healthcare-adjacent workflows — now AWS wants to sell AI versions of those operating patterns to other companies. That is different from saying “bring your own model and build an app.” It is much closer to “here is the workflow, the agent, and the business logic.” (aws.amazon.com) ### What’s the supply-chain product? Connect Decisions is the clearest example. AWS says it is generally available now, and positions it as a planning and intelligence system for supply-chain teams. The product combines more than 25 specialized supply-chain tools with AI “teammates” that learn from user decisions and help teams move from reactive firefighting to proactive planning. That is a pretty direct attempt to productize Amazon’s own logistics muscle. (aws.amazon.com) ### What’s the hiring product? Connect Talent is the other big new piece, and it is in preview. AWS says it can run structured voice interviews, administer assessments, and generate scored candidate evaluations for recruiters. Candidates can interview around the clock from any device, while human hiring teams review transcripts, scores, and summaries instead of doing every first-round screen themselves. In plain English — AWS wants AI agents doing the repetitive top-of-funnel hiring work. (aws.amazon.com) ### Where does customer service fit now? Customer service is still there, but it has been reframed. The old Connect heartland becomes Connect Customer, where AWS is leaning harder into AI agents that can talk directly with end users over voice and chat, assist human agents in real time, and now plug into external tools through MCP support. That matters because it turns the system from a scripted support layer into something closer to an action-taking agent platform. (aws.amazon.com) ### Is this tied to Bedrock and OpenAI? Yes — and that is part of the strategy. AWS paired this launch with a broader expansion of its OpenAI partnership, bringing newer OpenAI models and managed agent tooling into Amazon Bedrock. The implication is straightforward: Connect becomes the workflow shell, while Bedrock supplies model choice and orchestration underneath. AWS is trying to offer the full stack — model access, agent tooling, and business application layer. (docs.aws.amazon.com) ### So what is AWS really trying to become? Basically, more than the cloud where other people build AI products. This launch says AWS wants to be a direct vendor of production-ready business agents. That puts it closer to enterprise software companies that sell packaged workflows, but with AWS’s infrastructure and model ecosystem behind them. For buyers, the appeal is speed. The catch is lock-in — once the workflow logic, model orchestration, and operational data all sit inside one stack, switching gets harder. (aws.amazon.com) ### Bottom line? AWS did not just expand Connect’s feature list. It turned Connect into a branded agent suite for specific business functions. If you are evaluating enterprise AI agents, that means AWS is no longer just the place you host them. It increasingly wants to be the place you buy them. (aws.amazon.com 1) (aws.amazon.com 2)