Amazon just figured out what to sell next.

On Tuesday at the What's Next with AWS event in New York, Amazon Web Services announced four new agentic AI products under the Amazon Connect name. The original Connect, which has been a contact center tool for a decade, gets rebranded to Amazon Connect Customer. Three new siblings join it: Amazon Connect Decisions for supply chain and demand planning. Amazon Connect Talent for high-volume hiring. Amazon Connect Health for clinical documentation and patient coordination.

Skip the marketing language. Look at what those four products actually are.

They are Amazon. The whole company. Productized.

Selling the brain that moves a billion packages

Amazon Connect Decisions is not a generic supply chain forecasting tool. It is the operational logic Amazon uses to run its own logistics network, repackaged for everyone else. Thirty years of demand prediction, inventory routing, fulfillment optimization, the literal brain that decides which warehouse ships which package to which doorstep on which truck. Now available as an enterprise SaaS subscription.

Amazon Connect Talent is the same trick applied to hiring. Amazon hires hundreds of thousands of warehouse and delivery workers every year. They built voice-based AI interview tools to do it at scale, scoring candidates on skills rather than resumes. That tooling is now a product. Other companies can deploy AI agents that conduct structured voice interviews around the clock and run candidate assessments before any human recruiter is involved.

Amazon Connect Health takes the same playbook to clinical documentation, building on Amazon's existing healthcare bets through One Medical and AWS HealthLake. The pattern is identical: take a discipline Amazon has been forced to industrialize internally, extract the AI tooling, sell it.

The strategy nobody is calling out

OpenAI sells tokens. Anthropic sells safety. Google sells search-shaped AI. Microsoft sells Office-shaped AI. Meta sells personality-shaped AI. Every major AI company is selling some version of the same primitive: a model that does something cognitive on demand, charged by the call.

Amazon is not selling that.

Amazon is selling outcomes. The outcome is hiring 5,000 warehouse workers next quarter. The outcome is rerouting freight when a port closes. The outcome is documenting 200 patient encounters per shift. The model running underneath is whatever AWS thinks works best, often Anthropic's Claude through Bedrock, but the customer does not pay for the model. The customer pays for the result.

This is a fundamentally different go-to-market than the entire rest of the AI industry. It maps to AWS's original insight from 2006: nobody wants to buy compute, they want to buy the application that runs on compute. AWS sold compute anyway and made $100 billion a year doing it. Now they are skipping the compute layer entirely and selling the applications.

Who this kills

If Amazon Connect Talent works, it is an extinction event for a layer of HR tech. Pymetrics, HireVue, Workday's recruiting suite, the entire structured-interview SaaS category. None of them have Amazon's hiring volume, the proprietary data that comes from interviewing millions of warehouse candidates, or the AWS distribution channel that lets enterprise IT departments add it to their existing contract with one signature.

If Amazon Connect Decisions works, it is a direct shot at SAP's Integrated Business Planning, Oracle's supply chain suite, Blue Yonder, o9 Solutions, and a long tail of demand forecasting startups. SAP charges seven figures for software customers spend years implementing. Amazon will sell something that works on AWS day one, priced like a utility, backed by the operational data of the world's largest logistics company.

Health is the most contested. Epic, Oracle Cerner, and Microsoft's Nuance DAX dominate clinical documentation. Amazon comes in late. But Amazon also has more raw compute, more model access through Bedrock, and a One Medical fleet of clinics to test on. Late does not mean lost.

The take

Watch what Amazon is doing here, because it answers a question the rest of the AI industry has been pretending was not a question. What happens after the model layer commoditizes?

The answer Amazon is betting on: the company with the deepest operational data wins. The model is interchangeable. The training data, the workflow knowledge, the integration into a real-world business process, those are the moat.

OpenAI does not have that data. Anthropic does not have that data. Google has it for search and ads. Meta has it for social. Microsoft has it for productivity software. Amazon has it for retail, logistics, hiring, fulfillment, and increasingly healthcare. The list of operational domains where Amazon has more proprietary data than anyone else is long, and Amazon is now systematically turning each one into a product.

The era where AI companies competed on benchmark scores is ending. The era where they compete on which real business they can replace is starting. Amazon is the first major player that built an entire business empire before AI was a thing, and is now sitting on the operational data that empire generated.

Everyone else is selling tools. Amazon is selling the work the tools were supposed to do.