TL;DR
AWS launched AI-powered business applications, Connect Decisions for supply chain, Connect Talent for recruitment, joining Connect Health from March, entering the $300B SaaS market against Microsoft, Oracle, and Salesforce. AWS CMO Julia White says having no SaaS legacy is an advantage. But Amazon’s previous application products (WorkMail, Chime, WorkDocs) all failed and were shut down. The bet is that AI agents replace applications rather than augment them, and that Amazon’s operational expertise in logistics and hiring gives it a moat incumbents lack.
Amazon Web Services announced a set of AI-powered business applications on Tuesday that move the company from selling cloud infrastructure to selling the software that runs on it, a strategic shift that puts Amazon in direct competition with Microsoft, Oracle, and Salesforce in the $300 billion software-as-a-service market. The new products, Amazon Connect Decisions for supply chain management and Amazon Connect Talent for recruitment, join Amazon Connect Health, a healthcare agent platform launched in March, to form an expanding portfolio of AI-native business tools built on the same agentic architecture. AWS also renamed its flagship call centre product from Amazon Connect to Amazon Customer Connect, a rebranding that signals the company views its contact centre business, which hit a $1 billion annualised revenue run rate in 2025, as the foundation of a broader enterprise software strategy rather than a standalone product. “We don’t have a big legacy of SaaS or, frankly, a franchise to protect,” said Julia White, AWS’s chief marketing officer. “It allows us to really embrace this agentic-first approach in a way that is going to be harder for other people.”
The products
Amazon Connect Decisions is a supply chain optimisation platform that draws on more than 25 internal tools Amazon built for its own logistics operations, including the SCOT foundation models that power Amazon’s demand forecasting. The product uses AI agents to generate and tune demand forecasts, triage supply chain alerts, perform root-cause analysis, and create scenario-planning spreadsheets. The target user is a supply chain planner, not a data scientist, and the product is designed to compress workflows that currently require multiple specialised applications into a single agent-driven interface. Amazon Connect Talent conducts autonomous voice-based job interviews around the clock, scoring candidates on skills rather than resumes, and is aimed at high-volume hiring in manufacturing, logistics, retail, and hospitality. The platform schedules, calls, and interviews candidates without human intervention, making it the first AWS offering where an AI agent autonomously initiates and conducts real-time voice conversations on the company’s behalf.
Amazon Connect Health, which became generally available on March 5, delivers five AI agents for patient identity verification, appointment scheduling, medical history summaries, clinical notes, and medical coding, priced at $99 per user per month. Together, the three Connect products share a common pattern: Amazon takes an operational capability it built for its own business, retail logistics, high-volume hiring, healthcare management through Amazon Clinic and One Medical, abstracts it into an AI-native product, and sells it to enterprises through AWS. The approach is not new for Amazon, which has spent decades turning internal infrastructure into cloud services, but the application layer is. The $285 billion SaaS market anxiety that rattled software stocks in February reflects a genuine uncertainty about whether AI agents will replace traditional per-seat software or simply augment it. Amazon is betting on replacement, but only in categories where it has operational expertise that incumbents lack.
The competition
The new products put AWS into a competitive posture it has historically avoided. Amazon spent years assuring enterprise customers, including companies like Netflix that competed directly with Amazon’s consumer businesses while running their infrastructure on AWS, that it would not use its cloud position to undercut their software. AWS had been reluctant to commit to a broad business application suite, occasionally launching one-off products like WorkMail for email, Chime for video conferencing, and WorkDocs for document sharing, only to wind all three down after they failed to gain traction. WorkDocs shut down in April 2025. Amazon Chime was discontinued in February 2026. WorkMail ends support in March 2027. The previous generation of AWS application products failed because they tried to compete with Microsoft on Microsoft’s terms: generic productivity tools sold to knowledge workers. The new generation competes on Amazon’s terms: specialised operational tools sold to frontline workers in categories where Amazon has built the best internal systems in the world.
Microsoft’s largest-ever enterprise Copilot deployment at Accenture, which rolled out to all 743,000 employees with an 89 per cent monthly active usage rate, demonstrates the scale of the incumbent advantage AWS must overcome in the knowledge worker market. Microsoft has 450 million enterprise Microsoft 365 users, each of whom represents a potential Copilot upsell. Oracle’s Fusion Cloud applications and Salesforce’s Agentforce platform both have deep enterprise installation bases and years of customer workflow data. AWS has none of that. What it has is the largest cloud infrastructure customer base in the world and the operational expertise of the company that runs the most complex logistics network on Earth. The strategic logic is that AI agents do not need an existing software install base the way traditional SaaS does, because the agent replaces the application rather than sitting inside it. If that logic holds, AWS’s infrastructure relationships become the distribution channel and Amazon’s operational IP becomes the competitive moat. If it does not hold, the Connect products become the next WorkMail: ambitious launches that fade when enterprise customers choose to add AI capabilities to the software they already use rather than replace it.
The signal
AWS also announced Amazon Quick, a separate AI copilot that works across Google Workspace, Microsoft 365, Zoom, and Salesforce without requiring an AWS account, a product that takes direct aim at Microsoft Copilot’s cross-application integration. The breadth of announcements at the “What’s Next with AWS” event in San Francisco signals a coordinated strategy, not a collection of experiments. Anthropic’s zero-commission Claude Marketplace for enterprise software demonstrates that even foundation model companies are now building their own enterprise app distribution platforms, modelling them on the marketplace architecture that AWS pioneered. The competitive landscape for enterprise AI is no longer a three-way fight between hyperscalers. It is a multi-front contest in which cloud providers, AI labs, and incumbent software companies are all attempting to own the layer between the model and the business workflow.
Gartner estimates that companies spent roughly $300 billion on SaaS products in 2025. The symbolic death of SaaS as a category, marked by the retirement of Europe’s largest SaaS conference brand in favour of an AI-focused successor, reflects the industry’s belief that the per-seat, per-application model is giving way to an agent-driven model where value concentrates in the orchestration layer rather than the application itself. Amazon’s bet is that the company best positioned to win that orchestration layer is the one that already runs the infrastructure underneath it. White’s acknowledgement that AWS has no SaaS legacy to protect is both an admission of past failure and a claim of strategic advantage: the company that could not beat Microsoft at productivity software is arguing that productivity software itself is being replaced, and that the replacement favours the infrastructure provider with the best operational data and the broadest customer base. Google’s full-stack agentic AI play at Cloud Next 2026, which rebranded Vertex AI into the Gemini Enterprise Agent Platform with a no-code agent builder, makes the same argument from a different direction. The three largest cloud providers have converged on the same thesis at the same moment: that AI agents will eat SaaS. The question is whether the enterprises writing the cheques agree, or whether they will do what enterprises have always done and add AI to their existing software rather than replace it.
