There’s a telling statistic buried in AWS’s re:Invent 2025 Partner Advantage keynote that explains why the company spent most of the week talking about its ecosystem rather than just its own technology. For every dollar of AWS services a partner deploys at a customer site, that partner realises $7.13 in associated business. If AWS achieves its ambition of becoming a $400 billion business, that’s somewhere between $2 trillion and $3 trillion flowing through the partner channel.
That’s not an accident. It’s architecture.
Dr Ruba Borno, AWS’s Vice President of Global Specialists and Partners, delivered what amounted to a blueprint for how Amazon intends to commercialise agentic AI at enterprise scale, and the plan hinges almost entirely on partners doing the difficult work of implementation, integration, and ongoing management. The message wasn’t subtle. Customers experimenting with AI pilots need to move into production, and AWS believes the fastest path there runs through its certified partner network.
The company announced three new agentic AI competency categories (applications, tools, and consulting services) specifically to help customers identify which partners have demonstrated actual capability rather than just enthusiasm. Partners who achieve these specialisations get 50% more marketing development funds and a badge in AWS Marketplace. It’s a carrot designed to create a credible signal in what’s becoming an increasingly noisy market.
But the real work is happening inside Amazon Bedrock AgentCore, AWS’s platform for building, deploying, and securing AI agents at scale. The company is positioning AgentCore as the infrastructure layer for autonomous systems, complete with isolated execution environments, real-time policy controls, and continuous quality evaluations. Each agent session runs in its own MicroVM with separate CPU, memory, and file systems, a design choice that addresses one of the thorniest problems in agentic systems: how to prevent reasoning leaks and cross-session contamination when agents are making decisions autonomously.
AgentCore’s security model reflects AWS’s awareness that enterprises won’t hand over decision-making authority to systems they can’t audit or constrain. The platform includes deterministic policy controls using the Cedar open-source language, allowing organisations to define exactly what actions agents can take with specific tools and data. AgentCore Evaluations automates the previously labour-intensive process of inspecting agent behaviour for correctness, helpfulness, and potential harm, a capability that shifts quality assurance from a post-deployment scramble to a continuous monitoring function.
The customer stories AWS highlighted reveal what production-ready agentic AI actually looks like in 2025. Condé Nast, the 115-year-old media company behind Vogue, The New Yorker, and Vanity Fair, rebuilt its entire digital infrastructure on AWS and Databricks to deliver personalised experiences to over a billion people globally. Their transformation took them from fragmented data across 22 publications and 12 markets to a unified content platform that can syndicate stories across languages using AI-based translation. The Met Gala event they streamed in May 2025 generated 522% more engagement than the Super Bowl, a metric that speaks to what’s possible when media companies treat data strategy as seriously as editorial strategy.
Mater Dei, a major Brazilian healthcare platform, worked with AWS partner A3Data to build 12 autonomous AI agents across their entire revenue cycle, from contract analysis to billing. The result was a 517% return on investment and procedure authorisations that dropped from two business days to 40 minutes. Toyota Motor North America partnered with McKinsey, Deloitte, and AWS to integrate customer insights directly into supply chain planning, realising $1 billion in business value while reducing inventory by 60% and improving customer satisfaction by 15%.
These aren’t pilot projects. They’re production systems processing millions of transactions, handling regulatory compliance, and making decisions that directly affect revenue and customer experience.
AWS’s marketplace strategy has evolved to match this shift toward complex, multi-vendor solutions. The company launched multi-product solutions in AWS Marketplace, allowing customers to procure end-to-end offerings that combine software products, SaaS agents, and professional services from multiple partners through a single transaction. It’s a recognition that the most difficult enterprise problems require orchestrated responses from specialists with different expertise, and procurement friction has historically been a barrier to that kind of collaboration.
Express Private Offers, now generally available, uses AI to deliver custom pricing to customers in minutes based on predefined rate cards and qualification criteria. It’s a self-service mechanism designed to accelerate the long tail of smaller deals that would otherwise require weeks of email negotiation. AWS Marketplace also introduced Agent Mode, a conversational discovery experience that searches over 30,000 solutions using multiple AI models (including Anthropic’s Claude and Amazon Nova) to generate personalised comparisons based on uploaded requirements documents.
The marketplace now supports localised procurement in multiple currencies with region-specific tax treatments and local bank support, a detail that matters considerably more than it sounds. Global enterprises operate across jurisdictions with different regulatory requirements, and the ability to process transactions in yen or euros with proper local invoicing removes genuine friction from international deployments.
Salesforce is launching AgentForce 360 for AWS exclusively through AWS Marketplace early next year, built entirely on the Salesforce platform running on AWS with access to foundation models through Amazon Bedrock. The integration keeps customer data within the Salesforce-AWS trust boundary, provides transparency through the Atlas Reasoning Engine showing how agents think and plan, and grounds responses in the customer’s own data through the Prompt Builder. It’s the kind of partnership that demonstrates how AWS is positioning itself as infrastructure for AI systems built by other companies, not just its own offerings.
CrowdStrike’s Falcon next-gen SIM now includes automated discovery and deployment across multi-account, multi-region AWS environments with near real-time security findings through integration with Amazon EventBridge. The partnership reflects AWS’s strategy of making security integrations seamless enough that customers don’t have to choose between best-of-breed security tools and operational simplicity.
AWS Transform, the company’s AI service for assessing, migrating, and modernising customer workloads, has already saved customers over 800,000 hours of manual effort since becoming generally available earlier this year. The service now supports composability, allowing partners like Pega, Capgemini, and Accenture to integrate their own tools, agents, and knowledge bases directly into AWS Transform workflows. Western Union used Accenture’s financial services knowledge base agent to modernise their retail money order application, analysing 2.4 million lines of mainframe code and achieving a 2x increase in processing speed.
There’s a consistent pattern across these announcements: AWS is building infrastructure and tooling, then relying on partners to apply domain expertise, industry knowledge, and implementation capacity to turn that infrastructure into customer outcomes. The company launched Experience Based Acceleration (EBA) certification for partners in February 2025, codifying lessons learned from thousands of customer migrations with companies like BMW, SAP, Boeing, and Samsung. Over 70 certified partners worldwide are now scaling EBA to customers, bringing structured change management practices to what’s often the messiest part of cloud adoption.
Matt Garman, AWS’s CEO, was explicit about the opportunity size during his appearance at the partner keynote. AWS is a $132 billion business growing at 20% year over year, with that growth rate accelerating. The company is planning for $300 billion to $400 billion in annual revenue, which means massive capital investment in infrastructure and, more importantly for partners, massive expansion in the services, integration, and managed offerings that customers will need on top of that infrastructure.
The South African context for AWS’s global partner strategy is worth noting, particularly given the company’s long-term commitment to the region. AWS has maintained a development centre in Cape Town since 2004, and the AWS Skills Centre there has become a model for how cloud providers can address skills gaps while building local partner ecosystems. The centre has trained thousands of people on cloud fundamentals, security, and increasingly on AI and machine learning modules, creating a pipeline of certified professionals who can support partner organisations delivering solutions to African enterprises.
What’s interesting about AWS’s re:Invent 2025 partner messaging is what it doesn’t include: any pretence that AWS alone can handle the complexity of enterprise AI adoption. The company is betting that the path to production runs through specialised partners who understand specific industries, regulatory environments, and organisational change requirements in ways that a cloud provider simply can’t. The competencies, marketplace integrations, and security frameworks AWS announced are all designed to make it easier for those partners to build, sell, and support solutions without getting tangled in procurement bureaucracy or security compliance headaches.
The customer examples AWS shared demonstrate that agentic AI transformation is already happening in revenue cycle management, supply chain planning, content personalisation, and security operations. But those successes required partner organisations with deep domain knowledge and change management capability, not just access to AWS infrastructure. Mater Dei’s 517% ROI came from A3Data understanding Brazilian healthcare compliance and revenue cycle complexity. Toyota’s $1 billion in value required McKinsey’s supply chain expertise and Deloitte’s implementation skills working in concert with AWS’s technical foundation.
AWS is building the infrastructure and marketplace mechanisms to enable partners to scale that kind of outcome delivery. The company’s $2 trillion to $3 trillion opportunity projection assumes partners can consistently replicate what worked for Condé Nast, Mater Dei, and Toyota across thousands of enterprise customers in dozens of industries. That’s a bet on partner capability, not just technology superiority. Whether the AWS partner network has the depth, breadth, and execution capacity to deliver at that scale will determine whether AWS’s agentic AI strategy becomes the infrastructure layer for enterprise autonomy or just another well-architected platform waiting for adoption to catch up to ambition.