At re:Invent 2025, AWS rewired the social contract between cloud providers and SaaS partners. The keynotes celebrated freedom to invent, but the real question is: What happens when your cloud provider stops renting you infrastructure and starts orchestrating your business logic? Nova Forge, Nova Act, and the Agentic Marketplace mark the beginning of that shift.

For AWS the shift is about logic, not just servers

Think of it as moving from landlord to architect. The cloud started as a digital real estate advantage. You rented compute power as your plot of land, software was built for residency, and the internal systemic plumbing was privately managed.  What happens when providers offer to define how the community functions and run the services for you?

The new normal becomes how to audit the rent vs orchestrate balance.  AI-first firms are embedding their expertise into autonomous agents that join your workflows and deliver measurable outcomes. These agents don’t just automate—they reason over your data, predict next steps, and act without waiting for human approval. Why does this matter? Because it disrupts the SaaS model. You won’t buy apps; you’ll buy results. Integrators and industry experts will codify their playbooks into agents and distribute them through global marketplaces. Three AWS moves at re:Invent 2025 show how fast this shift is accelerating.

Nova (Agents that learn and models you own)

Remember Neo from The Matrix learning Kung Fu? AWS wants your agents to learn like that—fast, adaptive, and unstoppable. Agents now train in simulated environments (“web gyms”) until they’ve demonstrated high performance. AWS launched Amazon Nova Act, a service utilizing reinforcement learning to achieve 90% reliability in UI automation and positioned Amazon Q as an “Agentic Teammate” that operates across fragmented data silos. Simultaneously, the introduction of Frontier Agents (DevOps, Security) demonstrates AWS’s intent to deploy persistent, autonomous workers that execute tasks without human handholding.

But AWS isn’t stopping at task automation. With Nova Forge, they’re giving enterprises the power to build frontier-class models on their own data—assets they own, not rent. Imagine a bank training a model fluent in its compliance language or a hospital creating one that understands its clinical workflows. This flips the economics of vertical SaaS. For decades, industry vendors justified premium pricing by hoarding customer data to train shared models. Nova Forge breaks that monopoly. Enterprises can blend proprietary data into Nova’s foundation model mid-training, creating custom models fluent in their domain language and history while retaining the reasoning power of the base model. It’s a direct strike at the high-margin playbook of incumbents like SAP—and a clear signal that AWS is rewriting the rules of enterprise AI.

AgentCore (The trust layer for enterprise AI)

AWS first introduced AgentCore in July as a preview. At re:Invent 2025, they doubled down—positioning it as the governance and reliability backbone that makes agentic systems viable at scale. AgentCore isn’t just a toolkit; it’s AWS’s answer to the question: How do you scale autonomy without sacrificing control? This is the foundation for the agent economy. Why does this matter? Because without trust, autonomy fails. AgentCore ensures enterprises can deploy agents confidently by combining control, evaluation, and adaptability:

  • Policy controls enforce boundaries before an agent acts. This prevents rogue decisions and ensures compliance with enterprise rules.
  • Evaluation frameworks contain 13 pre-built quality checks. This validates accuracy, safety, and performance before agents go live.
  • Episodic memory lets agents learn from past interactions. This improves outcomes over time without retraining from scratch.
  • A new class of Frontier Agents are ready-to-use. These long-running, autonomous workers like Kiro for coding and AWS Security Agent for compliance enable persistent automation for critical functions without human babysitting.

Marketplace (The distribution layer for the agent economy)

AWS Marketplace is no longer just a catalog of apps—it’s becoming the backbone for distributing intelligence. This shift matters because marketplaces will define how enterprises source, scale, and govern agentic capabilities.

  • Agent Mode enables natural language discovery. This allows procurement teams to describe outcomes instead of hunting for SKUs, accelerating solution selection.
  • Express Private Offers streamline contracting. This reduces friction and speeds time-to-value for complex enterprise deals.
  • Solution bundles combine software and services. This makes it easier to deploy complete, outcome-driven solutions without stitching components together.
  • New Agentic AI partner categories. This gives SIs and industry experts a global route to market, enabling them to turn know-how into monetizable agents at hyperscaler scale.

AWS also introduced Agent Mode, a conversational procurement assistant that reads requirements and recommends solutions. The ambition is clear: transform buying from transactional to advisory. But here’s the challenge—AWS must deliver a trusted buyer experience with credible authority to architect and propose the best solution, or risk marketplaces becoming noisy catalogs instead of strategic platforms. You should:

  • Explore these marketplaces now — they’re becoming the distribution layer for the agent economy.
  • Plan for hybrid intelligence — buy agents for commoditized tasks, build your own for proprietary processes.
  • Rethink governance — agents will negotiate, collaborate, and learn; your trust layer must keep up.

 

What This Means for Enterprise Applications

The winners in this economy will master the art of composing agents into a fabric of intelligence.

If you’re building on AWS, you’re not just building an app — you’re building a node in an intelligent business fabric. The winners will be those who embrace this shift now.

Enterprise applications as we know them are about to disappear. Why? Because hyperscalers like AWS are turning their platforms into the agentic backbone for the next generation of business systems.

Every new AWS capability — fromAgentCore to Bedrock fine-tuning, from Trainium3 chips to AI-powered Marketplace — becomes a building block for software makers. ISVs will use these components to create agentic applications for the market, while enterprise developers will compose internal fabrics of AI agents, microservices, and data layers that learn, reason, and act.

This changes everything:

  • Applications become invisible. Users interact with outcomes, not screens. Agents orchestrate workflows behind the scenes.
  • Customization accelerates. Low-code and episodic memory let enterprises tailor agents to their unique processes without heavy coding.
  • Ecosystems explode. Marketplaces become the distribution layer for agentic capabilities, giving SIs and industry experts a global route to market.
  • Pricing shifts to outcomes. Seat-based licensing fades as consumption and value-based models take center stage.

Ready to Align Your Tech Strategy with the Agentic Future?

The shift from monolithic apps to agentic fabrics isn’t a distant forecast — it’s happening now. Hyperscalers like AWS are laying the foundation for this transformation with capabilities like AgentCore, Bedrock, and AI-powered marketplaces. The question is: Are you ready to rearchitect for this new reality?

Don’t wait until the fabric is woven around you. Schedule an inquiry or guidance session with me, Joe Cicman, or Faram Medhora to explore:

  • How to evolve your application strategy from SaaS to agentic.
  • Where to start with composability, data fabrics, and AI governance.
  • How to leverage hyperscaler ecosystems for speed and scale.

Forrester clients can read: The Evolution Of Modern Business Applications and The Agentic Business Fabric: AI’s Architectural Transformation Of Business Applications