AWS re:Invent 2025 - Transforming from SaaS to multi-tenant agentic SaaS (SAS304)

Transforming from SaaS to Multi-Tenant Agentic SaaS

Overview

• This presentation discusses the transformation from traditional SaaS to a more advanced "agentic SaaS" model, where AI agents are integrated into SaaS applications.
• The speakers cover the key considerations, technical details, and business impact of this transition, providing a practical guide for SaaS providers.

SaaS Fundamentals

• SaaS value propositions include scalability, efficiency, agility, and frictionless onboarding.
• These core SaaS principles still apply in the agentic SaaS model, but with additional considerations around agent-based automation and decision-making.

Agentic AI

• Agentic AI refers to systems with a high degree of autonomy and the ability to make decisions with minimal human oversight.
• This unlocks new capabilities for SaaS applications, such as 24/7 monitoring, process automation, and enhanced customer experiences.

Agentic SaaS Flavors

1. Agent-Ready SaaS: Existing SaaS applications wrapped with agent interfaces (MCP, A2A) to enable new consumption channels.
2. Embedded Agentic SaaS: SaaS applications with automated workflows and agent-based components integrated into the core system.
3. Human-Facing Agents: Chatbots and other agent-based user interfaces for interacting with SaaS applications.
4. Agent-Facing Agents: Agents that can communicate and coordinate with each other without human intervention.

Agentic Transformation Strategy

1. Domain Analysis and Modeling: Identify domains, subdomains, and bounded contexts within the SaaS application.
2. Agent Capability Development: Map agent responsibilities to the identified domains and build initial agent prototypes.
3. Infrastructure: Deploy agents in a secure, scalable, and reliable manner using a platform like AWS Agent Core.
   • Agent Core provides a managed runtime, identity management, observability, and other key capabilities.
4. Scaling and Monitoring: Implement strategies for scaling agent-based systems and monitoring their performance.

Technical Details

• Agent deployment models: Dedicated (per-tenant) or Shared (multi-tenant)
• Agent Core components:
  • Runtime: Executes agent code
  • Gateway: Enables agents to discover and access external tools and services
  • Identity: Handles authentication and authorization for agents
  • Observability: Provides visibility into agent performance and interactions
• Tenant isolation and security:
  • Workload access tokens for secure agent-to-agent and agent-to-service communication
  • Tenant-specific identity and access control policies
  • Dedicated agent runtime and memory for each tenant

Business Impact

• Improved customer acquisition through personalized offerings and integrations.
• Enhanced customer retention via reliable, automated workflows and improved customer success.
• Streamlined onboarding and reduced manual tasks through agent-based automation.
• Opportunities for new pricing models, such as outcome-based or per-agent pricing.

Examples and Use Cases

• Automating lead management, opportunity creation, and task assignment in a CRM SaaS application.
• Integrating agents with external services and APIs to unlock new capabilities for customers.
• Leveraging agent-based observability to improve reliability, identify issues, and enhance the customer experience.

Key Takeaways

1. SaaS is a business solution with technical implementations, and the agentic transformation should be driven by business needs.
2. Adopt a phased approach to agentic transformation, starting with domain analysis, agent capability development, infrastructure, and scaling/monitoring.
3. Leverage platforms like AWS Agent Core to simplify the deployment and management of agent-based systems, with a focus on tenant isolation and security.
4. Explore new pricing models and revenue streams enabled by agentic SaaS, such as outcome-based or per-agent pricing.
5. Integrate agentic capabilities seamlessly into existing SaaS applications to enhance customer experiences and operational efficiency.