AWS re:Invent 2025 - Integrated AI Agents: Moving Beyond Summarization to Automation (AIM207)

Integrated AI Agents: Moving Beyond Summarization to Automation

From SaaS 1.0 to SaaS 2.0

• The presentation starts by highlighting the evolution from traditional software running on local computers to cloud-based SaaS (Software as a Service) solutions.
• SaaS 1.0 provided benefits like scalability, security, and transparency, but also led to the creation of rigid data schemas and terminology specific to each department or tool.
• This presents an opportunity to rethink how AI can be leveraged to move beyond simple summarization and towards more integrated automation.

Lessons from the History of Computing

• The presenter looks back at the history of computing, noting how platforms catered to builders by providing the right separation of concerns and abstractions.
• This allowed developers and entrepreneurs to build innovative solutions on top of the platform, rather than having to manage the underlying infrastructure.
• However, as these solutions proliferated, companies faced challenges with security, deployments, and maintenance, leading to the rise of cloud computing.

The Challenges of Adopting AI

• The presenter highlights the current challenges faced by companies trying to adopt AI, particularly in the context of "agentic AI" or conversational AI.
• Many companies have rushed to implement AI solutions, only to find that 95% of these pilots fail, often due to issues with data accessibility, provisioning, auditability, and end-to-end use case integration.
• The presenter emphasizes the need to think holistically about the data layer, operational layer, and application layer when deploying AI solutions, rather than just building prototypes.

Towards a Modern "Computer" for AI-Powered Automation

• The presenter introduces the concept of "Computer" by DevRev, a platform that aims to address the challenges of integrating AI into enterprise workflows.
• Key features of this "computer" include:
  • Unified, bi-directional, and real-time access to data and metadata across disparate systems
  • Powerful search and contextualization capabilities powered by language models and SQL
  • Ability to build and upskill agents through a citizen development approach
  • Integration with external tools and data sources through an "MCP" (Microservices Connectivity Platform)
• The goal is to create a unified platform that can answer questions, synthesize information, provide recommendations, and take actions, all while respecting the existing data schemas and access controls.

Practical Applications and Benefits

• The presenter showcases examples of how the "Computer" platform can be used to answer questions, schedule meetings, and provide personalized information based on the user's context and role.
• The key benefits highlighted include increased productivity, faster decision-making, and the ability to bridge the gaps between different departments and systems.
• The platform is designed to be scalable, with the ability to deploy agents and skills at the personal, team, or enterprise level, while maintaining compliance and governance.

Conclusion

• The presentation emphasizes the need to move beyond simple AI-powered summarization and towards more integrated, automated solutions that can truly transform how organizations work.
• The "Computer" platform by DevRev is presented as a potential solution to this challenge, offering a comprehensive approach to unifying data, empowering agents, and enabling AI-driven automation at scale.