AWS re:Invent 2025 - IBM: Building blocks to scale AI agents: hybrid, integrated, automated​(HMC104)

Scaling AI Agents: Hybrid, Integrated, Automated

Adoption of AI in Banking

• AI has become a fundamental part of transformational plans for banking
• Many current use cases focus on efficiency and incremental improvements
• However, only 25% of organizations are using AI to drive growth and change business models
• Forecast: AI-powered workflows in banking will multiply by 12 by 2026, drastically changing how work is performed and customer interactions

Building an AI-Powered Housing Ecosystem

• Banco Estado, a 160-year-old Chilean bank, partnered with IBM Consulting and AWS to create a digital platform for the housing and construction ecosystem
• Key goals:
  • Reduce mortgage approval and processing time by 15 days
  • Reduce closing time by 30%
  • Decrease interactions between clients and businesses by 50%
  • Enable instant property appraisals for certain properties
  • Reduce processing costs by 30%
• The platform integrates banks, real estate companies, construction firms, notaries, and other providers using intelligent workflows and AI

Challenges in Scaling AI Agents

• Dynamic nature of AI agents - their behavior is complex and unpredictable, unlike traditional software
• Diverse infrastructure and tools that agents need to connect to, requiring loss of control
• Rapid pace of change in AI technologies, making it difficult for organizations to adapt

Strategies for Enterprise-Ready AI Agents

1. Reliability and Consistency: Ensure agents are reliable, consistent, and scalable
2. Performance and Value: Optimize agent performance and demonstrate clear ROI
3. Observability: Provide observability into agent behavior and decision-making
4. Automation: Enable agents to take automated actions, reducing human involvement
5. Governance: Implement secure policies and controls to govern agent actions

IBM's Offerings for Scaling AI Agents

1. IBM Instana: Application performance monitoring tool that provides full-stack observability, including for AI agents
2. Turbonomic: Resource optimization tool to ensure efficient utilization of compute resources (e.g., GPUs) for running AI agents
3. WebMethods: Secure, governed platform for running AI agents and integrating them with enterprise systems
4. Incident Investigation Agent: Helps improve incident investigation and root cause analysis within Instana
5. Auto Compliance and Resilience Agent: Automates compliance checks and ensures system resilience within Concert

Client Experiences

• Global Payments (Ganesh): Leveraging observability and auto-healing capabilities to ensure reliability and responsiveness of their mission-critical payment processing systems
• Flexivan (Sagar): Using AI-powered predictive logistics, computer vision, and language models to improve supply chain visibility, asset utilization, and process automation

Key Takeaways

• AI agents are becoming increasingly critical in banking and other industries, but scaling them comes with challenges around reliability, performance, observability, automation, and governance
• IBM offers a suite of tools and capabilities to help enterprises address these challenges and successfully scale their AI agent deployments
• Clients are seeing tangible benefits in terms of improved operational efficiency, reduced costs, and enhanced customer experiences by leveraging IBM's AI agent scaling solutions