Building the agentic operating system

IBM’s Agentic AI: Powering Enterprise Transformation with Trust and Precision

Zachary Karabell discussed IBM’s evolution with Mr. Matt Lyteson, CIO. Moving beyond its traditional image, IBM now leads in cutting-edge enterprise AI. Their work, less consumer-facing, focuses on secure, governed, and trusted AI solutions for complex enterprise workflows, establishing an “agentic operating system” at its core.

Mr. Lyteson detailed IBM’s foundation in enterprise computing, from mainframes to hybrid cloud and early AI with Deep Blue, now leveraging the Watson X platform. The “Client Zero” program involves IBM’s internal teams as first users, refining offerings and demonstrating practical AI adoption to clients.

IBM’s focus is distinctly on enterprise applications, differing from general consumer AI. An agentic IT support capability, “As IT,” deployed to 280,000 employees, deflects 86% of routine calls. In mergers and acquisitions, AI processes 100% of contracts with 95% accuracy, identifying 30 times more unfavorable terms, accelerating deal closures.

The company prioritizes cost-effectiveness using specialized, mid-tier models like the Granite family. These models run efficiently on internal infrastructure, such as the Z platform, providing necessary performance for specific use cases. This approach avoids massive capital expenditure, ensuring fiscal responsibility and scalability for AI solutions.

IBM fosters an “AI-first operating model” by integrating AI incrementally into workflows and providing individual productivity tools that enhance job satisfaction and employee engagement. Mr. Lyteson emphasizes delivering tangible value and a strong return on investment by managing technology costs through optimized model usage.

Advanced tools like “IBM Bob,” an agentic integrated development environment, are now accessible to a broader audience, including small businesses. This tool assists developers with requirements, architecture planning, code writing, testing, security vulnerability checks, and compliance, with a free trial available.

Addressing public concerns, Mr. Lyteson emphasized the critical importance of security, safety, governance, and auditability in enterprise AI platforms. He also highlighted the human element, stressing the need to manage workforce changes, facilitate skill adoption, and integrate new talent for continuous evolution and smooth transitions.

Globally, despite varying regulatory requirements, the core principles of AI value extraction remain consistent: driving revenue growth, improving operating performance, and managing risk. These universal business needs underscore AI’s broad applicability. Looking five years ahead, Mr. Lyteson anticipates discussions on hyper-growth and quantum computing’s transformative impact.