Digital Twins and AI Agents for IT Operations

Jun 21, 2025

UC San Diego CIO's Vision: Digital Twins and AI Agents for IT Operations

Dr. Vince Kellen, Chief Information Officer at the University of California San Diego, has proposed an innovative approach to addressing IT workforce challenges through the creation of digital twins and AI agents that replicate the expertise of top technical professionals1. Speaking at Cisco Live 2025, Kellen outlined a strategy that could fundamentally transform how universities manage their technology infrastructure and cybersecurity operations2.

The Core Concept: Cloning Technical Expertise

Kellen's proposal centers on extracting and digitizing the implicit knowledge held by experienced IT professionals to create AI-powered digital twins that can handle routine tasks and after-hours troubleshooting 1. This approach aims to address two critical challenges facing higher education: declining budgets and increasing cybersecurity threats 1 2.

The concept involves what Kellen describes as getting knowledge "out of their minds in a kind of drip irrigation fashion" and bringing it into AI systems to improve performance through "human/technical symbiosis"1. This knowledge extraction process would allow AI agents to replicate expert decision-making patterns and responses to common technical incidents 1.

The UC San Diego Context

Scale and Complexity

UC San Diego operates technology infrastructure comparable to a small city, serving 100,000 to 150,000 daily visitors across its campus 1. As a top-20 university worldwide with over $1 billion in annual research funding, the institution faces unique operational demands that make automation particularly appealing 3.

Kellen, who brings 25 years of executive-level IT experience and has served in advisory roles with major technology companies including Dell, SAP, Microsoft, and Apple, is well-positioned to implement such transformative changes 4.

Unique Security Challenges

The university faces distinctive cybersecurity threats stemming from its Scripps Institution of Oceanography, which conducts sensitive marine research 1. Kellen noted that "when you put sonar in the water, you discover more than fish, and other countries want to know about that" 1. This research activity has attracted sophisticated, well-financed foreign adversaries conducting long-duration campaigns against the university1.

The institution has experienced "very exquisite attacks coming out of foreign actors" that require specialized defense strategies beyond conventional cybersecurity measures 1. This threat landscape makes the automation of routine security tasks essential to free human resources for combating advanced persistent threats 1.

The Technology Framework

Agentic AI Integration

The digital twin approach leverages what industry experts now call "agentic AI" – autonomous AI systems capable of completing entire workflows without human intervention. This represents a shift from traditional AI assistance to systems that can "think and act independently" 5.

Kellen's vision involves creating AI agents that operate within digital twin frameworks, enabling proactive detection of infrastructure issues before they impact services. This approach transforms IT operations from reactive troubleshooting to predictive maintenance and prevention 6.

Infrastructure as Code Philosophy

Supporting this digital transformation, Kellen advocates for "Infrastructure as Code" as a fundamental requirement for managing hybrid cloud and on-premises environments 7. This approach enables seamless automation across different computing environments, from cloud to on-premises systems, supporting the distributed AI computing model he envisions 7.

Broader Industry Context

Digital Twins in Cybersecurity

The concept of using digital twins for cybersecurity is gaining traction across industries 8 9. Organizations are implementing AI-powered digital twins that simulate real-world threats and attacker behaviors within controlled virtual replicas of IT infrastructure 8. This enables proactive threat detection and security optimization before actual attacks occur 8.

Research from NIST and the University of Michigan has demonstrated the feasibility of using digital twins with machine learning for cyberattack detection, showing particular promise in manufacturing and operational technology environments 10.

The Knowledge Externalization Challenge

Kellen's approach addresses a fundamental paradox in professional expertise and AI collaboration: as domain experts share their implicit knowledge with AI systems, they potentially accelerate the automation of their own roles11. However, research suggests this collaboration can also create opportunities for the evolution of expertise and new forms of professional value11.

The process of "collaborative externalization" allows AI systems to extract relational tacit knowledge through demonstrations, corrections, and contextual guidance, creating more dynamic knowledge capture than traditional expert systems11.

Implementation at UC San Diego

TritonGPT Platform

UC San Diego has already begun implementing AI-driven solutions through TritonGPT, a suite of AI assistants designed for administrative tasks and institutional data analysis12. Hosted on-premises at the San Diego Supercomputer Center, this platform demonstrates the university's commitment to maintaining control over sensitive data while leveraging advanced AI capabilities12.

The TritonGPT system serves as a foundation for the broader digital twin vision, providing experience with AI integration and data security in a university environment12.

Hybrid Computing Strategy

Kellen's infrastructure strategy emphasizes a hybrid approach combining cloud and on-premises computing, particularly important given AI cost structures and data sovereignty requirements7. This hybrid model supports the distributed computing environment necessary for agent-to-agent communication and seamless automation across different platforms7.

Quality of Life Benefits

Reducing IT Professional Burnout

A key motivation for the digital twin approach is improving quality of life for IT professionals by reducing after-hours callouts and repetitive troubleshooting tasks1. Once an expert's knowledge and approach to solving incidents is captured, AI can replicate the solution for similar future problems, eliminating the need to repeatedly disturb human experts1.

This approach addresses a common source of burnout in IT operations, where experienced professionals are frequently called upon to resolve urgent technical issues outside normal working hours1.

Workforce Development

The digital twin model could also serve as a training mechanism, allowing newer IT staff to learn from the captured expertise of senior professionals13. This knowledge preservation aspect becomes particularly valuable as experienced workers retire or change roles13.

Challenges and Considerations

Higher Education Budget Constraints

Universities face significant budget pressures that complicate technology investments14. All IT leaders in higher education cite budget constraints as a primary roadblock for automation initiatives14. However, the long-term cost savings and efficiency improvements from digital twin implementation could justify the initial investment14.

Cybersecurity in Academic Settings

Academic institutions are increasingly vulnerable to foreign interference and cyber threats, particularly those involved in sensitive research15. The combination of remote work, large data volumes, and disruptive technologies like AI creates new attack vectors that require sophisticated defense strategies15.

Supply chain attacks, AI-powered threats, and insider threats represent growing concerns for higher education cybersecurity16. Digital twin approaches could help address these challenges by providing better monitoring and prediction capabilities16.

Future Implications

Industry Transformation

The concept of digital twins and agentic AI for IT operations represents a broader transformation in enterprise operations17. Organizations across industries are exploring similar approaches to create autonomous business processes and optimize operations through AI-driven insights17.

The networking industry specifically is moving toward "level 4+ autonomy" where networks can monitor themselves, act on intent, and recover from faults without operator intervention18. This aligns with Kellen's vision of proactive, AI-driven IT operations18.

The Human-AI Collaboration Model

Successful implementation requires careful consideration of how human expertise and AI capabilities complement each other19. The goal is not to replace human professionals but to augment their capabilities and free them for higher-value strategic work19.

As one industry observer noted, the key insight is to "respond to the prediction, not the incident"6. This shift from reactive to proactive operations represents a fundamental change in how IT organizations approach their responsibilities6.

Conclusion

Dr. Vince Kellen's proposal to create digital twins and AI agents from the expertise of top IT professionals represents an innovative response to the challenges facing higher education technology operations1. By combining advanced AI capabilities with deep institutional knowledge, universities could achieve unprecedented levels of automation while improving cybersecurity posture and staff quality of life1 8.

The success of this approach will depend on careful implementation that preserves the value of human expertise while leveraging AI capabilities for routine and predictive tasks11. As universities face increasing pressure from budget constraints and sophisticated cyber threats, such innovative approaches may become essential for maintaining effective IT operations14 15.

The broader implications extend beyond higher education, as the principles of knowledge externalization and agentic AI could transform IT operations across industries5 17. UC San Diego's implementation of this vision will likely serve as a crucial test case for the future of AI-augmented IT operations12.

Curious futures (KGhosh)