AI Agents Surge: IT Departments Transform into New HR Units

Agentic AI: The Backbone of Modern Enterprise Systems

Agentic AI is transforming the landscape of connected enterprises, moving far beyond the realm of simple helper apps. These AI agents are becoming the driving force behind the microservices that underpin enterprise systems. As their use expands, IT departments are evolving into a sort of "human resources" for AI, handling the acquisition, onboarding, and management of these digital assistants much like HR manages human employees.

These insights emerged from a panel discussion at the recent Mobile World Congress, organized by Deloitte, which delved into the evolving role of AI agents within enterprises. Panelists drew parallels between agentic architecture and the rise of microservices, noting how both break down large, monolithic applications into smaller, more manageable and flexible units.

From Monolithic to Microservices

"Agentic AI is the next step in breaking apart and solving problems," explained Bryan Thompson, vice president for GreenLake product management at HPE. He highlighted the potential of agentic AI to be segmented into microservices, enabling a more targeted approach to problem-solving. "It's about leveraging these models to break them into specialized services," he added.

Fred Devoir, global head of solution architecture for telco at Nvidia, agreed, emphasizing how agentic AI helps in stitching together enterprise workflows. "We integrate components into a RESTful architecture," he said. "Nvidia optimizes these microservices, assembling them into blueprints that quickly deliver value or results."

Capabilities Beyond Traditional Microservices

Abdi Goodarzi, head of gen AI products, innovations, and new businesses for Deloitte, pointed out that agentic AI offers capabilities far surpassing those of traditional microservice architectures. "We've never had a technology that could ideate or execute independently," he noted. "This independent execution is truly the power of AI."

Agentic AI services take over many of the tedious tasks previously handled by humans, forming a parallel workforce managed by IT rather than HR. "It's the same concept as human capital management, but applied to AI agents," Devoir remarked. "The IT department now acts as HR for these agents, curating, setting guardrails, training, and fine-tuning them for specific tasks and integration with human workflows. It's a complex job, akin to HR but on a much deeper technical level."

Organizational and Cultural Shifts

The rise of agentic AI necessitates significant changes within organizations. Goodarzi highlighted the emotional aspect: "Humans have emotions, agents don't. How do we incorporate emotions into the execution of work? As work processes change, so must the culture, talent strategies, and the way humans and machines collaborate."

Challenges in Implementation

Implementing agentic AI within enterprises comes with its set of challenges, particularly around data management, trustworthiness, and talent. Goodarzi mentioned the heavy investment enterprises have made in organizing structured data. "They've built ERP systems and systems of record and action, but these often result in data silos."

Agentic AI can help address this by deploying agents where the data resides. "You bring the AI to the data," Devoir explained. "When you make a service call, it queries the data agents for a response and then collates that data into a model."

Trustworthiness is another critical issue. Goodarzi stressed the need to verify data and results: "Are you dealing with the right data? Are the results correct? Traditional technologies focused on transactional activities, but agentic AI is probabilistic. It provides the best probable answer based on its training and data digestion."

Trust in these AI agents is paramount: "Can you trust this agent? Is the data accurate? These are questions that need to be answered."

The Road Ahead

Despite these challenges, the potential of agentic AI is undeniable. Goodarzi concluded, "These are new concepts for enterprises, which has slowed adoption. But the capabilities are real, and the technology is advanced enough for use in enterprise production systems. I believe this year will see a significant uptake."

The journey to an agentic AI-powered enterprise is complex, but the rewards in terms of efficiency and innovation are substantial. As enterprises continue to navigate these waters, the role of AI agents will only grow more integral to their success.