Home Bots in Society No human in the loop: the rise of AI Agents

No human in the loop: the rise of AI Agents

by Marco van der Hoeven

At the Gartner ITSymposium/Xpo 2024 keynote one of the most far reaching predictions from analysts Alicia Mullery and Daryl Plummer about AI was the expectation that AI agents are set to play an increasingly central role in organizations, evolving beyond general-purpose AI tools into specialized, expert-driven AI systems capable of handling specific tasks independently.

Gartner defines AI agents as distinct types of AI with specialized skills, broadly divided into two types. The first is the category of the ‘Guardian Agents’, AI systems designed for monitoring, security, and oversight tasks. Guardian agents can operate continuously, providing functions like real-time fraud detection, anomaly spotting, or identifying deepfakes. They can monitor data flows, flagging suspicious activity without human intervention. This helps bridge the gap where there may not be enough human resources to monitor all processes continuously.

The other category is that of the Expert Agents. Unlike Guardian agents focused on security, Expert agents are designed to perform complex, specialized tasks at a high level of expertise. The analysts envision Expert agents as AIs that go beyond answering questions to executing specialized functions – such as contract negotiations or completing detailed reports – with precision.

Roles

AI agents could take on several roles traditionally handled by specialized personnel, offering capabilities in areas like contract negotiation, where an AI agent trained in contract details and compliance could potentially review and negotiate contracts autonomously, ensuring all terms align with organizational standards.

In complex fields like tax filing, an AI agent could complete filings, review documents, and ensure regulatory compliance with minimal human oversight. In simpler, day-to-day contexts, AI agents could enhance productivity. For instance, a PowerPoint AI agent could quickly create complex slide designs or provide layout suggestions, automating repetitive and time-consuming tasks to save employees’ time and improve quality.

Experts

According to the analysts, AI agents are moving toward “expert” level functionality, evolving from generic automation to specialized roles that could handle entire jobs or processes. They predict that by 2025, AI agents will be designed to perform at a level comparable to human experts in specific domains, completing tasks with expert-level judgment and autonomy.

There is also a cautionary note regarding over-reliance on AI agents, particularly the risk of sidelining human expertise or facing ethical concerns around data privacy. For example trust and security. While Guardian agents are adept at monitoring, the reliance on AI to make security and compliance decisions could raise concerns about accountability and the potential for AI-driven errors.

Misuse

AI agents trained on an employee’s specific knowledge could eventually replace that individual’s role or expertise, potentially diminishing human input and decision-making. The analysts highlight the importance of negotiating appropriate terms for using AI agents to avoid unintentional knowledge misuse and ensure employees maintain control over their knowledge.

As AI agents become more embedded in business workflows, they’re expected to integrate seamlessly into daily operations, eliminating the need for users to engage them actively. This could drive operational efficiencies but may also add to costs as organizations deploy these systems at scale. The analysts point to the need for cautious budgeting, as these agents will require substantial initial investment and operational costs but could yield long-term productivity benefits.

In summary, AI agents are anticipated to transform organizational workflows by taking on high-stakes, expert-level roles, from monitoring to specialized tasks like contract negotiation and design. While offering substantial efficiency gains, they will require careful management to address cost, privacy, and ethical concerns in integrating AI-driven autonomy into businesses.

 

Misschien vind je deze berichten ook interessant