AI Agents: Transforming Work Lives with Independent Interaction

AI-based agents are emerging as a significant mode of delivery for AI systems, offering a transformative approach to automating tasks and interactions. Unlike traditional AI implementations, these agents are easier to set up and manage, capturing the attention of technologists and business leaders alike.

Rise of Agentic Systems

According to a report from McKinsey, AI-based agents represent the "next frontier" of AI. These "digital systems that can independently interact in a dynamic world" are predicted to grow in influence. The natural-language capabilities of generative AI unveil new possibilities, enabling systems to plan actions, use online tools, collaborate with other agents and people, and improve their performance over time.

Transformative Potential

McKinsey's report, led by Lareina Yee, suggests that generative AI is evolving from knowledge-based tools, like chatbots, to AI-enabled agents capable of executing complex workflows. "We are beginning an evolution from knowledge-based, gen-AI-powered tools to gen AI-enabled agents that use foundation models to execute complex, multistep workflows across a digital world. In short, the technology is moving from thought to action," the report states.

Survey Insights

A recent survey from Capgemini indicates that 82% of 1,100 tech executives plan to integrate AI-based agents within the next three years, a significant increase from the current 10%. The survey also found that 70% of respondents would trust an AI agent to analyze and synthesize data, and 50% would trust one to send professional emails on their behalf. Many respondents intend to deploy AI agents for tasks such as generating and improving code, drafting reports, and creating website content.

Diverse Roles and Applications

AI-powered agents can assume various roles. For example, a virtual assistant could plan and book personalized travel itineraries, while a programmer agent could code, test, iterate, and deploy new software features based on an engineer's descriptions. Qventus, a vendor, offers an AI-based assistant called the Patient Concierge, which manages patient appointments, care guidelines, and answering general care questions.

Levels of AI Agents

According to an AWS tutorial, there are six levels of AI agents, each offering increasing functionality:

• Simple Reflex Agents: Perform simple tasks like resetting passwords based on predefined rules.

• Model-Based Reflex Agents: Evaluate outcomes before deciding, building an internal model of the world to support its decisions.

• Goal-Based/Rule-Based Agents: Use robust reasoning for complex tasks like natural language processing and robotics applications. It compares different approaches and chooses the most efficient path.

• Utility-Based Agents: Compare scenarios to maximize desired outcomes, like finding the best airline deals for customers.

• Learning Agents: Continuously learn from experiences to improve results.

• Hierarchical Agents: Oversee other agents, deconstructing complex tasks into smaller ones and assigns them to lower-level agents.

Shift in Implementation

Historically, software agents required laborious, rule-based programming or specific training of machine-learning models. However, generative AI changes this dynamic. When built using foundation models trained on large, varied datasets, agentic systems can adapt to different scenarios much like LLMs (large language models) respond to diverse prompts.

Natural-Language Processing

AI agents' use of natural-language processing simplifies the automation of workflows. Previously, automating a task required breaking it down into rules and steps that could be codified into computer code—a costly and labor-intensive process. Now, with natural language as a form of instruction, even nontechnical employees can encode complex workflows quickly and easily.