We are entering the third phase of generative AI, characterized by the evolution from simple chatbots to more sophisticated AI agents. This new breed of technology not only seeks greater autonomy but also aims to work collaboratively, solving complex problems using tools and advanced reasoning.

The latest development in this arena is OpenAI’s ChatGPT agent, which integrates its previous products, Operator and Deep Research, into a streamlined system with the ability to “think and act.” Not just a mere conversational agent, this innovation represents a significant leap, enabling users to accomplish a wider array of tasks with a more effective and powerful AI.

The journey from chatbots to AI agents illustrates a remarkable evolution. ChatGPT sparked the chatbot revolution in November 2022; however, despite its effectiveness, the conversational interface imposed constraints on what the technology could achieve. The emergence of AI assistants or copilots transformed this landscape, utilizing the same underlying large language models while focusing on executing tasks with enhanced human oversight.

AI agents, on the other hand, are designed to pursue specific goals with varying degrees of independence. They are supported by an advanced suite of capabilities that include memory and reasoning, allowing for better decision-making processes. Furthermore, multiple AI agents can interact and collaborate, communicating to plan, schedule, and tackle complex challenges efficiently.

Another distinctive feature of AI agents is their ability to use various tools. This capability enables them to execute specialized tasks through software tools such as web browsers, spreadsheets, and payment systems, enhancing their functional versatility.

Rapid advancements in agentic AI have been evident since last year. A pivotal moment was marked last October when Anthropic empowered its Claude chatbot to interface with computers similarly to human users. This upgrade allowed the agent to sift through various data sources, retrieve pertinent information, and even submit entries on online platforms—a remarkable step toward bringing AI closer to human-like interactions.

The pace of innovation has since accelerated, with other AI developers quickly bringing their own agent solutions to market. OpenAI has introduced a web-browsing agent called Operator, and Microsoft revealed its Copilot agents, among others. In the same vein, Google unveiled its Vertex AI, while Meta presented its Llama agents, showcasing an impressive array of options available to businesses.

Notably, some startups across various regions have also entered the spotlight, demonstrating significant capabilities. In early 2023, the Chinese startup Monica showcased its Manus AI agent successfully buying real estate and generating summary notes from lecture recordings, while Genspark released a search engine agent that provides single-page summaries embedded with links to relevant online tasks, much like Google. Another intriguing initiative is Cluely, which offers a rather unconventional “cheat at anything” agent that has captured attention but remains to produce substantial outcomes.

It’s important to recognize that not all agents are suited for a general-purpose approach. Some have been tailored for specific functional domains. In the realm of software engineering, agents such as Microsoft’s Copilot and OpenAI’s Codex are leading the charge, demonstrating an ability to autonomously write and evaluate code while identifying human-created code for potential errors and performance issues.

The dynamic capabilities of AI agents extend beyond coding; they encompass various applications such as search and summarization, proving their value across multiple domains. Organizations must familiarize themselves with these new tools to explore their potential impact effectively.

In this era of AI agent development, understanding their functionalities, benefits, and risks is becoming increasingly vital for business leaders, product builders, and investors. As these systems evolve, they promise to redefine how we interact with technology and automate processes, making significant strides toward a future where AI can assist in intricate, goal-focused tasks.