What is an AI Agent? Definitions Vary Among Experts

AI agents are heralded as the next big thing in artificial intelligence, but there isn’t a universally agreed-upon definition. Different companies and experts have varying interpretations of what constitutes an AI agent.

Defining an AI Agent

At its core, an AI agent can be described as AI-powered software designed to perform tasks traditionally done by human customer service agents, HR personnel, or IT help desk employees. These agents can manage multiple systems and go beyond simply answering questions to performing complex tasks.

However, the lack of a cohesive definition has led to some confusion. For instance, Google views AI agents as task-based assistants, helping with coding, marketing, and IT troubleshooting. Asana sees them as co-workers handling assigned tasks. Sierra, a startup by former Salesforce co-CEO Bret Taylor and Google vet Clay Bavor, considers them customer experience tools that solve complex problems beyond the capabilities of traditional chatbots.

Varying Perspectives and Technologies

Rudina Seseri, founder and managing partner at Glasswing Ventures, notes that the lack of a single definition stems from the technology’s early stage. She describes an AI agent as an intelligent software system designed to perceive its environment, reason, make decisions, and take actions to achieve specific objectives autonomously. These systems use AI/ML techniques such as natural language processing, machine learning, and computer vision to operate in dynamic environments.

Aaron Levie, co-founder and CEO of Box, believes AI agents will become more capable over time due to improvements in GPU price/performance, model efficiency, quality, and AI frameworks. However, MIT robotics pioneer Rodney Brooks cautions that AI faces tougher challenges than other technologies and may not progress as rapidly.

Challenges and Realistic Expectations

One significant challenge for AI agents is integrating with multiple systems, especially older ones lacking basic API access. While improvements are being made, the complexity of accessing and solving problems across different systems remains a hurdle.

David Cushman, a research leader at HFS Research, compares current AI agents to assistants helping humans complete tasks to achieve strategic goals. The goal is for AI agents to operate independently at scale, but this requires further advances in technology and infrastructure.

Future of AI Agents

Jon Turow, a partner at Madrona Ventures, highlights the need for a dedicated tech stack for AI agents. This infrastructure must support AI agents and their applications, ensuring scale, performance, and reliability. Turow believes that multiple models, rather than a single large language model, will be necessary to create effective AI agents.

Fred Havemeyer, head of U.S. AI and software research at Macquarie US Equity Research, envisions a future where AI agents are truly autonomous, capable of taking abstract goals and independently reasoning through all necessary steps. However, achieving this vision will require significant advancements and breakthroughs.

While AI agents show promise, the technology is still evolving. It’s important to recognize that we are in a transitional phase, and the fully autonomous AI agents envisioned by experts are not yet a reality.