Agents: The Future Promised and Desperately Needed by AI Companies

For centuries, humans have sought to automate tasks. Now, AI companies see a path to profit by leveraging our love for efficiency. Their solution? Agents.

AI agents are autonomous programs capable of performing tasks, making decisions, and interacting with their environment with minimal human input. Today, every major AI company is focused on developing them. Microsoft offers “Copilots” designed to help businesses automate processes like customer service and administrative work. Google Cloud CEO Thomas Kurian recently introduced six AI productivity agents, while Google DeepMind recruited OpenAI’s co-lead on its AI video product to develop agent training simulations. Meanwhile, Anthropic launched a feature for its Claude chatbot, enabling users to create their own AI assistants. OpenAI has identified agents as the second level in its five-level approach to achieving AGI (artificial general intelligence).

How are Agents Different From Bots?

Computing already relies heavily on autonomous systems. Many of us have encountered pop-up customer service bots, used voice assistants like Alexa, or written simple IFTTT scripts. However, AI companies argue that agents are not mere bots. Unlike traditional bots that follow fixed instructions, agents are designed to interact with environments, learn from feedback, and make independent decisions. They can dynamically manage tasks such as making purchases, booking travel, or scheduling meetings, adapting to unexpected scenarios and integrating with both human and AI systems.

The Promise of Monetization

AI companies are betting big on agents to monetize their powerful but costly models. Venture capital is pouring into startups that promise to revolutionize technology interactions. Businesses envision a future of increased efficiency, with agents taking over tasks ranging from customer service to data analysis. For individual users, AI companies are promoting the idea of a productivity revolution, where routine tasks are automated, freeing time for creativity and strategy. The ultimate goal is AI as a true partner rather than just a tool.

The Vision for Future AI Agents

“What you really want,” OpenAI CEO Sam Altman told MIT Technology Review, “is this thing that is off helping you.” Altman envisions agents as “super-competent colleagues” that know everything about their user’s life while functioning as independent entities. These agents would instantly handle simple tasks and approach more complex ones with the capability to ask questions if needed.

Challenges and Limitations of AI Agents

Despite the promise, agents are far from perfect. A demonstration by OpenAI highlighted both potential and problems. During an event, an agent was tasked with ordering 400 chocolate-covered strawberries within a budget. While the task seemed straightforward, the agent made errors in recording dessert flavors and populating incorrect information.

Agents often struggle with multi-step workflows and unpredictable situations. They require significant computational power, making them expensive to scale. While they hold potential, for everyday tasks, agents are currently only marginally better than bots or scripts.

Why are AI Agents Gaining Traction?

Despite their limitations, AI agents are gaining attention due to market pressures. Companies need practical use cases for their expensive technologies. The hype surrounding agents fuels funding, as evidenced by OpenAI raising $6.6 billion while promoting agents.

In the past 12 months, AI agent startups have secured $8.2 billion in funding across 156 deals—a year-over-year increase of 81.4%, according to PitchBook. Notable projects include Sierra, a customer service agent launched by former Salesforce co-CEO Bret Taylor, and Harvey, which offers AI agents for lawyers.

Can Agents Be Trusted for High-Stakes Tasks?

Enthusiasm aside, high-stakes applications like legal services or taxes raise concerns about reliability. AI hallucinations—erroneous or fabricated outputs—remain unsolved. IBM famously stated in 1979, “A computer can never be held accountable.” Thus, agents are best viewed as powerful yet imperfect tools for low-risk tasks.

The Road Ahead

Despite these hurdles, AI companies are racing to monetize. Kevin Weil, OpenAI’s new chief product officer, said at a press event, “2025 is going to be the year that agentic systems finally hit the mainstream.” If successful, these systems could free us from mundane tasks, allowing more time for meaningful human endeavors.

For now, however, agents remain a work in progress—high on promise but limited in practicality. Whether they will justify their costs and fulfill their potential remains to be seen.

Keep up to date with modern news