Gartner: Over 40% of AI Agent Projects Canceled

The "agents" category attracts ambitions, but Gartner anticipates that more than 40% of agent-based AI projects will be canceled. To limit false starts, it is essential to distinguish between assistants, copilots, and agents, and to know when to deploy each.

Agent-based Projects: Anticipated Cancellations and Usage Context

Gartner predicts that over 40% of agent-based AI projects will be canceled. The integration of assistants, copilots, agents, and orchestration mechanisms can occur within the same system. The management of complex systems is cited as an example of application for agents. A week-long wedding at a palace hotel serves as an illustration to show how these different roles and orchestration can be coordinated.

What Assistants, Copilots, and Agents Do

AI assistants are designed to help users with specific tasks by answering questions or providing recommendations based on data. Copilots act as collaborators in professional activities, offering instant proposals and recommendations to optimize efficiency. Agents correspond to autonomous systems that make decisions and perform actions independently, particularly in complex situations.

Choosing the Tool: Simple Tasks, Co-Creation, or Total Autonomy

An AI assistant is used for simple tasks or to quickly obtain information. A copilot is suitable for interactive support during group projects or creative processes. An agent is relevant for uses requiring total autonomy as well as immediate decision-making capability. These distinctions allow for choosing between assistants, copilots, and agents based on the need.