As generative AI continues to evolve, 2025 is anticipated to witness a surge in high-value use cases for AI agents, which are becoming increasingly relevant in professional settings.
AI agents, initially developed alongside various modalities such as image and audio processing, now encompass systems that facilitate action and communication through large language models (LLMs). These agents are defined as comprehensive AI systems that integrate multiple tools to perform complex tasks beyond simple Q&A interactions.
Nicolas Gaudemet, Chief AI Officer at OnePoint, describes AI agents as cognitive architectures that utilize AI models to orchestrate various tasks, including planning, online research, and document analysis.
OpenAI has quickly capitalized on the potential of AI agents by launching plugins that offer numerous tools for developers, including code interpreters and web search capabilities. Similarly, French startup LightOn is preparing to offer clients tools to create customized agents by the end of the year.
LightOn's offerings will include Websearch for online information retrieval, Executive Python for data analysis, and Search Document for document queries. Laurent Daudet, co-CEO of LightOn, notes that the current phase of generative AI represents a shift towards more sophisticated uses of LLMs capable of managing diverse data.
OnePoint has developed demonstrator agents that successfully perform tasks such as analyzing bids and developing business models. These agents automate processes like summarizing documents and generating relevant questions.
Looking ahead, AI agents are expected to enhance their capabilities with new tools and connectors, facilitating integration with various data sources such as corporate CRM systems. This development is crucial as it will expand the operational possibilities for businesses.
Upcoming advancements include improved interaction capabilities, such as real-time communication and autonomous collaboration between agents, moving closer to the vision of fully autonomous assistants.