Agricultural robotics and the Internet of Things (IoT) require intelligent interfaces that combine artificial intelligence with human involvement to efficiently manage data flows, robot control, and precision farming operations. This PhD research is conducted within the European MSCA GreenFieldData project and is jointly supervised under a cotutelle agreement between the University Lyon 1 Claude Bernard and the Università degli Studi di Milano. The dissertation aims to design conversational layers based on Large Language Models (LLMs) to support sustainable, inclusive, and adaptive smart agriculture by facilitating interactions among users, robotic systems, and IoT infrastructures.