Pain detection is critical for individuals unable to correctly self-report, such as young children or patients with intellectual disabilities. This paper introduces an Integrated AI System for the physiological assessment of Pain and social robotic intervention, which transcends traditional monitoring by combining real-time physiological measurements with AI assessment and immediate behavioral support. The system architecture enables a multimodal approach, currently integrating electrodermal activity and heart rate signals to feed a Random Forest Classifier model for pain level estimation. Pain assessment results are communicated through a dual-interface framework: a mobile application provides caregivers with intuitive visualizations and longitudinal history for clinical oversight, while a social robot (MiRo-E) can utilise the assessment for interactive interventions, such as distraction, to assist patients in coping with pain. As this study serves as a proof of concept for the integrated system, achieving high diagnostic accuracy was not the primary scope of this work. While this framework establishes a functional foundation for integrated pain management, testing is required in clinical settings to validate the practical value of pain assessment and the therapeutic impact of the robotic interventions in healthcare.