The dynamic development of 5G and future 6G networks has turned AI-driven resource allocation into a pillar of effective telecom operation. Nevertheless, fairness and transparency in these AI systems are largely unexamined. The research introduces a thematic analysis method to detect algorithmic disparity and injustice in resource distribution among various Quality of Service (QoS) classes. As a novelty, this analysis runs on a qualitative telecom dataset with parameters such as throughput, latency, and packet loss categorised by QoS levels High, Medium, and Low. The approach incorporates the six-stage thematic analysis scheme, tailored for numerical data interpretation. To develop the model, codes were pulled from trends affecting throughput, latency, and the number of packets lost among different QoS groups. The main ideas were used to enhance the AI resource scheduling model to check that the method worked. Based on the results, there was a clear pattern of high-priority services receiving better throughput, lower latency, and fewer packet losses compared to the lagging and compromised services for the medium and low-tier categories. As a result, the fairness in resource allocation improved by 9.6%, as shown by a steady throughput across all QoS classes and a noticeable reduction in the difference between the latencies of different classes. The study highlights how qualitative thematic findings can be incorporated into AI optimisation, making the process more equitable. The findings help ensure AI is used fairly and ethically in future telecom networks.