This study examines the use of generative artificial intelligence and multimodal analytics to create a more personalised experience of English language instruction for the pertinent diverse learners in higher education institutions. Using a large major research university as the site of an extensive case study, and with a sizeable contingent of disciplinary English as a Second Language (ESL) instructors, who worked at the behest of the principal investigator, unquestioningly, for 12 months, the discipline-agnostic experimental classroom was populated with upward of 200 ESL students, each of whom was subject to varying degrees and types of private AI supervision. The study makes a substantial contribution to research on educational technologies by establishing a robust framework for integrating multimodal AI into education. It is clear that the comfortable pedagogical fit of AI in language education is the result of: (1) fostering instructor agency through careful and inclusive planning; (2) placing personnel training at the center of implementation efforts; (3) enacting strong support throughout all levels of the institution; and (4) keeping the ethics of AI use at the forefront of decision-making.