The COVID-19 crisis has presented severe challenges to our education system. In particular the crisis has highlighted and exacerbated the sharp inequalities within education, with those from poor backgrounds suffering most. The problems with administering exams have shone a light on the difficulties we have in assessing student ability and achievement in a fair and consistent manner. At the same time there has been some, albeit patchy, success in introducing online learning to temporarily replace and support classroom teaching. This suggest that now would be a good time to explore how new types of AI tool can be developed to provide high quality, personalised learning in the future so that all can benefit from the highest quality teaching delivered in a way that will suit personal learning styles. Developing this type of capability may offer new, better and fairer methods for continually and objectively assessing student performance. And building effective forms of online learning might make our education system more resilient against future pandemics and other forms of disruptive event.