This chapter explores applicable principles of AI innovation, which may be used in radiology, and follows the AI product lifecycle starting from development to testing, validation, and different model types to inform the reader. Medical devices must be certified before clinical use and are subject to regulation, the specifics of which may vary in different geographical locations. Applications incorporating AI are classed as high-risk medical devices when developed for use in healthcare. Once developed, algorithms or applications must, therefore, be subject to extensive and transparent evaluation and validation prior to consideration of adoption into clinical use. There are many more hurdles to overcome in commercialisation of any AI model, from funding to the achievement of regulatory clearance in multiple geographies, and then educating the market, marketing the product and post-market surveillance of performance; all are explored here through the description of two specific real-world case studies. We hope this learning will support those of you who might be thinking of embarking on an entrepreneurial journey as AI innovators, using your clinical expertise as a starting point and this chapter as a rough guide and inspiration. The pace of AI development is faster than any publishing cycle, so we also explore the near future of AI, though we are not quite yet able to discuss generative artificial intelligence—which may well be in full swing by the time the reader reaches these pages.