Cracking the code of failed AI pilots

In this episode of the Practical AI podcast, co-hosts Daniel Whitenack and Chris Benson discuss AI news and trends, focusing on the impact of AI on jobs and the high failure rate of AI pilot projects. They analyze an MIT report highlighting that 95% of AI pilots fail, not due to model capabilities, but because of a learning gap in understanding how to design effective AI workflows. They explore the misconception that general-purpose models can be directly applied to complex business processes, emphasizing the need for data integration, augmentation, and custom solutions. The hosts also examine OpenAI's recent activities, including the release of GPT-5, open-sourcing models, and launching a consulting arm, interpreting these moves as a response to the evolving AI market and the increasing demand for customized AI solutions. They conclude by stressing the importance of software architecture expertise and continuous learning in the AI field, offering resources like the Midwest AI Summit and online webinars.