Elon Musk's comment that humans are underrated (Humans replace robots at flagging Tesla plant, 17 April) doesn't come as much of a surprise, even though his company is at the forefront of the technological revolution. Across industries, CEOs are wrestling with the balance between humans and increasingly cost-effective and advanced robots and artificial intelligence. However, as Mr Musk has discovered, the complexity of getting a machine to cover every possibility results in a large web of interconnected elements that can overcomplicate the underlying problem. This is why so many organisations fail when they try to automate everything they do. Three key mistakes I see time and again in these situations are missing the data basics, applying the wrong strategy, and losing the human touch. There are some clear cases where automation works well: low value, high repetition tasks or even complex ones where additional data will give a better outcome, for example, using medical-grade scanners on mechanical components to identify faults not visible to the human eye.