When an AI tool is usually right, people tend to accept its suggestions without checking them. How can organisations make sure that the professionals using such tools keep thinking critically and remain genuinely in control?