Count the hours before you buy anything

Most tool decisions are made on a demo and a feeling, which is why so many licences sit unused. Before anything is bought or switched on, have somebody watch a real month of commercial work and write down where the hours actually go: the re-keying, the chasing, the rebuilding of numbers that already exist somewhere else. The count does two jobs at once. It tells you whether there is anything worth automating at all, and it gives you the baseline that proves, three months later, whether the change paid for itself.