The common mistake with a first automation is picking the most painful process. The most painful process is usually also the most complex, the most full of exceptions and the hardest to define — so it ends in three months of work with no result to show.
The right process to start with meets three conditions: it repeats often, it's almost always the same, and it takes no judgment. Even if each run saves two minutes, a hundred times a month that's already a full working day.
At a trading company we worked with, the first candidate they suggested was quote pricing — a complex process with hundreds of exceptions. Instead, we chose intake of orders arriving by email: three entries for the same order, about a hundred times a month, with no decisions along the way.
The first automation went live within two weeks and ran in parallel with the manual work for another three weeks. Because of that, by the time we reached quotes — the complex process — there was already trust, and it was clear where a person had to stay in the process.
Rule of thumb: if you can't explain the process in one sentence, it isn't the first one. The first one is there to prove it works, not to impress.
The bottom line
Pick the first automation for frequency and simplicity, not pain. The pain comes next, once there's trust.