Will AI Replace Credit Controllers? We Read the Actual Data.

Not dodges. Refuses — explicitly.

The Index measures which occupational tasks show up in AI conversations. It does not measure jobs, employment, or who's using the AI. Its own methodology states, in writing, that the data cannot support conclusions about job displacement or job security — because a conversation matching an accountant's task usually isn't an accountant automating themselves. It's often a random person asking an accountant-shaped question.

Sit with that. The only rigorous, monthly, global measurement of AI-at-work says the replacement question cannot be answered from usage data. Everyone claiming otherwise is selling something.

That's shock number one, and it's the most valuable thing in this article: the next time a headline tells you a percentage of finance jobs will vanish, you now know no measured dataset on Earth supports that number.

Three measured facts:

Accounting tasks rank 147th of 718 occupations by AI usage share. The occupations topping the list are researchers, writers, librarians and programmers — because their tasks (searching, drafting, explaining) are what everyone asks AI to do. Accounting-specific work is genuinely in the data, but it's mid-table, not ground zero.

When accounting tasks do appear, humans stay in the loop 55% of the time — collaboration, iteration, review — versus 45% full delegation. Slightly more supervised than the global average.

The tasks that do get delegated got dramatically faster. The 5-hours-to-40-minutes ratio applies across the board.

Data mode off, practitioner mode on. What follows is our professional read after decades in collections — clearly our judgment, not a dataset finding.

Look at which kind of credit-control work matches the AI-heavy task families, and which doesn't:

Notice something about the left column? It's the part of the job every credit controller complains about. The lookup, the letter-drafting, the status-update treadmill.

And the right column is the part that collects the money.

That's our honest read: the measurable AI wave in finance tasks maps almost perfectly onto the drudgery, while everything that determines whether an invoice actually gets paid — judgment, negotiation, the phone, the relationship — is nowhere on humanity's top-50 delegation list. Not because it's protected. Because a language model can't do it.

Here's the spicier truth the replacement panic hides.

The risk to a credit controller in 2026 isn't a robot taking the chair. It's the job quietly becoming letter-processing — spending the day feeding a machine that exchanges correspondence with the debtor's machine, while nobody picks up the phone and DSO climbs.

Because on the other side, debtors are automating too. Two AIs writing each other polite letters is not a collections process. It's a stalemate with stationery.

The credit controllers who thrive in this decade will be the ones who hand the left column to the machine ruthlessly — and spend the reclaimed hours doing more of the right column than was ever possible before. Five hours of record-chasing compressed to forty minutes is five extra hours of phone calls per file, if you choose to spend it that way.

The data's honest answer is "unknowable from usage — and mid-table anyway." Our practitioner answer is sharper: the machine is taking the part of the job you hated, and the part that collects money just became the whole job.

The threat was never the robot. It's spending 2026 exchanging letters with one.