Eliminate the non-accounting first

Someone asked accountants what they’re actually automating. Not theoretically — what they actually built and use.

The most useful answer came from someone who builds automations for accounting teams for a living.

His framing: eliminate the non-accounting.

Not the analysis, not the judgment, not the professional work. The other stuff. Moving files from inbox to shared folder. Cleaning date formats and extra spaces before data will load. Running the same report every Monday. Copying numbers from one system into another. Writing check marks on printed PDFs to track what reconciled.

That last one he couldn’t believe either. A team exporting from the bank, printing PDFs for each company, checking boxes by hand. He connected the systems and set up an email alert when anything didn’t match. They’ve been getting that email ever since.

What strikes me about this framing: it resets the question from “how do we use AI in accounting?” to “what is the accounting work, and what is everything around it?”

The second category is larger than most people admit.

I’ve seen the same thing — a bookkeeper spending an hour a week downloading statements, renaming files, and moving them into folders. An accountant who reconciles by copying figures into a template because the ERP export doesn’t match the required format. A finance manager who manually updates a tracker every time an approval comes through.

None of that is accounting. It’s handling.

And handling is exactly what automation is good at.

The question worth asking before any automation project: what percentage of this job is the actual work, and what percentage is moving things around so the actual work can happen?

Start with the second part. The first part is usually more interesting than people expect.