Hiring an accountant when the processing is automated

Most accounting interviews were designed to find someone who processes accurately and quickly. There is often a test: code these transactions, find the error, reconcile this account.

Those tests measure something real. They also measure the thing you are no longer buying.

If your processing is automated, the person you need is a reviewer, an investigator and an explainer. Here is how to find one.

What the traditional test misses

A candidate can be excellent at producing correct entries and poor at spotting an incorrect one someone else produced. These are different cognitive tasks — production is constructive, review requires reconstructing an expectation and comparing.

They can also be flawless on a clean test and helpless with the ambiguity of real exceptions, where the right answer depends on information nobody supplied.

So keep an accuracy test if you like, but weight it low. It confirms competence, not fit.

Four things worth testing instead

Whether they catch a plausible error

Give them a completed reconciliation or a set of coded transactions with two errors in it: one obvious, one plausible. The obvious one confirms attention. The plausible one — a transaction coded to a sensible-looking but wrong account, an amount that balances but is wrong — is the real test.

Most candidates find the first. Fewer find the second, and that difference is the single most predictive thing you can measure.

Whether they ask what is missing

Give them an incomplete scenario: an invoice with no purchase order, a payment that does not match, a supplier nobody recognises.

Weak candidates decide. Strong candidates ask questions first — was there a goods receipt, has this supplier been used before, what did the buyer expect to pay. The instinct to establish facts before deciding is exactly what exception handling requires, and it is largely untrainable.

Whether they can explain to someone non-financial

Ask them to explain, out loud, why profit and cash differ — as if to a shop owner with no accounting background.

This is remarkably discriminating. Many technically strong accountants cannot do it, because they have never had to. In a role where the numbers assemble themselves, this is a substantial part of the contribution.

Whether they describe processes precisely

Ask them to describe how invoices were handled at their last employer, then push on the edges: what happened when there was no PO, who approved above what limit, what happened if the approver was away.

You are testing whether they hold processes as vague summaries or as actual mechanisms. The second is what makes someone able to improve a process rather than only operate it.

Questions that reveal more than they look like they do

"Tell me about a time the system was wrong and you caught it." A candidate who has never questioned system output has either not been paying attention or has only ever done data entry. Either is worth knowing.

"What would you check first if the bank reconciliation did not balance?" Tests whether they have a method or guess randomly.

"What is a process at your last job that you thought was badly designed?" Reveals whether they think about how work is organised or only execute it.

"How do you know when a number is wrong before you have checked it?" Gets at calibration. Good answers are specific: I know what this supplier normally bills, this margin is normally in this range.

What to stop over-weighting

Speed. You are not buying throughput any more.

Software-specific experience. It is a two-week problem, and weighting it heavily narrows your pool for no lasting benefit. Judgement transfers between systems; product knowledge does not need to.

Immaculate accuracy on clean data. Real work is not clean. A candidate who is careful with messy, ambiguous input is worth more than one who is perfect with tidy input.

What juniors should be hired for

If you hire trainees, hire them for what they will become rather than what they can process, because the processing is exactly what is going away.

Look for curiosity about why rather than what, willingness to say "I don't know" instead of guessing, and the ability to follow a question to its cause. Those predict a good reviewer in three years. Speed at data entry predicts nothing that will still matter.

And be honest with yourself about the obligation: hiring juniors into an automated function means committing to teach them what the processing used to. See what juniors learn when AI does the entries.

Common questions

What should I look for when hiring an accountant now?

The ability to spot a plausible error in someone else's work, the instinct to establish facts before deciding on an ambiguous case, the ability to explain financial results to non-financial people, and precision in describing how a process actually works including its edge cases. These predict performance in review and exception-handling roles far better than speed and accuracy at processing.

Is a data entry test still useful in an accounting interview?

It confirms basic competence but measures the capability you are no longer buying, so weight it low. A more predictive test is giving a candidate completed work containing one obvious error and one plausible one — most find the obvious error, and the ability to find the plausible one is what distinguishes a genuine reviewer.

Does software-specific experience matter when hiring?

Much less than it is usually weighted. Learning a particular system is typically a couple of weeks, and insisting on it narrows the candidate pool for no lasting benefit, whereas accounting judgement transfers between systems and takes years to develop.

What should I look for in a trainee accountant?

Curiosity about why something happened rather than only what to do about it, willingness to say they do not know instead of guessing, and the ability to follow a question through to its cause. Hiring trainees into an automated function also carries an obligation to teach them deliberately what processing used to teach as a by-product, since they will not build that calibration by volume.

Related: what juniors learn when AI does the entries · when the accountant becomes the reviewer · how AI changes the finance org chart