4 finance jobs you shouldn't be doing manually anymore

Finance teams have spent years investing in better accounting software, ERP systems and automation, yet a surprising amount of the working day is still devoted to tasks that feel like they should have been automated a decade ago.

An invoice arrives in the wrong entity. A shared subscription needs splitting between departments. A reconciliation is out by $47 and someone spends the next hour investigating.

These aren't necessarily simple problems. They require knowledge of how the business operates and enough judgement to recognise when something doesn't look right.

That's historically been the sticking point with automation. Software handles predictable processes well, but introduce an unusual supplier, missing reference or exception to the normal accounting treatment and suddenly someone needs to get involved.

AI agents are beginning to change that, taking on more of the work and bringing difficult decisions back to the team.

Here are four jobs worth reconsidering.

1. Figuring out where invoices belong

For businesses operating across multiple entities, processing an invoice involves a surprising number of decisions. Which entity? Which account code? What tax treatment?

You've thoughtfully set up a dedicated inbox for every legal entity. Unfortunately, nobody appears to have told your suppliers, who continue to send invoices wherever the mood takes them. Meanwhile, your 110 legal entities have somehow acquired 346 different spellings of their names.

Someone needs to work out where the invoice belongs, check previous transactions and make sure it hasn't already been entered elsewhere.

AI agents can do much of that investigative work, reading invoices, checking supplier records and preparing bills in the appropriate accounting system. When something doesn't add up, they can bring the relevant context back to the team.

The finance team still makes the unusual calls, but it shouldn't need to spend its morning playing invoice detective.

2. Preparing accruals and prepayments

The accounting principles behind accruals and prepayments are well established. Under IFRS, expenses need to be recognised in the appropriate periods, which doesn't necessarily correspond with when an invoice arrives or cash changes hands.

Applying those principles requires judgement around service periods, recognition and materiality. But once the treatment is established, there's still plenty of administration involved.

Identifying transactions, checking contract dates, calculating adjustments, maintaining schedules and preparing journals. Often that means exporting data into a spreadsheet before entering the results back into the accounting system.

AI agents can handle much of this preparation, applying established policies and preparing draft journals with supporting calculations.

Accountants remain responsible for reviewing the treatment, but don't necessarily need to do all the legwork.

Nobody went into accounting because of a deep personal attachment to maintaining a prepayments spreadsheet. Although some of those spreadsheets have been with the company longer than most of the staff.

3. Allocating costs across departments and entities

Consider a $12,000 software invoice that needs splitting across three entities according to an allocation method agreed six months ago.

The calculation might be straightforward. The challenge is finding the right percentages and ensuring the entries make their way into the ERP.

Get it wrong and the GM of your Outer Mongolia division will be on the phone demanding to know why their department is paying for everyone else's software. They may not know the right allocation, but they're absolutely certain it isn't this one.

Multiply that across dozens of recurring expenses and it's easy to see why finance teams maintain elaborate spreadsheets to hold everything together.

AI agents can apply established allocation methods, prepare entries and flag anything unusual. They can also retain the context behind those decisions, so an explanation given this month informs how the same expense is handled next month.

There's not much value in automating a task if someone needs to explain the same allocation for the sixth time.

4. Investigating reconciliation differences

One Sterling customer used to have a "reconciliation ruler". An actual 30cm piece of plastic they'd run down the screen, checking each invoice against a supplier statement line by line.

Several decades of accounting software development and we've arrived at holding a ruler against a computer monitor.

Traditional automation handles straightforward reconciliations well, but exceptions still require investigation. A payment without a useful reference, a supplier under a different name or a transaction missing from another system.

AI agents can increasingly handle that investigation, identifying likely matches, checking supporting records and preparing reconciliations for review.

There's a meaningful difference between deciding how to treat an unexplained payment and spending 40 minutes finding the PDF that explains it.

The reconciliation ruler, thankfully, can retire.

The bigger opportunity

These jobs are repetitive, require business context and are just messy enough that traditional automation has struggled to handle them end to end.

That's exactly what we're building Sterling to take on.

Sterling is an AI assistant accountant that works across existing accounting systems, ERP platforms and spreadsheets. Show it how you want a job done and it takes things from there, recording its work and bringing exceptions back when your judgement is needed.

Importantly, it retains what it learns. Explain how a supplier should be treated or a shared expense allocated and that knowledge can inform the next occurrence.

For years, finance automation has focused on making individual tasks faster. The bigger opportunity is handing off entire jobs so accountants can focus on decisions that actually need them.

There will always be hard calls to make. Finding the bloody attachment shouldn't be one of them.

What would you give Sterling?

Sterling takes on repetitive finance jobs across your existing systems, bringing exceptions back when needed.

See Sterling in action