Should You Let AI Run Your Foreign Currency Revaluation?

Foreign currency revaluation is one of the few month-end tasks where your ERP probably already does the math, and the reason it still eats an afternoon is everything wrapped around the math: the rate you used, the open items that didn't get picked up, the intercompany balance that's off by a few hundred dollars for reasons nobody can explain. That wrapper is where AI earns its keep, and I'd argue it's a better first project than most of what gets pitched, because the rules are mechanical, the evidence is a rate table and a subledger, and you can check every number it hands you. I've watched teams spend a full damn day on this at close and I think most of that day is avoidable.

What does foreign currency revaluation actually involve at month-end?

You take every open balance denominated in a currency other than the entity's functional currency (foreign-currency payables, receivables, cash, intercompany loans), restate it at the period-end rate, and book the difference to the income statement as a gain or loss, usually with a reversal on day one of the next period. The functional currency is, in the IFRS Foundation's wording for IAS 21, the currency of the primary economic environment in which the entity operates, and the standard also covers which exchange rates to use and how to report the effects of rate changes, which is the part that trips up every new multi-entity controller and generates a bullshit amount of confusion for something so mechanical. Translating a foreign subsidiary into the group's presentation currency for consolidation is a related but separate step, and mixing the two up is the most common mistake I see in first-pass workpapers.

Why is revaluation a good job for AI?

Because it's a rules problem with a paper trail. The inputs are a list of open foreign-currency items, a rate table, and the prior period's balances, and the output is a short set of journal entries with a calculation behind each one. A model working from those inputs can build the schedule, compute the revaluation, draft the entry with a memo, and flag anything that looks off, and you can recompute any line in a spreadsheet in under a minute. Where I'd be more cautious is judgment calls like determining an entity's functional currency or deciding whether a long-term intercompany loan is effectively part of the net investment, and those stay with a person who knows the structure.

It also matches where the market sits. Across the 80,000+ postings we track, about 73% of the senior roles that own the books (controllers, accounting managers, senior accountants, assistant controllers) name month-end close or reconciliation work, so the close is still the core of the job, while only about 1 in 6 postings mention automation at all and fewer than 1 in 10 name AI directly. Multi-currency work doesn't get its own line in most of those descriptions, which is a good reason to be the person on the team who has quietly made it faster.

What should the AI actually do, and what should it leave alone?

I'd break it into three jobs. The first is a completeness check: compare the list of foreign-currency accounts and open items in the subledgers against what the ERP revalued, and surface anything that fell through, like a vendor set up in the wrong currency or a bank account added mid-quarter. The second is rate validation: confirm the rate source, the rate date and the quote direction (is that USD per EUR or EUR per USD) on every currency, since an inverted rate is the sort of error that looks plausible until a variance review. The third is commentary: draft a short explanation of the gain or loss by currency, tying the swing to rate movement on the largest open balances.

What I'd leave alone is posting, rate selection policy, and anything touching hedge accounting. The model drafts, a person with release rights posts, and your treasury or controller decides the policy. If you're already thinking about continuous reconciliation, the same staging-and-review pattern we use for rolling reconciliations applies here, and the intercompany side is covered in our piece on intercompany reconciliation and consolidation.

Where does AI revaluation go wrong?

In three places, from what I've seen. It will happily apply a single rate to everything when your policy calls for average rates on the income statement and closing rates on the balance sheet, so spell out which rate goes where in your prompt and check one example by hand. It can miss that an intercompany balance is booked in different currencies on each side, which creates a difference no amount of revaluation will clear. And it inherits whatever mess is in your subledgers, so a vendor coded to the wrong currency three years ago gets revalued wrong every month with total confidence. That last one is a pain in the ass to find the first time, and it's worth finding.

The audit trail matters too. Keep the rate table, the source of each rate, the model's draft and your edits together in the close folder, since somebody will eventually ask where the number came from, and we've covered the control side in how to let AI into the close without blowing up your SOX audit.

Do you need new software for this?

Probably not to start. Most mid-market ERPs, including NetSuite and Sage Intacct, handle the revaluation calculation natively, so the AI layer can sit on top of exports from the system you already have, and that's the cheapest way to find out whether it helps. If you're running several entities and currencies on something that doesn't revalue natively, that's a legitimate reason to look at a platform change, and our guide to choosing multi-entity accounting software walks through what to compare, with the platforms listed in our software directory.

What I'd do

Take last month's close and rebuild the revaluation schedule for your two largest non-functional currencies, using an AI assistant and exports of the open items and the rate table. Compare its totals to what your ERP posted, and chase every difference, because each one is either a mistake in the model or a gap in your setup, and both are worth knowing about. Do it once more next month, and if it keeps matching, add the completeness check and the commentary. I wouldn't hand it posting rights, and I also wouldn't wait for a vendor to package this for you. If you're hiring for or looking at multi-entity finance roles, the Audit Friendly job board links straight to employer career pages.

Frequently asked questions

Can AI do foreign currency revaluation?

It can build the schedule, compute the revaluation from your rate table, draft the journal entry and flag exceptions, and a person should still review and post. Most ERPs already calculate revaluation natively, so AI is most useful for completeness and rate checks.

What is foreign currency revaluation in accounting?

It's the month-end restatement of open balances denominated in a foreign currency at the period-end exchange rate, with the resulting gain or loss recorded in the income statement, usually reversed at the start of the next period.

What is the difference between revaluation and translation?

Revaluation restates foreign-currency balances inside an entity's own books, while translation converts a foreign subsidiary's financial statements into the group's presentation currency for consolidation.

How do you check that exchange rates were applied correctly?

Verify the rate source, the rate date and the quote direction for each currency, and recompute a sample of lines by hand. An inverted rate is the most common error and usually shows up as an unexpectedly large gain or loss.

Should an AI agent post revaluation entries?

I'd keep a human on release. Drafting into a review queue keeps segregation of duties intact and gives auditors a clear approver for every entry.

Pull your two biggest foreign-currency balances, rebuild their revaluation by hand and then with an AI draft, and see where the two disagree - the disagreements are your to-do list.