Accounting is a strange and often very stingy profession. Much of the accounting information managed by AI goes on rails from the transaction stage to formal journal entries. Account entries have protocols and essential checks and balances that forgive the puns.
The current generation of Accounting UA is also under continuous intensive scrutiny for reliability and productivity and is gaining increasing acceptance in varying degrees of difficulty. It is achieving this in the face of a kind of obsessive professional care that makes other sectors look positively serene.
money BEN matter. If you’re an accountant and you log into an account, it’s all yours. The scale of the responsibility is extraordinary and it is brutal. Everything on every account MUST stack up and be reliable before it goes to level C. This is not a sector where “AI destruction” or “AI hallucinations” can be tolerated. It’s a true test of operational business AI at its most basic level.
Yes, AI accounting works and is undoubtedly useful
some routine day-to-day accounting practices of AI are very predictable and are really just doing the hard work. This is definitely a significant plus, ironically “rehumanizing” the work for accountants and reducing the extremely tedious element of digging the ditch of maintenance and maintaining accounts. Accountants seem happy with this approach.
Important note: All this level of work is done under human supervision. It is only automated at the dashboard level, not the actual accounting.
Interestingly, AI is clearly and demonstrably productive and highly efficient in this area. It’s not hard to generate efficiency metrics based on time saved on reporting, organizing work, and even the potentially super-sharp phase of coding spreadsheets.
At that level, AI is winning the efficiency argument, and it’s not actually replacing anyone. It can’t. Many accounting firms are in the process of adopting AI for their core businesses, including tax and other heavy compliance work.
Identified problems with AI accounting and their fixes
It’s just getting things done faster, not checking, asking and verifying, which is the gut-level reality of accounting. This work requires a hierarchical range of expertise that cannot easily be accommodated in AI accounting.
The argument that AI cannot replace accountants has merit and can prove it. AI isn’t doing an intense internal critique of “Do I trust these numbers?” as a higher level accountant would do. It cannot ask where some of the numbers that appear in the account come from.
For example:
Suppose a first and pristine $600,000 appears on a government authority’s balance sheet. The accounts are supposed to be in strict accordance with public accounting practices.
This adorable figure, which just happens to miraculously bring a balance to the accounts, also comes with no explanation, no prior entry, and absolutely no other qualifying information. There is not so much as a subatomic hint where this number came from.
What should AI do about a case like this? How can it be answered, if any? Can it also flag the account entry for attention?
The above example comes from direct experience. This is what accounting is all about. It’s also where old-style accounting asserts itself, for any number of good and potentially expensive reasons. The level of pure fiction that can slip into the accounts can only be understood by accountants.
This level of oversight also happens to be the epitome of best-practice AI management.
The takeaway here is that AI has freed the accountant to focus clearly by better managing the workload. It is very much a matter of opinion whether AI can do forensic accounting at any level. It can provide numbers, but can it track them?
AI and the future of accounting
The big and potentially fatal mistake with endless predictions of the future of AI in accounting it’s assuming that automation somehow does the work. No, it doesn’t and it can’t. Most accountants would agree that complacency is a recipe for fraud. Cheating doesn’t take vacations. Even serious accounting mistakes don’t suddenly become harmless because of automation, whether the mistakes are honest or otherwise. Even basic data entry can do a lot of damage.
AI is clearly going to become a major asset in accounting. This raises more than a few other issues:
Training: Will AI become another “continuing education” element in accounting? Perhaps, although at least the training can be structured to meet the needs of the business. This essential training will have to be factored into future needs as a cost.
AI Agents: This is complicated. Custom AI agents are given a set of functions that will constantly change and may become redundant. At what point should AI agents be modified, improved or retired? In accounts, their roles are critical at the most basic levels.
Compatibility: There are two bandwidths here, and neither is simple. Compliance is not optional. Accounts that don’t match are instantly and rightfully suspicious and lack credibility. AI regulation, in whatever form, is inevitable, whether by AI sovereign rules or from the usual tectonic changes in accounting laws. This means that the AI accounting setup will need to match both.
The comfort zone in accounting only goes so far.
