You Run Your Business Like a Drug Dealer. Good Luck Replacing Your Accountant With AI.
I’ve seen general ledgers at perfectly legitimate, above-board businesses that would make Pablo Escobar’s CFO beam with pride. And this is exactly why I’m not convinced AI is going to replace all accountants anytime soon. Accounting data is created by humans behaving badly, unpredictably, creatively, and sometimes stupidly.
Here's why AI still needs an accountant—and why the best bean-counters are going to thrive in an AI world.
AI loves predictable, structured transactions. Businesses and their owners are neither. Every SaaS finance company out there is selling agents that predict and code everything, but their software relies on patterns and only has access to the data you give it. Their marketing is great; business owners are told they can totally replace their accounting team since SynerFinanze™ software has agents that can ‘do it all’. But have you seen the havoc a non-finance-minded human can do to a general ledger system?
Example 1: The ol’ “I forgot my corporate card”
Before heading to a company happy hour, your head of Strategy accidentally leaves their company card at the office and asks the VP of Engineering to pay for their portion with the Engineering card. In the best-case scenario, the VP of Engineering puts a memo in the transaction that clearly says, “Strategy budget,” and the AI hopefully overrides the mountain of transactions telling it to code all restaurant spends over $300 on the Engineering card to the Engineering budget.
Do we think that’s all going to happen? Will AI override a mountain of historical data based on one memo that the head of Strategy will probably forget to write because, frankly, they have better things to do than explain to the accounting robot why they drank three cross-functional beers with the team?
Example 2: Payment Roulette
Despite all this AI talk, a lot of businesses are still in love with their checkbooks. Writing a paper check makes them feel in control of their cash, and ironically there is a belief that checks are safer since they’re not “in the cloud.”
What happens when they pick up the wrong checkbook to pay the office cleaner? How is AI going to detect a check written out of a personal account? You could tell the system to expect a cleaning check twice a month and flag it if it’s missing, but that’s an extreme outlier.
I’ve seen businesses pay for the same, recurring expense out of personal accounts, family member accounts, business credit cards, personal credit cards, and IOUs from the vendor.
Credit cards with sign-up bonuses have made payment method roulette commonplace in the SMB space as owners aggressively take out new cards and switch payment methods to trigger the sign-up bonus. I’ve seen ChatGPT subscriptions paid out of 7+ credit cards and banks, making SaaS pollution easier to proliferate if you aren’t flagging the duplicates.
Example 3: AI-Generated Nothing-Burger Transaction Memos
AI doesn't just fail to understand accounting data—it can make accounting data look better while actually making it less useful. A client of mine recently had an event called CAB and wanted me to go in postmortem and search all the memo fields including the word “CAB” to calculate the total budget spent. Can you believe how many “business cab” memos I saw across hundreds of taxi rides?
AI is allowing users to auto-generate expense memos that sound compliant but don’t necessarily tell you anything useful. Employees used to have to tell us WHY they took a cab. Now, AI software can suggest a memo that clears the memo requirement hurdle while providing zero information about why the expense was a proper business expense. These nothing-burger memos make T&E harder to police because the transactions appear legitimate. Employees are stoked because their expense reports are basically done for them, and at first glance, accounting teams are celebrating because they’re finally getting the required memos. But ultimately, they’re getting very little useful data.
None of these problems are insurmountable, but in the end, humans are messy and unpredictable. From a client paying a girlfriend’s rent with business funds to playing roulette with checkbooks for expenses, you’re doing pretty well if you can correctly tag 80% of the transactions without manual intervention.
As much as I’d love the robots to totally take my job so I can relax at my mountain house all day, there’s now a new niche in accounting – AI Robot Herder.
You must be able to combine the messiness of humans with AI’s need for standardized, repeatable transactions. Accountants now need to think more deeply about accounting architecture: how each account is used, how it could be misused, and how to structure it so AI can understand it. Formerly, you could roll with the punches and rely on instinct to code transactions as they came in. Now, you need to go a step deeper and set up your clients’ accounts so they are easily digestible by AI.
Key takeaways for business owners: before you whip out that new credit card or grab your sister-in-law's Venmo because your phone is charging in another room, ask yourself whether the transaction you're about to create will make sense to the AI that's supposed to account for it.
Key takeaways for accountants: dive into your clients’ general ledgers and make changes that allow AI to do more of the work for you. If you have sloppy meal accounts—travel meals, business meals, team office meals, new customer meals—figure out exactly what information is needed to automate them 80% of the time.
I’ve heard plenty of pushback from accountants who say, “The customers are messy” or “there are too many exceptions; we have to do it manually.” But the goal is not perfection; the goal is setting up an accounting architecture that gets you to 80% so you can just review the outliers.
Being detail-oriented and good at following a checklist isn’t enough anymore. You need to become an AI Robot Herder: figuring out which robots can be trusted to roam free, which ones need a leash, and which ones you absolutely cannot leave alone with the corporate credit card.
