Opinion: AI: Replacing workers or a useful tool?

John Taylor

It’s getting pretty difficult to know what to think about artificial intelligence these days.

On the one hand, it sure looks like we’re on the brink of a full revolution. Take an article by Matt Schumer published earlier this year on the social platform X and in Fortune magazine called “Something Big Is Happening.” In it, Schumer relays his experience in the AI space, emphasizing how capable AI has become in just a few years and how quickly it’s now improving. His essay reverberates with a sense of fear and awe, including an unsettling conclusion: “If AI is smarter than most PhDs, do you really think it can’t do most office jobs?”

Pretty scary stuff. It’s probably one reason why political leaders at the federal level have been calling emergency meetings with tech leaders this summer. It’s also been the justification cited by companies for recent rounds of office layoffs.

But then there’s this, from Fortune: “[Experts] say companies with the highest gains were those using AI as a form of ‘people amplification,’ implementing the technology to make workers more productive rather than outright replacing them.’”

Let that sink in: In the midst of all the hand-wringing and doomsaying, there’s a parallel sense emerging that AI is probably just another tool. Right now, that seems about right. In my experience so far, AI sadly feels longer on artifice than it does on intelligence.

To illustrate: I interact quite a bit with QuickBooks Online, a web-based accounting platform that Intuit markets for its automated workflows and AI agents. Users can connect their bank and credit card accounts to the program, and QuickBooks will import transactions and recommend how to record them in the books. They can also view AI-generated insights about income and expense trends, customer receivables and vendor payables. A big part of the entire value proposition is that the software can turn hours of bookkeeping into just a few clicks per day.

As bookkeeping software, it’s fine. As an AI workhorse, it’s a little underwhelming.

The software can get all over the place in trying to code transactions. Part of the problem can be how the bank describes transactions. Part of the problem can be how the user originally set up the chart of accounts. But part of the problem is that the software has no idea what that ACH charge from Walmart could possibly have been for.

When the AI analyzes your finances, things aren’t a lot better. It might point out a year-on-year increase to utilities expense; what it can’t know is whether it was due to unseasonably warm weather last month, because of an undetected water leak or simply thanks to its own bad bookkeeping (see paragraph above). It can tell you that 14% of your customers have overdue balances, which is almost less helpful than just running an AR aging report to find the culprit customers yourself. It doesn’t have a way to see someone accidentally spilling raw materials during a production run; it only knows that “your cost of sales has increased 7% in the past week” and might offer that you ask suppliers to charge you less for those raw materials all over the floor. In short, AI might be able to get the idea behind how accounting works, but it hits a wall when it tries to engage with the real, lived world of day-to-day business.

And that’s the problem: AI impresses PhDs because it seems able to understand, explain and perhaps even advance pretty complex ideas. But, to answer the question at the beginning of this article: No, I really don’t think AI can do an office job very well.

Don’t let the big fears about AI become overwhelming. But also: Keep an eye on things. To an owner or manager: Stay aware of what’s out there and keep asking how it might improve workflow. Adopt the tools of AI one at a time, one implementation at a time, and be clear (including with yourself) about what you expect. Be clear-eyed about whether it’s working. Use AI to supplement work, not as a substitute for workers, and hold employees accountable for the information used for reporting to you. And if your business strategy relies more on your conversations with ChatGPT than on your interactions with peers and the folks at your company, know the risk of interacting with an AI that is programmed to suck up to you.

Right now, there’s a lot being written about AI. Some of it is hyperbole, even when the author doesn’t mean it to be. A little of it is probably in the service of covering up earlier bad business decisions. But a lot of it points to a truth: Technological advances tend not to retreat in the face of stubbornness, and businesses stand to gain from AI by viewing it as nothing more than a humble tool in the hands of their hard workers.

John Taylor is a consulting specialist at Elliott, Robinson & Co. He can be reached at jtaylor@ercpa.com.