Use It, Don't Worship It: What the AI Conversation in Accounting Means for Your Business

I read a fair amount of profession trade press so you don't have to, and a recent dispatch from the Institute of Management Accountants' annual conference caught my attention — not because it said anything shocking, but because it said something sensible. The headline framing was that the profession favors AI use over AI abuse. (Accounting Today's coverage is here if you want the full account.) That distinction is doing a lot of work, and I think it's exactly the distinction most business owners I talk to are missing.

So let me tell you what the people who actually run the accounting profession are saying — and then let me tell you what I think you should do about it.

The room wasn't asking "should we use AI." It was asking "how do we use it well."

Here's the part that struck me. According to the officials at the conference, finance and accounting professionals have moved past the breathless "will the robots take my job" stage. The questions in the room were practical and a little skeptical: What are we actually trying to accomplish? What problem does this solve? Is this cost-effective? Can we trust the output?

That's a mature conversation, and it's the right one. Notably, the new chairman of the Public Company Accounting Oversight Board — the body that regulates the firms auditing public companies — used his time to emphasize that professional skepticism and human judgment remain critical even as the technology advances. When the chief referee of the audit world tells a room full of accountants that the machine doesn't replace judgment, you should listen. He's a 40-year veteran of one of the Big Four firms. He's not a Luddite. He's just right.

The conference produced a handful of through-lines worth repeating, because every one of them applies to a contractor in Torrance or a real estate operator in the South Bay just as much as it applies to a Fortune 500 controller:

• AI cost and return-on-investment are getting real scrutiny. Companies are nervous about building entire workflows and headcount plans around tools whose long-term pricing nobody can predict.
• The fundamentals can't be allowed to atrophy. There's genuine concern about training people who can run the AI but can't function — or catch the error — when the tool is wrong or simply unavailable.
• Hype is out; use cases are in. The appetite is for narrow, practical applications that solve a specific problem, not grand theoretical transformations.
• Data quality is the whole ballgame. AI cannot produce a reliable output from incomplete, inconsistent, or poorly governed data. Garbage in, garbage out — now at scale and with a confident tone of voice.

Why this matters more for a construction or real estate business than you'd think

Let me get to the opinion you came here for, because I do have one.

I think the small-business world is about to split into two camps. The first camp will treat AI like a slot machine — feed it a messy QuickBooks file, pull the lever, and trust whatever comes out because it sounds authoritative. The second camp will treat it like a power tool — enormously useful in trained hands, dangerous in careless ones, and useless if the material you feed it is junk.

I want you firmly in the second camp, and here's the unglamorous reason: AI does not change a single one of your obligations. It changes how fast you can do the work, not what the law requires the work to be.

Consider recordkeeping. Internal Revenue Code §6001 and the regulations under it (Treas. Reg. §1.6001-1) require you to keep records sufficient to establish your income and deductions — full stop. An AI bookkeeping assistant that miscategorizes a $40,000 equipment purchase as a repair expense hasn't saved you time; it's handed you an exposure. The same logic runs through the long-running fight over capitalization versus current deduction under the tangible property regulations (Treas. Reg. §§1.263(a)-1 through -3) — a distinction that hits construction and real estate businesses constantly, and one that a generic AI tool will cheerfully get wrong because it doesn't know your facts. Which purchases must be capitalized depends on your facts and your elections, such as the de minimis safe harbor in Treas. Reg. §1.263(a)-1(f), so have a person review what the tool coded.

Or consider your financials. If you report on an accrual basis and you're trying to honor the matching principle — recognizing the cost of a job in the same period as the revenue it generated — then the quality of your output depends entirely on the discipline of your inputs. GAAP doesn't get easier because the software got smarter. A percentage-of-completion calculation on a long-term contract is only as good as the cost data underneath it. The conference takeaway that "data quality is becoming a prerequisite for AI value" isn't a tech-industry slogan. For a contractor, it's the difference between a job-costing report you can bid against and a number that quietly lies to you.

This is where I'll get mildly political, since you asked. I think there's a cultural pressure right now — pushed hard by vendors with something to sell — to believe that software can substitute for expertise. It can't, and the people regulating this profession are saying so out loud. The accounting talent shortage is real, and AI is a genuine and necessary lever for getting more done with fewer hands. But "fewer hands" is not "no judgment." A tool that lets a skilled person do the work of three is a blessing. A tool that convinces an unskilled person they don't need the skilled person is a liability waiting to mature. The profession's instinct — embrace the technology, guard the fundamentals — is the correct instinct, and I'd urge you to adopt it as a business owner too.

What I'd actually do, if you're sitting where you are

A few concrete moves:

1. Use AI for the first 80%, not the last 20%. Drafting, summarizing, organizing, surfacing questions — terrific. Final classification of a transaction with tax consequences, or sign-off on a financial statement you'll hand to a lender — that's where a human who's accountable needs to stand.
2. Fix your data before you automate it. If your chart of accounts is a mess, AI will simply make the mess faster. Clean inputs first.
3. Keep a human who can catch the wrong answer. The error you can't catch is the one that costs you. That's the whole argument for not letting the fundamentals erode.
4. Be skeptical of pricing promises. The professionals worried about uncertain long-term AI pricing aren't paranoid. Don't rebuild your entire back office around a tool whose cost in three years is a guess.

I use these tools every day. They've made me faster and, frankly, better. But the value isn't in the tool — it's in knowing which output to trust, which to verify, and which to throw out. That judgment is the product. The software is just the lever.

Let's make sure your numbers can actually be trusted

If you're adopting new tools and you're not certain your underlying books, job-costing, or tax positions are solid enough to build on, that's exactly the kind of thing worth sitting down for. I'd be glad to talk through an advisory consultation — or to walk you through our financial reporting package, which is built to give you numbers you can make decisions against, AI or no AI. Reach out here and let's set something up.

A note on logistics: if any part of that review involves sensitive financial documents, please send them through our secure client portal rather than email.

This article is general information and reflects my own views and reading of public reporting. It is not legal, tax, or accounting advice, and it does not create a client relationship. Tax provisions and inflation-adjusted figures change; anything cited here should be confirmed against current primary authority for your specific facts before you act on it.