Opinion: Why can’t we just use AI?

Gabriel Cassady

“You’re looking at around $6K, all in” are not words you want to hear from any professional, but especially not from a plumber standing in your backyard. Unfortunately, our sewer line is failing, and fixing it is going to be — pardon the crude pun — crappy.

Naturally, the AI-obsessed corner of my brain piped right up: Gabriel, why pay a contractor? Rent a backhoe. Buy the pipe. Let ChatGPT guide you. It can explain trench depth, pipe slope and code requirements. It’s a big ditch and some pipe; how hard could it be, right?

If you’re a plumber, you’re probably already laughing or screaming at your screen.

But believe it or not, this kind of thing is already happening, every day, all across the world. Folks are using AI to fix their cars without mechanics, interpret contracts without lawyers, treat conditions without doctors and complete projects without contractors. I am absolutely not recommending you do any of that. But I know it’s happening because I have witnessed it happening. And the results vary wildly.

Beyond my own experiences, McKinsey’s 2025 global survey found 88% of respondents’ organizations were regularly using AI in at least one business function, but 51% of those organizations using AI reported at least one negative consequence, with nearly one-third reporting consequences from AI inaccuracy. The Stanford Social Media Lab and BetterUp surveyed 1,150 U.S. desk workers and found 40% said they had received AI-generated “workslop” from colleagues in the past month alone, and worst of all, they reported spending nearly two hours cleaning up each instance on average.

Setting aside economic, ecological and other AI-related concerns, the biggest problem with using AI for expert tasks in which you have no expertise isn’t what you know or don’t know about the task. The real problem is that you don’t know what you don’t know. And more often than you realize, what you don’t know could literally kill you.

It isn’t that AI can’t tell you accurate and useful things; it can tell you plenty. The problem is that you may not even know what questions to ask or what context to give, and, crucially, you may not be able to judge whether the AI’s response is correct or just confident-sounding hokum. In my case, I am not a plumbing professional, and the cost of being wrong could be catastrophic and potentially permanent.

Generative AI tools like ChatGPT, Claude, Gemini or Grok shine when mistakes are cheap and straightforward to catch and correct before anything bad happens: drafting an email you’ll read before sending, brainstorming product names, summarizing a report you’ll double-check anyway. Using AI gets dangerous when the opposite is true, when mistakes are expensive and hard to spot. For me, sewer line replacements fall into that category. And, for most people, so do contracts, tax filings, hiring decisions, your clients’ (and the public’s) perception of your businessee and anything directly touching somebody’s health or safety.

That said, I know it’s not easy to resist the allure of increasingly powerful and pervasive generative AI models. As an AI and automation consultant, I struggle with it myself, which is why I’ve developed a handy mnemonic to structure my decision-making. The next time you’re trying to decide how heavily to lean on AI for some task, remember the acronym “SLOP” — i.e., Stakes, Liability, Opportunity costs, Proficiency:

• Stakes: What are the potential costs of failure? What are the likely costs of failure?

• Liability: Who is responsible for success or failure? Who takes the blame or gets the credit?

• Opportunity costs: What might you be missing out on by handing the task to AI? (For example, if you want to connect with someone online, having AI write all your messages is probably counterproductive.)

• Proficiency: How familiar or experienced are you with the task at hand? Can you reliably spot mistakes, inaccuracies or other issues?

By keeping this framework in mind, you can sidestep the messes too many folks are thoughtlessly wandering into these days.

Anyone who knows me knows I’m a big fan of generative AI, so I’m not saying you should never use it. To me, it’s the most powerful utility humanity has ever invented, on par with other general-purpose technology like electricity. And much like electricity, unwise or improper use of AI can be damaging and even deadly.

Nevertheless, I can easily imagine a better world in which AI is used to help humans do more and better things together rather than deepening divides and doing harm. We are at an inflection point as a species; the choices we make regarding AI today, as individuals and as a society, will have deep and lasting impacts on the world our grandchildren inherit. I pray we choose wisely.

As for the sewer line? I’m calling the pros.

Gabriel Cassady co-owns local marketing and communication agency 2 Oddballs LLC with his wife, Kylie, and additionally provides AI consulting services. He can be reached at gabriel@2oddballs.com.