From Call Centers to ChatGPT: Using AI for Everyday Household Repairs
Once the province of company call centers, is basic troubleshooting now something you can do with an AI assistant in your pocket?
Many years ago, I worked for a company that pioneered electronic publishing via CD-ROM technology. Our initial projects included electronic encyclopedias and U.S. and European national bibliographic databases. Later, we developed network-accessible and portable parts catalogs and repair documentation for a range of industries, including appliances, automobiles, medical equipment, aircraft, and trucks.
Some systems documented the structure of customized long-haul trucks based on each chassis’s unique bill of materials. Those projects drew us into the manufacturing process itself, which made data capture more controllable.
Among the most interesting projects were those for large call centers. These centers helped consumers diagnose and repair their appliances—or, when repair wasn’t possible, referred callers to local service providers.
These projects were fun, but even then it was becoming clear that increasing automation, early experiments with AI, and the shift of U.S. manufacturing overseas would eventually make large, company-operated call centers—staffed by knowledgeable product specialists—obsolete.
Today, many companies have realized it’s often simpler and cheaper to replace entire units rather than attempt repairs. As a result, much of what we built back then for diagnostic- and repair-focused call centers has become unnecessary for many products.
I was reminded of this recently when I found myself using ChatGPT Plus to help diagnose a string of problems with my Mac Mini, my new Canon office printer, my DSLR, and my Whirlpool washer and dryer.
I’ve used AI tools extensively for business and personal applications, but this was the first time I relied on them heavily for “domestic tech support.”
From the experience, I’ve taken away some insights on using one particular AI tool (ChatGPT Plus) for diagnostic and repair applications:
Be explicit and detailed in describing the problem. The more context you provide, the better. Go beyond basic year-make-model descriptions to include purchase date, physical location, and even recent events such as neighborhood power outages. Explaining what you’ve already tried also helps. (Having started out with traditional database, keyword, and full-text search, I find this ability to communicate in whole sentences a gamechanger.)
Expect to refine your inquiry. The conversation evolves step by step. Each action (“remove screw A before screw B”) and its outcome gives the system more information, shaping subsequent diagnostics. You can even add clarifying comments like, “Tell me in more detail how to remove panel X from the frame since I don’t see any screws,” and the system will respond.
Consider how the AI was trained. The AI seems to rely on an internal model or framework for diagnosing common issues across product types, supplemented by details of specific models and refined by user interactions. Your feedback may help shape its guidance.
Recognize gaps in model coverage. Small product variations can significantly affect diagnostic steps. For example, while troubleshooting password and screensaver settings on my M4 Mac Mini, ChatGPT sometimes noted, “This version of the OS changed the system’s behavior.”
Expect frequent safety warnings. This was especially true during washer and dryer repair. Step one was always, “Be sure the appliance is unplugged.” That also made me wonder what my ChatGPT Plus user agreement says about liability!
Take advantage of memory. Being able to pause and return days later to continue the same thread was invaluable, especially when waiting for spare parts or juggling client meetings.
Use portability. At one point, I was crouched behind the washer during a test run, interacting with ChatGPT on my phone via speech-to-text while watching the (dripping) drain pump in operation. That saved a lot of time.
You still have to do the work. And, you still have to get the parts. Twice so far I have been sent either the wrong part by an online vendor or a damaged part, even though AI was correct in identifying the correct part number.
I admit I’m a bit of a do-it-yourselfer. I enjoy getting away from the computer now and then. Doing the washer and dryer repairs myself (instead of buying expensive replacements or hiring a hopefully well-trained technician) gave me the chance to learn new skills—and even buy a multimeter at ChatGPT’s suggestion, which I selected with ChatGPT’s realtime help while standing in the electrical tools aisle at Home Depot.
Such interactions do raise broader questions:
How will this impact employment for trained repair technicians, especially in areas less suited to consumer involvement?
Where do the diagnostic details for specific model variations come from?
What is the cost in terms of electricity and water usage at data centers supporting these transactions? (I live in Northern Virginia where disputes over data center locations are common.)
What about legal liability?
How will AI-assisted repair affect the “right to repair” movement for industrial products?
Will companies prefer—for financial or customer loyalty reasons—to take responsibility for repairing their own equipment, or will they encourage displacement of that responsibility to generic tools like ChatGPT?
For companies that incorporate AI diagnostic tools into their own website chatbots, how will they govern the pont where human involvement is required?
The last three points are important. Many companies factor post-sale repair revenue into their financial projections. In such cases, questions about repairing smaller appliances like washers and dryers may pale compared to issues with complex mechanical and electronic systems.
Still, I do appreciate it when a repair doesn’t require an obscure, manufacturer-only tool!
Copyright 2025 by Dennis D. McDonald
