For years, I paid for an annual appliance-service contract with a major national company. When my washing machine failed, I paid the required service fee and expected the company to honor its agreement.

Instead, I entered a maze.

Technicians came to my home and tried to repair the machine. They were pleasant, professional people, and I do not fault them. But their authority was limited. They could diagnose the problem and install authorized parts. They could not settle the warranty issue, control the supply chain, or stop the company from canceling appointments.

Then I waited weeks for parts while appointments were confirmed, canceled, rescheduled, and canceled again. On one occasion, a technician arrived at my door only to learn the company had canceled the appointment. He advised me to call customer service.

That was the problem. The only human representatives I could reach were the technicians standing at my door. Once they left, I was returned to an automated customer-service system with no practical human off-ramp.

Another major retailer gave me a different version of the same experience. I ordered one large item and several smaller related items online. The large item arrived at my home. The inexpensive accessories were routed to a pickup box nearly 50 miles away.

I was not about to drive 100 miles round trip to retrieve a $6 item. I tried to cancel the smaller items. The system would not allow it. I tried to have them shipped instead. That option was blocked too. When I asked for a person, the AI informed me that no human representative was available.

Different companies. Different purchases. The same message: the machine had made the decision, and no responsible human being was available to reconsider it.

Many consumers recognize this pattern. A chatbot can answer a routine question but often fails when the problem does not fit the menu. The website sends the customer to a phone number. The phone system sends the customer back to the website. The system may be working exactly as designed while the customer’s problem stays unresolved.

Artificial intelligence is becoming a corporate firewall between the customer and anyone empowered to act. The problem is not that a machine handles routine work. The problem begins when the company makes the machine the final authority.

The rush toward automation is understandable. AI can lower costs, answer large volumes of inquiries, and perform work that once required sizable staffs. McKinsey notes it now lets small teams accomplish what once required entire departments. Meanwhile, a Gartner survey found 91% of customer-service leaders faced executive pressure to implement AI in 2026.

Corporate leaders can easily measure reduced payroll, fewer calls transferred to employees, and greater use of self-service systems. What they do not measure as carefully is the work transferred to the customer.

I spent hours repeating information, reviewing emails, preserving service records, rescheduling appointments, waiting for parts, and correcting contradictions the company’s own systems created. The automated process left me to correct an absurd pickup arrangement I had not created. The work did not disappear. It became mine.

The customer is becoming the company’s unpaid employee.

That burden can mean missed work, wasted gasoline, delayed repairs, inaccessible refunds, unresolved warranty claims, and money tied up in goods or services the consumer cannot use. It weighs especially heavily on older Americans, people with disabilities, and those who lack the time or technical confidence to wrestle with automated systems.

The problem is measurable. Qualtrics reported that nearly one in five consumers who used AI for customer service saw no benefit from the experience, and half feared AI would prevent them from reaching a human being. The Consumer Financial Protection Bureau, examining chatbots in consumer finance, warns that customers can become trapped in automated “doom loops” — repetitive and unhelpful exchanges with no off-ramp to a live representative.

That point is crucial. AI can schedule a routine appointment, report an order’s status, or provide standard troubleshooting instructions. It is far less dependable when it must reconcile conflicting repair records, interpret a warranty dispute, expedite delayed parts, or recognize that requiring a 100-mile trip for a $6 purchase is unreasonable.

Artificial intelligence can process the rule. A human being is still needed to recognize when applying the rule produces an absurd result.

The technicians who visited my home demonstrated another problem. The company had not eliminated human beings altogether. It had separated human contact from human authority. The people I could speak with were courteous but lacked power to resolve the larger matter. The authority to correct the process was effectively hidden behind software and policy.

Efficiency is a legitimate business goal. Christians should appreciate good stewardship, productive innovation, and the wise use of technology. But efficiency does not cancel responsibility.

A company that accepts payment assumes an obligation to deliver what it promised. That obligation does not vanish because customer contact has been delegated to an algorithm. Software can generate an apology, but it cannot bear moral responsibility for a broken promise. The executives who authorize these systems remain accountable for what those systems do to customers.

Philippians 2:4 instructs, “Let each of you look not only to his own interests, but also to the interests of others.” A corporation that carefully counts its savings while ignoring the costs and frustration imposed on its customers is looking only to its own interests.

The answer is not to abandon AI. Properly used, it can make customer service faster and more accurate. It should handle ordinary requests and equip human agents to solve difficult ones. Gartner’s research emphasizes that AI and human expertise must work together, with people supplying context, empathy, and judgment.

Every automated customer service system needs a plainly accessible human off-ramp. Repeated failures should trigger an automatic transfer to a person — not another menu, chatbot, or recorded instruction. Contract disputes, significant purchases, disputed charges, repeated repairs, and unusual delivery problems should reach someone with authority to decide.

Companies should measure success by problems resolved, not merely by calls diverted.

OpenAI’s own proposed scorecard calls this “useful intelligence per dollar.” Its first test is whether AI completes work that matters. For a customer support operation, OpenAI suggests that “done” should mean the customer’s issue was resolved. A customer who gives up was never successfully served.

I was not asking either company for special treatment. I wanted an appliance service agreement honored and an unreasonable pickup arrangement corrected. In both cases, the machines could process information but could not exercise judgment, and no responsible human being was practically available.

That should not become the new normal. Businesses should use AI to become more responsive, not more unreachable. Technology may answer the telephone, schedule the repair, or route the package. But when the system fails, someone with a human voice and genuine authority must still answer for the outcome.