Anyone who has called a restaurant to book a table or to tell them you’re running late knows the experience.
It can be particularly frustrating to troubleshoot a product that arrived broken from a company where AI use is the dominant means for service.
Chances are, you didn’t speak to a person. You talked to a chatbot — and if you tried to reach a human, you may have found it nearly impossible.
This isn’t an accident. It’s the culmination of a decades-long structural shift in how companies think about customer service. The arc is familiar. In the 1990s and 2000s, companies offshored their service centers — first to India, then to the Philippines and beyond — chasing labor costs that were a fraction of domestic rates. Companies saved millions that way, but the quality varied enormously.
Agents were often poorly equipped to handle anything context-specific. I once had my luggage delayed for days during a major snowstorm, and the agent kept insisting it would arrive that evening — even though the airports were closed and the roads were impassable. The agent didn’t account for the weather.
Then came the interactive voice response era — the phone trees designed less to help you than to wear you down before you reached a human. Customer service stopped being a relationship and became a cost center.
Now generative AI is accelerating the trajectory faster than anyone anticipated.
Large-scale disruption
The numbers are stark. The U.S. customer service sector employs an estimated 2.8 million people. Industry analysts project that up to 80% of those roles are candidates for automation — not eventually, but within this decade, according to research by Demand Sage. The business logic is straightforward. Human customer service can cost upwards of $3 per minute to deliver, according to Nextiva.
AI handles the routine stuff — order status, password resets, returns, policy questions, basic troubleshooting — for a fraction of that, around the clock, with no hold times.
Phone agent pricing models
For most companies, the financial case is overwhelming. In all fairness, things did advance significantly during the 2024–25 surge of generative AI. Responses got more natural. Bots could actually take action, not just point you to a list of frequency asked questions. Resolution rates on routine issues climbed.
But the improvement didn’t close the gap between what customers wanted and what they got. In some ways it widened it — because expectations rose faster than capability did.
Growing frustrations
Research from analysts like Gartner and Forrester put the “prefers human” number anywhere from 60% to 80%, depending on how the question is asked.
That preference has not softened as AI has improved. The frustrations cluster around a few patterns. Getting stuck in loops. The AI doesn’t understand what you actually need, keeps offering variations of the same wrong answer, and there’s no obvious escape hatch. This is especially infuriating when you’ve already tried the self-service options — that’s why you reached out in the first place. Being walled off from humans.
Many companies have made reaching a live agent deliberately difficult — buried phone numbers, the absence of a “talk to a person” option in chat, opaque escalation paths.
Customers recognize this for what it is: a deliberate design choice that prioritizes cost savings over the relationship. Have you ever tried to speak to someone at Google or Meta? It’s nearly impossible — and that’s the model other companies are quietly adopting. No memory, no context.
Even with all the AI progress, customers often have to re-explain their problem when they get escalated to a human, or when they return to the same chat the next day. The continuity that would make AI support actually feel intelligent isn’t there yet.
Two-tiered future
Here is where the trajectory leads, and why it matters beyond the frustration of any single dropped call.
I envision a world where AI bots become the primary customer service mode for most consumers. But I also envision models emerging where customers will be able to pay for human service — or where affluent, high-value customers get human service as a built-in benefit.
The gap between “talk to a bot” and “talk to a person” becomes the new class divide in service. This isn’t hypothetical. Airlines already charge nuisance fees for phone bookings. Banks tier their service levels by account value.
Some platforms — such as Google and Meta — have effectively eliminated consumer customer service altogether. The AI revolution doesn’t disrupt this logic. It accelerates it. Costs drop even further, which justifies more aggressive automation across the mainstream, while the economics of premium human service become more attractive as a way to capture high-margin customers.
The divide won’t map neatly onto wealth alone. It’ll also separate the digitally fluent from those who aren’t, and the high-value customer from the ordinary one. Older customers, customers with limited English, customers dealing with emotionally charged or genuinely complex problems — these are the people most dependent on human support and most likely to find it unavailable.
Consequences for Maine
The effects aren’t abstract. Maine has historically been a prime state for customer service outsourcing, with an estimated 10,000 to 12,000 jobs in the sector and roughly 284 customer service and telemarketing businesses registered in the state, according to IBIS World.
These are real jobs, often in communities without a lot of alternatives. As AI-driven automation continues to compress demand for human agents, the regional economic consequences will be concrete and concentrated.
Bottom line
Customer service is in the middle of a real structural shift, not just a tooling upgrade. AI will handle most routine interactions, and for simple queries it’ll do so faster and more cheaply than humans ever could. That’s real progress.
But the companies that treat AI as cover for eliminating accountability — hiding escalation paths, removing human access, degrading the experience for everyone outside their most profitable accounts — are making a short-term calculation with long-term consequences. Trust is hard to rebuild once customers feel they’ve been deliberately abandoned.
The companies that get this right will use AI to handle the routine so that human agents can focus on the consequential — the complex, the emotional, the situations where judgment and empathy actually matter.
That’s the version of this future worth building toward.
