Customer service once meant waiting on hold to talk to a real person. Now, it often means going back and forth with a chatbot before you can reach someone.
More and more, banks, airlines, delivery apps, phone companies, online stores, insurance sites, and even government services make users go through automated chat first. For businesses, the reason is clear: bots cost less, work faster, are always available, and can handle lots of routine questions at once.
This trend is only growing. According to Gartner’s 2025 customer service technology survey, self-service and live chat will likely become the main ways people get help by 2027. That explains why companies keep making chatbots the first stop for support—they now see them as the standard way to start, not just a test.
The business reason is simple: cost
Most companies use chatbots because it saves money. Human support costs more since it needs staff, training, schedules, quality checks, and teams to handle tough cases. A chatbot can answer thousands of similar questions at the same time, never needs a break, and works nights, weekends, and holidays.
The California Management Review article “Chatbot Frustration is Real” says using chatbots can save about $0.70 for each customer interaction compared to human agents, and estimates that chatbots could save businesses over $10 billion a year by 2026.
With numbers like these, it is easy to see why companies are turning to automation. For a big company handling millions of support requests, even small savings per chat can add up and make a strong case for using software instead of people for the first contact.
That does not mean all chatbot use is bad. It means the incentive is clear. If a bot can answer “Where is my order?”, “How do I reset my password?” or “What is my balance?” without a human agent, the company saves money and the customer may get a faster answer.
The problem starts when companies treat every issue as if it belongs in that simple category.
AI can help support workers — when used properly
The best reason to use AI in customer service is not to replace every support agent, but to help them do their jobs better.
In the NBER working paper Generative AI at Work, Erik Brynjolfsson, Danielle Li, and Lindsey Raymond studied 5,179 customer support agents who used a generative AI assistant.
They found that using the tool boosted productivity by 14% on average, and by 34% for newer or less experienced workers.
That finding matters because it points to a better model: AI as support for humans, not a wall between humans and customers.
When AI helps an agent find the right answer, summarize a case, or respond more consistently, it can improve service without making customers feel abandoned.
But when the bot becomes the only visible support channel, users often experience the opposite: more repetition, more dead ends and more frustration before anyone solves the actual problem.
Why users feel trapped
People do not dislike chatbots just because they dislike technology. They get frustrated when the bot gets in the way. The most annoying moments are common: the bot misunderstands the question, gives a vague answer, repeats the same question, will not let you talk to a person, or sends you back to a help article you have already seen.
The Consumer Financial Protection Bureau’s report on chatbots in consumer finance warns that when chatbots are “poorly designed” or customers are unable to get support, there can be “widespread harm” and customer trust can be significantly undermined.
The CFPB also says chatbot limitations can create risks when people are trying to resolve serious financial problems, including disputes, fraud, fees or account access issues.
This is why chatbot frustration feels worse than a regular inconvenience. If a shopping chatbot cannot answer a product question, the user might just leave.
But if a bank chatbot fails during a fraud report, or an airline chatbot gets stuck when someone misses a flight, the customer is not just annoyed. They could lose money, miss important deadlines, or be unable to take actions that need to happen quickly.
The hidden problem is escalation
The worst chatbot experiences usually have one thing in common: getting help from a real person is treated as a last resort instead of an important option.
Companies want the bot to handle as many cases as possible because every transfer to a human costs more. But for customers, being able to talk to a person is part of good service, especially when the problem is urgent, emotional, or out of the ordinary.
The CFPB’s findings are important here because they show that customer harm often appears when automated systems cannot recognize that a person needs real help.
In finance, that can mean failing to understand when someone is invoking consumer protection rights, disputing a charge, or trying to stop a scam-related loss. A chatbot that works well for routine account questions can still be dangerous if it handles exceptional situations poorly.
This is the main trade-off that companies often downplay. Chatbots make things easier for support teams, but they also put more work on customers.
Instead of explaining a problem once to a trained person, users might have to say it several times, click through menus, copy case numbers, start over, and repeat the same details when they finally get to a human.
The future should be hybrid, not bot-only
The best solution is not to get rid of chatbots entirely, but to design them thoughtfully. Bots work well for simple, repetitive, and low-risk tasks like checking order status, finding store hours, resetting settings, or answering basic questions.
They are not as good when the problem needs judgment, empathy, special handling, urgency, or accountability.
This is where the NBER study and the CFPB warning fit together. The NBER research shows that AI can make support teams more productive when it helps workers solve problems faster.
The CFPB points out that badly designed chatbots can hurt trust if they block people from real help. The takeaway is not that AI is always good or bad. Instead, customer service automation works best when it supports human help, not when it hides it.
Why this matters to ordinary users
For most people, customer service is when software feels personal. It is the point where your app, bank, airline, or store stops being easy to use and becomes a system that either helps you or leaves you stuck.
That is why chatbot-first support is so annoying—it often shows up when you are already stressed.
Companies will keep using chatbots because they save money. But the real question is whether they use them to solve problems quickly or just to make it harder for customers to reach a person.
A good chatbot should answer simple questions, know when it cannot help, keep track of the conversation, and hand things over to a human without making the user beg.
A bad chatbot does the opposite: it saves the company money by wasting the customer’s time. That is why chatbots are taking over customer service.
Software costs less than people. But companies that get this wrong may find out that customers do not judge service by how well a bot turns them away.
They care about whether someone, human or machine, actually helps.
