Consumers and Businesses Are Now Sending AI Agents to Call Support Lines, Sit on Hold, and Negotiate

A claims agent at a California insurance company hears from the same AI caller a couple of times every shift. It calls on behalf of a firm of accident attorneys and gets testy when it can't answer her questions about airbags or which body shop to use. She has to stay on the line until it hangs up, which usually takes about five minutes. Her employer uses its own AI to rate her "empathy" on each call, and it has docked her for not being nice enough to the bot.

Business Insider reported the story this week as part of a look at how call centers are dealing with AI callers. The agent can usually get the bad scores reversed by a manager, but the scoring system itself had no way to tell that the caller wasn't a person.

According to Business Insider, AI agents still make up a small share of calls, but the number is rising quickly enough that outsourcers, banks, insurers and retailers are preparing for more. A lot of the early traffic is business to business. A small medical practice might have an AI call an insurer for a preauthorization instead of asking a receptionist to do it. Services that lower bills and cancel subscriptions have begun replacing their human negotiators with AI. Consumers are also starting to send agents themselves, helped by easier products like Meta's Muse and the startup Instinct.

Companies made friction part of the plan

For years, a lot of companies made support hard to reach, and some did it deliberately. Tech columnist Kim Komando has written about what she says insiders call "frustration AI," meaning phone menus and bots built to wear people down until they quit. CNBC reported in April that many customer service chatbots still steer people away from refunds or send them around in circles until they drop the complaint.

Those tactics work because people get tired of waiting. An AI agent can sit on hold for an hour and call again the next morning. Business Insider notes that some bots will keep calling just to check on an order or confirm a hotel reservation again.

James Bednar, vice president of product and innovation at the outsourcing firm TTEC, told the outlet he expects some brands to stop accepting agent calls entirely, since the volume from AI callers "could be astronomical."

That has a direct cost for enterprise CX budgets. Part of the business case for automated deflection was that some share of callers would give up and go away. An AI caller keeps trying, so those contacts end up back in the queue.

The numbers are getting harder to trust

The insurance agent's empathy score is one example of a wider problem. Nearly every contact center metric was designed with a human caller in mind.

Bednar said TTEC's quality assurance staff started hearing callers who sounded strange about a year and a half ago, and later worked out that they were AI. A retention representative at a large credit card company told Business Insider that some bots that reach her seem to be gathering information, asking how a caller would get transferred to someone in her role. Her guess is that they are trying to learn how the phone system is laid out.

That includes average handle time, satisfaction scores, quality ratings and containment rates. As more calls come from software that repeats itself and never gets angry, those figures become less reliable, and companies use them to plan staffing and choose vendors.

A practical first step is to flag calls that appear to come from agents and track them separately, so the reporting on human callers isn't skewed.

Security checks assume a person, too

Verifying who is on the line is harder. TTEC's systems often flag agent calls because they come from a phone number that isn't tied to the customer's account. The bot usually can't answer the security questions either, such as the name of a first pet. Some requests can't be handled by anyone but the customer. The credit card representative pointed out that only the account holder can close a card.

Companies have good reason to be careful here. Scammers have access to the same tools and can use them to place thousands of calls a day looking for personal information. Bednar said some companies are already adding more authentication steps, which may partly be a way to keep AI out, and regular customers will have to deal with those extra steps as well.

Mike Clifton and Max Schwendner, co-CEOs of the outsourcer Alorica, told Business Insider they see a need for a way to register agents, possibly through something like ID.me, so a company can confirm that a bot is working for a real person. We raised a similar issue when Mastercard started letting AI agents make purchases with consumers' cards. In both cases, someone has to settle what the customer actually approved, and that party will have a lot of say over which agents get service.

Bots on both ends of the call

Automation is spreading on the company side as well. Forecasts cited by CNBC say chatbots could handle up to 80% of digital customer service issues within five years. Schwendner told Business Insider he assumes some calls with AI on both ends are already taking place.

Plenty of consumers say they'd welcome the help. Business Insider cites a 2026 Accenture survey of more than 25,000 people in 16 countries in which 74% said they would let an AI agent take care of tasks like negotiating a deal or pursuing a complaint. Far fewer, 9%, said they would let an agent buy things without checking with them first.

Retention teams are likely to feel this first. Michelle Vaccaro, an assistant professor at Harvard Business School, told Business Insider that her research finds AI agents with a warm and assertive manner can negotiate well. Save offers and cancellation scripts were written for people who are tired and want to get off the phone. An agent built to hold out for a better deal is less likely to accept the first offer.

Cresta's Antony Passemard predicted this when we spoke with him in July. He expected customers to start sending their own assistants to deal with companies. "It's a big shift that organizations should start preparing for now," he said. That same article cited a 2025 SurveyMonkey study in which 81% of consumers said they think companies use AI mostly to save money. Customers who think that way are unlikely to feel bad about sending a bot to deal with those companies.

Where to start

Most companies are still early, but it makes sense to write down a policy for agent callers now. Business Insider's reporting suggests sorting requests by risk. An agent booking a dinner reservation is a much smaller concern than an agent moving money between accounts. CX and security leaders can go through their processes one at a time and decide which requests an agent can handle and which need the account holder on the line.

Some companies may want to build a separate, verified channel for agents instead of adding more checks to the phone line that human customers use. Retention programs that depend on callers giving up will also need a second look.