I Scammed My Internet Provider to Try to Lower My Bill | Invesloan.com

My internet bill crept up by about $15 a month over the summer. I know I can likely lower it by comparing rates to a competitor company, but I’ve been pushing off that phone call because I’m exhausted from dealing with automated systems, long call wait times, and automated voices that can’t decipher what I’m saying when I beg to speak with a human.

It all seemed more work than it was worth. But what if I could outsource this chore, for free?

Using a generative AI voice tool, I was able to turn myself into an agent. Agentic me was tasked with trying to get my internet price lowered and programmed it to be short on patience, leading to a bossy tone that made me cringe and feel sympathy for the human worker on the other side. Agentic me often restated some version of what the customer service rep told it, then said it was disappointed, and started hallucinating lower rates offered by competitor companies to compare prices. “I’ve been a loyal customer for a while now and I don’t think it’s fair that new customers get better deals,” my agent snapped. “It sounds like you’re telling me I should just go ahead and cancel.”

In the AI boom, just as customers are straining to get human representatives on the line, human representatives are now straining to decipher if they’re talking to human customers. Call center workers are up against convincing deepfake tech as the realistic distance between human and agentic callers narrows. As AI tools for consumers become widely available, there’s also opportunities for people to use them not just for malicious fraud, but to troll call centers by sending agents to attack them en masse and waste time. Or, as in my case, as a regular, frazzled customer turning to them to make calls on their behalf.

“We are in the early days of a large emerging problem,” says Brian Levin, chief customer officer at Reality Defender, which makes software for governments, financial institutions, and other businesses to detect the likelihood that content was generated by AI in real time (Reality Defender helped me navigate the process of making myself an agent for this story). And the development of agentic callers “is moving much faster than these contact centers typically do.”

“Across industries, there’s a surge in AI voice agents targeting customer support service lines, driving new forms of fraud while dramatically increasing call volumes and operational strain,” says Emily Fontaine, global head of venture capital at IBM, a partner and investor in Reality Defender, tells me in an email.

Patrick Carroll, founder and CEO of deepfake detector ValidSoft, tells me in an email that call centers are increasingly fielding calls from agents that attempt to thwart security protocols and authentication methods. “Even companies with strong defences are seeing rising call volumes as voice automation tools become more accessible and sophisticated,” he says.

AI is a catch-22 for customer service. A report released earlier this year by the IT research firm Gartner predicts that AI agents will resolve 80% of common issues for customers by 2029, potentially cutting operating costs by 30%. Another Gartner survey of customer support leaders found that 85% of them are interested in or using AI chatbots, but only 11% report that the investment so far has “mostly” or “fully” met expectations. As they try to figure out how automation in customer service works best, fraudsters and callers who want to use agents themselves could upend the model, and trigger an AI vs. AI arms race.

AI is a catch-22 for customer service.

The easier it becomes to have a no-nonsense, hardline negotiating AI version of you call up your cable provider, your dentist, or Williams Sonoma, the number of customers creating AI versions of themselves will multiply. Daniel O’Sullivan, a senior director analyst at Gartner, tells me that “customers want to outsource these types of requests,” and companies will eventually “have to get with the program.”

But, O’Sullivan adds, allowing agentic callers poses risks for companies: They need to find ways to evaluate calls for fraud, may receive more calls seeking service as the lift to deal with minor billing or service problems lowers, and they could lose out on building relationships with customers, something that human agents do when they go above and beyond for callers.

Attempts to thwart call waiting aren’t new. About half of customers say they try third-party channels, like Google, Reddit, or ChatGPT, to try to resolve a service issue before they even bother contacting the company, according to another 2025 survey from Gartner. In 2010, a startup called LucyPhone tried to best call waiting. It allowed people to hang up on customer service, and then receive a call back when it was their turn in line. The Reserve with Google tool can book restaurant reservations on your behalf. DoNotPay, a startup that uses AI to help people fight fees, tickets, and search for money users may be owed, has gained attention for successfully acting as an advocate for individuals against big corporations that often have roadblocks to contact.

Setting the agent up to do my bidding took some finessing, and about an hour’s worth of work to clone my voice from audio in a radio interview I did, and then program the necessary info for my account, so my agent knew it like the back of its cold, coded hand. It wasn’t a time saver, but this voice generating tech has become remarkably more sophisticated and simpler than when I tried a similar experiment with my bank just six months ago. Back then, I had to painstakingly generate each sentence and all of the information I might need, anticipating the questions I would be asked by the representative before getting on the call. I had to save them as audio files, and then hit play on each quickly when asked. But programming this conversational agent was more like creating a character with a few boundaries and setting it loose to chat in real-time.

“It’s entirely feasible for somebody to do that just as a hobby project just to really make a point,” says Matt Smallman, a call center security expert. “People are a bit nervous about what the future does hold,” when it comes to calling, he says. It’s easy to imagine a world in which these tools become so easy to use, more people would take them up — both to troll call centers, and to handle the mundane and tedious tasks of contacting help lines for banks, healthcare systems, and retail companies.

Agent me was not able to convince the call center worker that I deserved to have my bill lowered back to the promotional rate (the customer service rep told me that the company doesn’t match competitor prices, and once my introductory rate came to a close, there wasn’t a better offer to give me). But it did try to keep the representative on the line, who showed no signs of suspicion that they were speaking to an agent. In a five-minute-long, sometimes tense back-and-forth, my agent repeatedly threatened to cancel service and said the rates offered just weren’t good enough for a “loyal customer” like myself.

When the agent exhausted all options, it asked to be transferred to someone who could get the job done. I quickly cut it off, fearful that the agent might be too effective in canceling my much-needed wifi. The agent had a singular goal, and was willing to stay on the phone indefinitely to achieve it. Given the nuance of the situation, I’m not ready to unleash an agent to do my bidding — I’d rather call for myself and ask nicely.

Amanda Hoover is a senior correspondent at Business Insider covering the tech industry. She writes about the biggest tech companies and trends.