A few weeks ago, Amazon delivered my order but it was completely wrong. I sighed and shuddered knowing I would have to call customer service and talk to a bot for at least ten minutes before getting to a real person.
Despite talking to and testing AI all day, when faced with the wrong package and an AI customer service agent, I usually hang up. I'd rather give up and let the broken thing stay broken or the wrong package stay wrong rather than shout "AGENT! HUMAN!" into my phone or chat box. Maybe you know the feeling?
But recently, I decided to stop surrendering and start testing. Over the last week, I threw every supposed "secret trick" at the bots guarding Amazon, Optimum, Walmart, AT&T and my local electric company. Some of the famous hacks are nonsense. A few work almost every time. Here's what worked best when trying to contact someone with a pulse.
Why AI customer service seems to be getting worse
It's not your imagination — the wall is getting taller. Companies are handing customer service to AI because it's dramatically cheaper than staffing call centers, and the bots are getting better at stalling you in polite loops. The frustrating irony is that most people don't want this.
Surveys of consumers consistently show that most people would still rather talk to a real person than an AI chatbot, and regulators are beginning to pay closer attention to how companies deploy AI in customer-facing roles. A 2025 Gartner survey found that 64% of customers would prefer companies not use AI for customer service at all, while a separate survey found that 93% of consumers prefer interacting with a human over AI. Meanwhile, lawmakers and regulators in both the U.S. and Europe have increased scrutiny of AI systems used in customer interactions, particularly around transparency and accountability.
The good news: the escape hatches are real, because the systems are designed to escalate. You just have to speak their language.
What actually worked
1. The magic words (the most reliable trick by far)
Forget being clever first. The fastest route is usually the bluntest one: say or type "agent," "representative," "human" or "escalate." These are deliberate trip-wires built into most systems. If the first attempt bounces, repeat it. The pattern I kept hitting: after two failed bot replies, restating "I need to speak to a human" flipped the system into transfer mode almost every time.
Verdict: Worked on 6 out of 10 AI customer service agents I tried. It worked on Amazon every time. Start here.
2. Use the words that scare them
This was the upgrade that genuinely surprised me. The bot treats "I have a question about my bill" very differently from "I want to cancel my service" or "dispute a charge."
Companies route the threat of a lost customer straight to a human — and fast. The same goes for "billing" and "legal," which many systems flag as higher-risk and hand off quickly. I started leading with "cancel account" even when I had no intention of canceling, then simply explained my real issue to the human who picked up. They can transfer you internally without dumping you back into the maze.
Verdict: Fastest escalation I found. Mildly cynical, extremely effective.
3. Ditch the app, use web chat
A quieter finding was that the chat widget on a company's website routinely routed me to a person faster than the same company's mobile app. Web chat tends to be set up for the messier, higher-stakes questions, so the threshold for handing off to a human is lower. If you're stuck in app-chat purgatory, just switch to a browser.
Verdict: Underrated. Cost me nothing and shaved real time off every time.
4. Time it right
Obvious in hindsight, but worth stating: the trigger words only summon a human if a human is awake. Calling or chatting during standard business hours got me to live agents noticeably faster, because the system isn't trying to route me to an empty queue.
Verdict: Not a "hack," just reality. Call right when the customer service hours begin.
5. Go public on social media
This is my favorite and forever my immediate go-to solution. When I am truly stuck, like the time my mom burned her mouth on a mug from Vera Bradley — the nuclear option worked: a polite-but-pointed public post on X (and a DM on Instagram/Facebook).
Even if you get an automatic reply at first, plenty of brands still staff their social accounts with real people, and a public complaint is a reputational itch they want to scratch quickly. Something like "@Company, 40 minutes stuck in your chatbot trying to fix [issue] — can a human help?" got a response when nothing else did.
Verdict: Slower to start, but a reliable last resort for the truly stuck. Plus, you're more likely to get "apology" perks beyond just a refund.
A few other options that worked some of the time
You could always "act broken." For instance, stay totally silent for 30–60 seconds, mash 0 or #, or type gibberish like "asdfghjkl" until the bot gives up. And yes — sometimes the system assumes you've got a bad connection or an old phone and kicks you to an operator. When I tried this it only worked twice.
In other words, it's a coin flip, but still worth a try. But be warned, when I did this with my cable company, the bot simply disconnected the call and I had to restart in the menu again. So, treat these options as a Hail Mary, not a strategy.
Another options is the "pick a different language" trick (choosing a language menu option that doesn't match your accent to force a human) is in the same boat — clever in theory, inconsistent in practice, and you may end up routed to a queue you can't actually communicate with.
Another shortcut is to let a service do it. If you'd rather not play any of these games, GetHuman is the old reliable. Its free tools can navigate phone trees, wait on hold and call you back once there's an actual agent on the line, and it keeps a big directory of direct numbers and the current known shortcuts for major companies. The trade-off is that it's ad-supported and you hand over your phone number, so weigh that against the convenience. Use that option if you simply don't have time to wait but really need to speak to someone.
Final thoughts
Here's the thing none of the "hacks" lists tell you: how you behave once you reach the human decides how your problem ends. While it's probably fine to yell at an AI customer service agent, especially after 40 minutes of robot tennis — avoid it. Because when the actual human finally picks up, you might still be tempted to keep going.
But that customer service agent didn't cause the problem and they have enormous discretion over whether you get the bare minimum or the goodwill credit. Be brief, be specific, be kind. It's the most effective "trick" of all, and it's the only one that's also just decent.
The bottom line here is this: the fastest, most dependable path is unglamorous: say "agent," and lead with a word like "cancel" or "dispute," use web chat over the app, and do it during business hours. Save the gibberish and the silent treatment for emergencies, keep GetHuman in your back pocket and stay civil when you break through.
What's your favorite tip for beating the bots and getting to speak to an actual, live human? Let me know in the comments!
Follow Amanda Caswell and stay ahead of the AI curve
