Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
You’ve been there. You’ve said “human” once, twice, three times into the phone, each time a little more clipped, each time getting a cheerful redirect back to the same menu. By the fourth attempt, the word comes out differently.
According to a Parloa survey of 1,001 US adults, reported by Futurism, 17% of customers eventually cross that line. They type or say something into a chat window they’d never direct at another person. That number looks small on a data slide. It isn’t.
Parloa sells conversational AI for customer service, so take the study for what it is: a vendor confirming that its market exists and its competitors are underperforming. Naming that upfront is the honest move. It also doesn’t make the data wrong.
More than half of the adults surveyed try to bypass customer service chatbots entirely. Nearly 44% repeat “human” or “person” trying to reach a live agent. Many give up after three minutes. Only 13.6% trust AI to handle a complex request.
Then there’s the other number: 85% say they’d welcome AI that actually resolved their issue.
That’s the important metric. Not 13.6% versus 85% as some kind of contradiction. It’s the space between what experience has trained people to expect and what they’d actually want if they got it. These aren’t two different populations. They’re the same people, and their behavior is a rational response to a track record.
A chatbot isn’t installed by accident. Someone decided to put one there. Someone chose what it could and couldn’t do. Someone set the metric it would be evaluated against.
In most cases, that metric wasn’t resolution. It was deflection: how many contacts got routed away from a live agent, how many callers never reached a phone rep. That math looks fine on a cost spreadsheet. It produces a very specific kind of failure inside a customer’s experience.
When someone types something profane into a chat window at 10 p.m. trying to fix a billing error, they’re not having a technology meltdown. By that point, they’ve already been deflected, restated the problem three different ways, and exhausted what the bot would give them. The profanity isn’t the start of a bad experience. It’s the conclusion of one.
The 17% who curse at the bot are the ones who stayed long enough to be frustrated. In a strange way, they’re the most committed customers in the data set because they were still trying.
I build with these tools. As an AI keynote speaker who works with teams thinking through how they’re deploying AI, I keep coming back to the same question: did the deployment actually solve the problem?
A chatbot built to serve customers and a chatbot built to protect the company from customers can look identical in a demo. They behave very differently at 10 p.m. when someone’s billing is wrong.
The design choice reveals the priority. And customers, whether they know it or not, read the priority correctly. They’ve been trained by experience to recognize the difference between a system that’s trying to help and a system that’s trying to close the ticket. Not because they dislike technology, but because they’ve learned what the technology was built to do. More than half skip the bot entirely for exactly that reason.
When more than half of your customers are actively trying to route around the system you built, that’s feedback. When 43.9% are repeating a single word into their phones, that’s feedback. When 17% resort to profanity, the temperature is higher, but the message is the same.
Companies spend real money on NPS surveys, satisfaction scores, and focus groups. Meanwhile, their chat logs contain unfiltered, real-time data about exactly where their service fails. The people cursing at the bot aren’t edge cases to be managed. They’re the most honest respondents a company will ever get, and they’re delivering that verdict for free.
Taking it seriously means asking a different question than “how do we reduce inbound volume?” It means asking whether the bot actually helped, then being willing to change something when the answer comes back no.
Customers were never asking to be served by a person instead of a machine. They were asking to be served.
The company that internalizes that difference has a clear path forward: define resolution as the goal instead of deflection, and hold someone accountable for whether the bot hits it. When the incentive changes, the behavior of the system changes. A bot optimized for resolution looks and functions nothing like one optimized to keep contacts out of the queue.
The 17% cursing at chatbots aren’t an indictment of AI. They’re an indictment of a choice about what AI was for. The 85% who’d use AI that worked are saying the same thing from the other side.
Joel Comm is a columnist at Grit Daily, New York Times bestselling author, internet pioneer, and keynote speaker who has been helping people understand emerging technology since the early days of the web. Best known for making complex topics accessible, Joel speaks and writes about AI, entrepreneurship, digital media, and the future of technology in everyday life. He is the co-host of The Bad Crypto Podcast and host of AI for Everyone, where he explores practical, human-centered uses of artificial intelligence.
Grit Daily News is the premier startup news hub. It is the top news source on Millennial and Gen Z startups — from fashion, tech, influencers, entrepreneurship, and funding.
Get Grit Daily’s latest stories every day in your email.
Thank you for signing up!
© Grit Daily Group. All Rights Reserved.
