Bev and Xavier
“I need to speak to a human,” I snipped back for the third time to the voice operated call routing program.
“Alright. Please hold while I connect you to a live agent!”
After a brief pause, a cheerful voice joined the line.
“Hi! My name is Bev! This is a live interaction. How can I help you today?”
Immediately, my hackles rose.
“I need to speak to a human.”
“Yes, I understand. I can assure you this is a live call!”
“Well, I’m not so sure you’re a human. You sound like AI.”
“Oh, yes, I understand your concern, ma’am, but this is a live call! We use voice isolation technology that could be, um, altering the sound of my voice as it comes through. How is your day going today?!”
“Fine,” I replied curtly.
“I’m glad to hear that! How can I help you?”
Bev never confirmed or denied her humanity.
She did, however, punctuate her relentless cheer and helpful responses with a plethora of ums, uhs, and let’s see here.
“Um, let’s see here… May I have your updated address?!”
“No. I don’t want to give that to you.”
“Okay! No problem! I totally understand that you don’t feel comfortable sharing that information. Thank you so much for being a loyal customer!”
Despite her surly customer, Bev was as efficient as she was ebullient. In ten minutes, she solved a problem that Xavier, the human I spoke to one week prior, took an hour to address and never resolved. Live agent Bev did more than just answer simple questions and kill me with kindness. She produced results her human predecessor did not.
Over the past few months, I’ve collected a bevy of Bev-like experiences.
Slop and Synthesis
I spent the morning hunched over a laptop, attempting to surgically dissect which parts of a client deliverable were human insights and which were AI-generated. I asked the team directly, receiving a demurred response. I turned to my next task: completing a strategic brief. The brief template was AI-generated, and its structure didn’t quite make sense. Summing up the hours my brain had spent untangling AI slop versus human thought, I banged my proverbial head against the wall. When all the AI-assisted work was done, I fired up Claude to begin my messy, personal Substack authorship process. Claude never writes for me, but supports my synthesis… a thin line of differentiation from what riled my frustration hours earlier.
Prognoses and Palm Springs
During a stroll through the park, my physician friend shared her professional experience with AI. An AI scribe records her patient visits and transcribes notes—a task that used to take her precious time between patients. Soon her employer will deploy AI to offer healthcare providers a series of additional medical considerations or diagnoses to explore beyond their original diagnosis. Have you thought of [insert obscure medical diagnosis here] as a possibility, Dr. So and So? She was open to AI’s support, hoping it would improve patient outcomes and reduce human error. But she also questioned how far we should go. When might AI entirely replace knowledge work like hers? “Does anyone really want AI to tell them they have cancer?” she pondered. As we concluded our walk, we moved on to discussing our upcoming trip to Palm Springs. We opened ChatGPT to review our Airbnb options and verify the best stay based on flight patterns, wind patterns, and walkability.
Claude versus The Couch
A friend, who is a therapist, called me to discuss a big life change. “I’ve been talking to Claude a lot about this,” he began. “Wait,” I interrupted. “Isn’t that exactly what you would tell your patients not to do?” He chuckled. “God, yes. Every day, multiple patients come in and say, ‘Well I’ve talked to Claude…’ If you want to talk to Claude, what do you need me for?!” They say doctors make the worst patients. We spent another thirty minutes talking about the life choices he had been ruminating over with Claude for weeks.
Wrecking Balls and Remodels
I took a colleague out to lunch on her last day at our consulting firm. She shared her upcoming plans: take a 6-month sabbatical; travel solo to Japan; buy and flip houses. Her eyes sparkled. After years spent in technology consulting, she yearned to work with something tangible. She gestured excitedly as she explained the nuances of tiling. She beamed about the joy and pride in crafting something handmade, in obsessing over the details. We briefly mused on the trajectory of AI, and she nodded to her desire to do something entirely different, trading AI transformation roadmaps for hardwood flooring and drywall.
In each of these interactions, we find ourselves relieved, unburdened, annoyed and fed up by AI, all at once. We’re relieved to take the mundane details off our hands; we’re unburdened when we can pour our heart out to an endlessly patient listener; we’re exhausted by wading through the noise; and we’re annoyed when nuance gets lost. I’ll never know with 100% certainty if I was curt to human Bev or AI Bev. But, I know with complete certainty that Bev was a poor judge of vibes. Our frustrations with AI circle around its lack of humanity even as we actively engage it as we would a human.
On the surface, AI does a great job of making us feel seen. Claude has agreed with me, validated me, pandered to me, and complimented me countless times. As we engage with AI, AI responds with big Bev energy. AI doesn’t sigh in annoyance. AI doesn’t roll its eyes. AI doesn’t get distracted. AI doesn’t tell us “Fortheloveofgod please stop talking about this.” Humans do. Trust me, I know.
AI offers each of us immediate, seemingly complete responses to our trivial and not-so-trivial questions and queries. With those responses, an exponential volume of content lands at our feet. As we flood ourselves and others with AI-generated content, we are individually heard and collectively drowning. The human on the other side of AI doesn’t have an endless well of cognitive capacity and patience to draw from. Our only self-defense is to skim, to brush off our lack of attention with “human-in-the-loop” disclaimers, and move on as we attempt to tread water. Our ability to listen degrades. And we return towards the cheerful, reliable listener once again.