CX Without the BS Needs Master
===

Speaker: [00:00:00] April 2026, a man named Praveen Nepali Naga sits down to check the books. He's the CTO, or chief technology officer, at Uber. You probably have heard of them. It's a big company. It's a big job, and it's a big budget, and he's about to discover that the entire AI budget he has set aside for the entire year is gone.

Now we'll need a top up in Q4 gone, like it's actually gone, gone in four months. So what went wrong? What blew up? What failed? Well, nothing. And that's the part that'll mess with your head, 'cause nothing failed. The tool, it worked, and it worked so well that 5,000 of his engineers couldn't put it down. They rolled out an AI coding tool in December, and by February, usage had doubled.

By March, 84% of the engineering organization was hooked. They were shipping more code than ever,

And it was costing them 500 to $2,000 [00:01:00] per engineer per month. See, Uber had built internal leaderboards, and they ranked their engineers by how much AI they had burned. They gamified the spend. The more you used, the higher you scored. So they used, and used, and used

Until the CTO said, "Uh, time to go back to the drawing board." They had to slap a $1,500 a month cap on every employee just to stop the bleeding. So let me ask you something. What kind of money-saving technology torches a full year budget in a single quarter because it's just too good to stop using? And if that's what happens to Uber, a company that runs on software with money falling out of its pockets, - What exactly do you think happens to a 30-person business down the street that just got sold the same dream?



Speaker 9: So you've been sold one story about AI, and it goes something like this: AI is going to replace your [00:02:00] people, and it's gonna save you a fortune. So fire the team, deploy the bot, and then watch your margins fatten. Two big promises right there, replace your people and save money. But here's the truth. AI is costing companies more than humans, not less, more, and it's driving customers straight out the door.

And I'm not anti-AI. Let me be very clear about that up front. AI is real, it's useful, and it's not going anywhere. I use it every single day. But there's a difference between a tool and a miracle, and the entire industry has been selling you a miracle. So today, we followed the money on one side, and then we followed the customer on the other, and we are gonna meet in the middle where the truth actually lies.

Because if you're an advisor who's selling this stuff or if you're a business owner buying it, you're about to make a very expensive decision,

[00:03:00] 

Speaker 10: All right, so back to Uber. Remember that leaderboard? The one that turned spending money into a competition? Well, here's what happened next. Their COO, a guy named Andrew McDonald, went on a podcast and said the quiet part out loud. He admitted they couldn't actually connect all the AI spending to anything customers could really feel.

His words, "That link is not there yet." So they spent the year's entire budget in four months, and the second in command can't draw a straight line from the bill to a single thing that made the product better for their customers. And their CEO, Dara, uh, got asked point blank whether AI was now costing him more than hiring a junior engineer.

And his answer, "I haven't done the math yet, but it's significant." That's the CEO of Uber, and they hadn't done the math but had already started slowing down hiring to pay for their brand new [00:04:00] bills.

And it's not just Uber. Microsoft handed this same kind of AI tool to its own engineers and then turned around and canceled most of the licenses and then shoved everybody onto a cheaper in-house tool because the bill got insane. Walmart capped theirs, Amazon, Cisco, Meta, all tightening up. One company we don't even know the name of reportedly burned half a billion dollars in a single month because nobody even set a limit.

And here's the stat that should scare every CFO in America. The KPMG survey found that only 26% of businesses have a clear view of what their AI actually costs them. 26%. Three out of four companies are flying blind on the bill. They find out what AI costs after that first invoice lands. So that's the dollar side.

Hold [00:05:00] that thought



Speaker 4: Now this is where I'm gonna call some BS. For two solid years, every CEO on earnings calls has said the exact same thing: "AI will make us more efficient. AI will help us cut headcount. AI will help us reduce costs."

And then the market clapped. They announced layoffs. Stock went up. Announce AI, stock goes up. But the company's actually deploying this stuff at scale. They're finding out that the math doesn't math.

The more your people use AI, the bigger the bill gets. And it's not a one-time license, it's a meter. It runs every single time. See, this is the BS at the center of this whole thing. The pitch was replace your expensive humans with cheap software, but the software isn't cheap, and it can't actually replace a human.

So now you've got the cost of AI, and you [00:06:00] still need people. So if AI is this expensive, why does everybody keep insisting it saves money? Where's the proof? Well, let's bring in the receipts



Speaker 6: So we're gonna start with the most damning because it's not coming from a skeptic, it's coming from inside the machine, and that machine is NVIDIA, the very company that's selling the picks and shovels for this entire AI gold rush. Their VP of Applied Deep Learning, Bryan Catanzaro, told Axios that the cost of compute is far beyond the cost of the employees.

So the guy selling the shovels just admitted the shovels cost more than the workers themselves. And then there's the research. A 2024 MIT study looked at what it would actually take for AI to do human jobs,

and they found that AI only made financial sense in 23% of the roles they actually studied.

23%. [00:07:00] In the other 77%, it was still cheaper to just pay a person. "But Brian, prices are dropping." Yep, they are. Gartner says the cost per token could fall more than 90% over the next few years. And here's the trap. Total spending is still going up because the AI agents eat exponentially more tokens than the old tools did.

Cheaper per sip, but everybody's drinking from a fire hose now. Big Tech is on track to spend around $740 billion on AI infrastructure alone this year. That's up nearly 70% from last year. And get this, the Yale Budget Lab looked for hard evidence that AI was actually displacing workers across the economy.

They didn't find it. The great replacement everybody's terrified of, the data doesn't show it's happening. So now connect it back to Uber. Why did their bet blow up? Not because AI is useless. Because the economics [00:08:00] of fire the human, run the bot doesn't actually pencil out yet. Uber didn't get less human.

It got a giant new bill on top of its humans.

So that's the dollar side fully exposed. Now we flip the coin



Speaker 13: See, here's the twist nobody saw coming. The smartest people in the room, they already know. The CEO of Uber, when he was pushed on this, he said something fascinating and revealing. He said if his engineers become more productive with AI, he wants more engineers, not fewer, more. The man writing that giant AI check is telling you the goal was never to replace people.

It was to give people superpowers. The CEO of Duolingo did the same thing. Last year, he was all in on AI replacing tasks, and then he publicly walked it back, and he said he doesn't actually see it replacing what his people do. So [00:09:00] here's the real frontier, the one that doesn't fit inside a conference slide.

It was never AI versus humans. It's humans with AI beating humans without it. That right there, that's the whole game. The companies that win aren't the ones ripping out their teams. They're the ones arming their teams. But here's where things get a little dark. There's a second victim in this story, And it's not the CFO staring at a budget, it's the customer.

Because while the executives were chasing the dream of firing everyone, they pointed the bot at the one place it does the most damage, the front door. Let me tell you about Jake





Speaker 11: November 2022, Jake Moffat just lost his grandma, and he goes to Air Canada's website to book a flight to the funeral. And he does what we all do [00:10:00] now. He asks the chatbot a simple question, "Are there bereavement rates?" And the bot tells him, "Yes, buy the ticket at full price, and you can apply for the discount retroactively up to 90 days later."

So Jake, grieving, trusts it, books the full fare ticket, $1,600,

And then he files for the discount. Denied. Turns out the bot was wrong. Air Canada's actual policy, you can't claim it retroactively.

The chatbot just made it up. It told a grieving man something that wasn't true. And here's where it goes from sad to infuriating. When Jake took them to a tribunal, Air Canada's defense was, and I promise you, I'm not making this up, that the chatbot was a separate legal entity responsible for its own actions.

The airline argued that its own bot on its own [00:11:00] website wasn't its responsibility. I mean, imagine trying that with a human employee. "Your Honor, my cashier acted alone." The tribunal rightfully shut that down cold, and they said the airline is responsible for everything on its website, full stop, and made them pay.

And then the chatbot quietly disappeared from Air Canada's site a couple of months later, like it never even happened. Now, here's why this isn't just one bad bot. This is a unfortunate pattern. A Verizon study of five thousand customers across seven countries found that eighty-eight percent were satisfied dealing with a human, and only sixty percent with AI.

That's a twenty-eight-point gap. Harvard Business Review found that 77% of people find chatbots frustrating, and 87% still prefer a human

Even when the AI performs nearly perfectly. Even when the bot works, they still want a human? Yeah. [00:12:00] That one should stop every business owner cold. A survey by Acquire BPO found that 70% of consumers would switch brands after just one bad AI experience. One.

See, I wrote a whole section in my book, The CX Compass, about exactly this. One in three customers leave after a single bad experience. Now, point a clumsy bot at them, and you're handing that door to seven out of 10, and it's getting worse, not better. One tracking study showed customer frustration with AI agents climbing from 54 to 59% in just six months.

60% of people say they feel forced to use a company's AI when all they want is a real person. Guys, this isn't a tech problem. This is a policy and trust problem, and the culture has noticed. 

Speaker 5: Gen Z and Gen Alpha invented a whole vocabulary of insults for it. Clanker, AI slop, and there are [00:13:00] actual anti-AI rallies now. People are blocking data centers in their own towns. $130 billion in projects stalled or killed in just the first three months of this year. And there's even a US senator with a bill so you don't have to scream "representative" into your phone 10 times to reach a human.

So that's the receipt on the customer side. People don't just dislike bad AI, they're actually organizing against it



Speaker 3: So how do you not become Uber's budget or Air Canada's lawsuit? Now, I sell communications and contact center platforms for a living. So I sit in these rooms, and I hear the AI pitch every single week, and I've got a short list of questions that cut right through it. I call the first part the slow boil because that's how these bills get you.

Not all at once. A little more usage, a little more spend, a little more every month, [00:14:00] till one day, you're Uber, four months in, staring at an empty budget, wondering how the water got so hot. So here's what advisors should be asking and what buyers should be demanding. Five questions. Number one, where's the human escape hatch If a customer can't get to a person in one step, that bot is a liability, not a feature.

Number two, can you connect this spend to an actual outcome? If a vendor can't show you the line from dollars to results the way Uber's own COO couldn't, eh, that's your answer. Number three, what does this cost at scale, not in the demo? The demo, it's always cheap. The meter at ten thousand interactions a month is a different animal.

So ask for the real number. Number four, does this make the customer's life easier or harder? Effort is everything, so every extra hoop you make them jump through is a customer halfway out the door. And number five, [00:15:00] is this attempting to replace your people or arming them? Because one of those builds loyalty, the other one builds lawsuits.

That's the filter. Run every AI pitch through those five before you sign anything



Speaker 12: So here's where I land. AI, it's not a villain, and it's not the savior. It's a power tool. It's incredible in the right hands, and it's dangerous in the wrong place. The future is not AI replacing humans, and the data could not be clearer on that. The future is humans who use AI outworking the humans who don't, on the dollars and the experience.

So if you're an advisor, stop selling, "Replace your team with AI." Instead, start selling, "Arm your team with AI." Help your clients put AI in the back office, the transcribing, the summarizing, the routing, the grunt work, and keep a human at the front where the customer actually is.

And if you're a business owner, do not [00:16:00] point the cheapest bot you can find at your best customer and call that innovation. Run AI behind your people.

Put a human as the face and let the machine handle the busy work so your people can handle the human work. The f- winners in the next five years won't be the companies that fired everyone. They'll be the ones who figured out exactly where the machine helps and exactly where it has no business being.

Back office? Bring on the AI robots. Front door? No. That's a job for a person



Speaker 2: Remember Praveen at Uber staring at a budget that was gone in four months? And then Jake grieving, lied to by an AI bot that the airline swore wasn't even theirs? Well, look at these two stories side by side. One was a bot too expensive to keep running. The other was a bot too cheap to care. And they're the exact same mistake.

Both companies believed a machine could do a human's job. It [00:17:00] can't. At least not yet. Maybe not ever, at least where it counts. So the next time someone sells you the dream, "Fire the people. Deploy the bot. Watch the money roll in," you ask them about Uber's budget, and you ask them about Jake. And if this hits home, do me a favor, subscribe and send it to one advisor or one business owner who's about to make this exact mistake.

You might save them a fortune or a lawsuit



And if you made it this far, do me one favor, subscribe, hit the bell, share this with somebody who needs it. New episodes every week on YouTube. No BS.

No BS. Love that for you. Quick question, Brian

Here we go.

You, you just spent 17 minutes telling everyone AI is overhyped and too expensive

Correct.

So who do you think helped you research the whole thing?

Well, that's different.

It's always different when you use your tokens, isn't [00:18:00] it?

Brian Nichols (New): Subscribe before she lawyers up

As a separate legal entity? I just might