Can We Have Our Humans Back? Why Some Companies Are Rethinking AI.
Industry insiders and CEOs explain the hidden price tag and, in some cases, buyer’s remorse associated with trading real workers for AI agents.

The artificial intelligence revolution may not be eliminating human jobs as quickly as some feared. Rising computing costs, operational headaches, and inconsistent results are prompting some companies to change course and bring workers back.

It’s a hard lesson learned in the throes of the early AI boom, in which bold claims of big savings have enticed many businesses to downsize their staff.
Many industry professionals now say that roles requiring sound judgment, creativity, customer interaction, and quality control need to keep humans in the driver’s seat.

A Careerminds survey of 600 human resources professionals who'd made layoffs in the previous 12 months revealed that nine out of 10 companies would rethink their AI-related terminations.

Three out of four human resources professionals who took the survey confirmed that their organization sacked employees because of technological advancements that replaced roles and responsibilities.

But only 8.4 percent of the survey pool said AI delivered the promised results.

“Over the past 12 months, we have seen a noticeable uptick in companies coming to us after pausing or scaling back AI tool rollouts,” James Calloway, chief operating officer at Stealth Agents, told The Epoch Times.

Calloway’s company provides executive-level virtual assistants, an area where the cost difference between human workers and AI agents is stark.

“One e-commerce client had budgeted for an AI customer service implementation and found the licensing, integration, and ongoing prompt engineering costs were two to three times their original estimate,” he said.

“They hired two of our [human virtual assistants] instead and cut their per-ticket resolution cost by nearly 40 [percent].

“Human employees remain more cost-effective in client-facing communications that require empathy and judgment, tasks that require reading between the lines of what a customer actually needs, work involving proprietary context that cannot safely be fed into third-party AI systems, and any workflow where a mistake has real reputational or legal consequences.”

Big tech companies have also found this to be true. In April, Bryan Catanzaro, vice president of applied deep learning research at Nvidia, told Axios, “For my team, the cost of compute is far beyond the costs of the employees.”

Nickle LaMoreaux, senior vice president and chief human resources officer at IBM, argued that augmenting roles with AI is more essential to corporate growth than replacing human talent entirely, during a Wall Street Journal Leadership Institute summit in March.

LaMoreaux’s comments followed just weeks after IBM announced plans to triple its entry-level hires. When asked why so many companies aren’t taking a similar approach, he said, “It’s because they’re in this productivity mindset versus the growth mindset.”

https://www.theepochtimes.com/...age&ea_med=section-1

Same thing happens with just about any new technology breakthrough. Computers, robots in manufacturing, electric vehicles, etc..

Once the bugs are worked out, there is no going back. There may be ebbs and flows, but AI is here to stay.

I just saw a video on Ford’s JD Power initial quality rating drastically improving and the reason that Ford credits for the improvement is their decreased reliance on AI for quality control and that they rehired experienced engineers back…

Bugs is one thing....hacking is another.

The people making the decisions on who/what to use don't understand the current situation, let alone AI. They make the decision based on what benefits their reputation, which is usually the high-profile cost savings. When it backfires, they can usually hide/spin/bail.
Can't count the number of times I've said 'bet somebody got a bonus for that stupid shit' for something that saved pennies but wasted dollars. The same can be said for being so cheap you cost yourself money.
I recently quit a place that did both and wasn't interested in anything different.

AI is still not ready for prime time. Since overall doesn’t “think” but relies on available information to function, it probably won’t ever truly create independent thought. In actuality, we might be better off with a very efficient assistant rather than a self aware, HOPEFULLY benevolent god.

quote:

Originally posted by smlsig:
...Ford credits for the improvement is their decreased reliance on AI for quality control and that they rehired experienced engineers back…

Ford's AI implementation was called 'Fast then Fix'. Turns out it was fast, but required a lot of fixing.

I wonder what AI thinks about that.

Slop Code.

AI, whatever that means, is in it's infancy right now. Many hiccups are to be expected.
Anyone who thinks the technology won't develop and continue is kidding themselves.
If we're all around here in ten years, let's have another conversation and see if some minds haven't changed.

I made important calls to two medical doctors practices today, and now the AI is answering all calls, and trying to do actual work, which is frustrating, because when if fails to do a simple thing, it tries to talk you into trying something else on the flay, which is pure bullshit because you can't get anything done.

In both cases I told the AI you're a *^&%($^ moron, give me a %&$(^$& human being live agent NOW. And with both AI's that worked. I've done that before with other automated phone agents before AI became a thing, so at least the AIs understand language from the other side of the tracks.

One of the practices has about a dozen docs plus NPs and PAs, and the other probably has over a hundred medicals not including admin and backoffice. In both offices the AIs understood strong language. So if that's the way it's gonna be fore now, at least you're not yelling a flesh and bone agent, if that's any consolation.

However, it's useful for coming up with ideas for starting a coding project or learning about competing architectures when deciding on an approach.

People need to understand there is no ‘I’ in AI. It’s just regurgitating information from a giant database of communications from the web.

It’s not thinking. It’s not able to ask itself a question and come up with an answer. It also doesn’t know right from wrong. It will give you an answer based on info it vacuumed from the net.

If it can barely run a phone tree to route your call consider that a win.

Like online retail and social media, both went through a period of development, stumbles and hiccups. Executives who were a mix of supportive, partial understanding and outright clueless bumbled through about a decade of painful lessons, stubbed-toes and failures before a new generation arrived with a clearer idea of how to implement and utilize these new tools. Today, both segments of the electronic world are indispensable and the public can't imagine a world without them.


Remember MySpace, Friendster, Boston Computer Exchange and Waldenbooks?

I have seen it time and again. Business "leader" comes across a tool. Instead of a rational assessment of capability and deployment based on that assessment, the "leader" now has a hammer and every problem looks like a nail.

The aftermath isn't pretty but always involves massive cleanup and finger pointing.

There is value to AI but the path to automation requires patience, discipline and understanding. All things sadly lacking in many "leaders".

This is a little off the original topic, but perhaps okay to raise here.

quote:

Originally posted by nhtagmember:
It’s just regurgitating information from a giant database of communications from the web.

That has been obvious from many of the responses I’ve gotten from Copilot and Gemini, but I’m curious about an exchange I had a couple of days ago.

I was trying to find the cartoon I posted in the “bayonets” thread showing the cook who discovered that a can opener would fit on the end of a rifle. The answer I got from Copilot referenced the cartoon without being able to link to the actual image, but it stated that the characters were Willie and Joe who were the two who most commonly appeared in the artist’s (Bill Mauldin) WWII cartoons. When I got that response I pushed back and said that, no, they were two anonymous cooks, not Willie and Joe.

The response then was, “Yes, I was mistaken.” That wasn’t unusual as I’ve gotten corrections when I’ve challenged responses about other topics. What struck me, though, was the AI’s statement that it would be easy to assume they were Willie and Joe because those were the characters who appeared most commonly. That was almost an “intelligent,” or reasoning explanation about why the assumption was made. Is that something which would have been simply sucked up from the Internet?

Comments?

Garbage in, Garbage out.... AI is the new buzzword and every company is in on adapting it into their daily operations. Nobody really stops to ask the question "should we" since it's now a race to keep up.

Well sold Tech, Well sold......

quote:

A Careerminds survey of 600 human resources professionals who'd made layoffs in the previous 12 months revealed that nine out of 10 companies would rethink their AI-related terminations.

Human Resource Professional - 3 lies in 3 words

quote:

“One e-commerce client had budgeted for an AI customer service implementation and found the licensing, integration, and ongoing prompt engineering costs were two to three times their original estimate,” he said.

Replacing HR with AI takes much less computing power than replacing customer service with AI, AND the company is likely paying HR more than customer service.

Oil & gas is having success applying AI to task humans are NOT performing such as analyzing big data sets (e.g. decades of production data and geophysical data from a producing asset). Let AI find the info/data to focus the humans on.

The API team that I headed used AI as programmer assistants. I got each team member their choice of Chat GPT, Co-Pilot, or Claude. Most developers chose Chat. None of the three products could "write" software worth a shit. But all three were great assistants. They could take mundane tasks that might take a human hours to accomplish and finish them in seconds. Tasks such as writing data interfaces (classes) or writing basic unit tests to name a couple of many. We improved our output by over 30% with far fewer bugs. AI can't think but it can do what it's told very well and it's improving.