Have we finally come full circle with artificial intelligence (AI)? Probably not, not yet, at least. But there is somewhat of an awakening, a small market self-correction of sorts that highlights a need for both organisations and nations to rethink how they approach AI.
In the early days of the current iteration of AI, and by that I mean the accelerated evolution — and still rapidly changing — that began when generative AI (GenAI) and large language models (LLMs) surfaced into prominence, tech vendors had assured the world that the technology wouldn’t lead to job cuts.
And when the layoffs did come, organisations would make the effort to explain the cuts weren’t due to AI, but the result of a resizing…a restructuring…a refocusing of the business.
Some time passed and companies didn’t shy away from associating AI with the job cuts. They still took the time, though, to reason why roles were affected by their adoption of AI, saying that some skills could not be realigned…reskilled…reused.
More time passed and businesses got bolder, with cuts hitting tens of thousands of jobs in one swoop.
The layoffs were due to AI, they said, which produced greater operational and cost efficiencies than the human roles could. In fact, the technology meant “lower value human capital” was no longer necessary…no longer essential…no longer cost effective.
However, as the number of retrenched and unemployed grew, and stayed jobless, and as public perceptions of AI turned sour in some quarters, the whiplash hit.
And businesses were hit with the reality of hitting the accelerator on a vehicle that hasn’t been fully tested.
No real AI returns, as costs climb instead
In recent weeks, organisations realise AI isn’t perfect and can lead to costly mistakes, including wiping out an entire production database and deleting an entire codebase.
Forrester in a January report said 55% of employers regret laying off workers for AI, with two in three already rehiring — most of them within six months of cutting the roles.
The cost savings these companies had hoped for also didn’t come through. A third say rehiring cost more than the original job cuts saved them, while 42% broke even.
That means three quarters gained nothing financially and lost institutional knowledge, Forrester noted.
Some also are finding out, to the detriment of their pockets, that their spanking new AI workers cost more, way more, than their human peers.
One CFO in US reportedly chalked up an AI bill of $500 million because it didn’t limit the company’s use of AI tokens, according to Axios.
Some organisations have since tightened their purse strings.
Uber just capped its monthly spending on AI to US$1,500 per employee and per agentic coding tool, with usage tracked via an internal dashboard that can be accessed by every employee. The limit can be exceeded only with permission, according to the ridesharing operator.
The move came after Uber in April revealed it busted its annual AI budget in just four months, after its staff were encouraged to use AI tools as much as possible.
Amazon also pulled down an employee leaderboard that tracked staff’s AI token consumption, first set up apparently to motivate its workers to use AI to carry out tasks and climb up the ranks on the leaderboard. They are now told not to use AI just for the sake of using AI.
Amazon had reportedly cut 30,000 jobs to direct more investments towards its AI buildout.
Unironically, and not unexpectedly, hard cold cash is the kicker that has slapped businesses awake and to finally rethink their AI adoption.
Simply cutting back on human labour in favour of AI isn’t quite the easy money-saving move organisations assume it would be.
But while AI has proven to be more expensive to ‘hire’ than humans, the cost structure may change as the price of compute falls and AI vendors get more ‘innovative’ in how they charge customers for usage.
Jump first, ask questions later
And the race will continue to intensify, as companies fight to outrun the next competitor.
Already, businesses are feeling the heat and are compelled to jump head-in, often without first understanding what exactly it is they’re jumping into.
Some 70% are investing in AI, pushed largely by its potential or fear of falling behind their competition, according to research from a study of 800 multinational organisations in Asia-Pacific, Europe, and the US. Conducted by IDC, the research was commissioned by Expereo.
One in five, or 20%, acknowledge that the fear of being left behind drove them to invest aggressively in the technology with little evaluation.
This figure is the highest in Asia-Pacific, at 37%, compared to 13% in Europe and 10% in the US.
Across the board, 51% plan to prioritise investment in AI or machine learning over the next 12 months, but only 24% say their AI deployments have exceeded expectations.
These numbers are higher in Asia-Pacific, where 61% are prioritising AI investments and 40% indicate such deployments have exceeded expectations.
Amongst respondents in this region, 54% cite higher-than-expected costs or unachieved ROI (returns on investment) as the reason for their underperforming AI initiatives, while 49% point to inadequate or subpar training data. Another 46% say AI simply has not performed as expected.
Globally, 39% are worried about losing track of related costs and ROI once AI Is embedded across their business.
Cognitive impairment cause for concern
And it’s not just organisations that are losing sight of what actually matters amidst the gold rush.
Countries, along with their populations, are rushing to consume all things AI, and this has led to some trepidation about its impact on societies.
Introduced too early to kids, in particular, AI can affect their cognitive development, noted Mohan Kankanhalli, director of NUS AI Institute, who is also deputy executive chairman of AI Singapore.
This is an area the professor believes should be studied more closely, and sooner rather than later.
“We want our future citizens to be fully developed physically, mentally, in every possible way, and having AI access too early may have negative consequences,” said Kankanhalli, whom I spoke with on the sidelines of ATxSummit 2026 in Singapore.
He pointed to calculators, which had triggered calls for its ban when it was invented, over concerns that it could stunt mental growth.
“As a civilisation, we figured out [then] that it could be damaging to our [ability] to develop mathematical maturity and ideas, if you gave calculators too early to schoolkids. But, if you delayed it too long, we would be wasting many years on pen-and-paper calculation that could have been used more productively,” he said.
Consensus then was reached and calculators were permitted in classes when students entered middle school or later, depending on each country’s eventual decision.
“It is not a perfect consensus, but still a reasonable one,” Kankanhalli said. “A similar consensus about using AI models now has to be established.”
“This isn’t happening yet. I don’t think we’ve had serious enough conversations about it,” he said. “You don’t want to give preschoolers access to the AI models and then permanently damage their development.”
He called for more caution here, so such access can be introduced in a safe way and without causing any harm in the process.
“This is one example of where and how we should introduce and diffuse AI in society, in a responsible manner,” Kankanhalli said. “I think somebody needs to pay attention to this.”
And they should do so, before it’s too late, and not just for kids.
An MIT study has found that “excessive reliance” on AI-powered tools may lead to “cognitive atrophy”.
It discovered that participants who used ChatGPT to write short essays exhibited less brain activity, by up to 55%, compared to their peers who didn’t have AI to assist in a similar task.
The “AI group” also were less able to recall what they wrote and quote from their essays after submission.
The MIT study findings are extracted from a small sample base, with just 54 participants, but it does suggest there’s real cause for concern.
“While LLMs offer immediate convenience, our findings highlight potential cognitive costs,” the MIT researchers wrote. “Over four months, LLM users consistently underperformed at neural, linguistic, and behavioural levels. These results raise concerns about the long-term educational implications of LLM reliance and underscore the need for deeper inquiry into AI’s role in learning.”
Prepping next generation of workers
It seems the initial glow of AI may be dampening, with growing backlash especially amongst the youth and those in the arts, who have expressed disdain over suggestions AI could replace human creativity.
Students at the University of Arizona, for one, gave a less-than-enthusiastic reaction to former Google CEO Eric Schmidt’s commencement speech, where he mentioned the rise of AI.
Kankanhalli described the students’ outcry as a symptom of anxiety over the technology’s impact. “It’s real and we cannot just wish it away,” he said.
Just like many countries that realised, over time, the impact of globalisation was not all positive and resulted in pockets of society being left out, they now will need to recognise the same will be true with AI, he noted.
There is increasing anxiety about jobs, and like other universities worldwide, Kankanhalli said NUS also is deeply concerned about the impact AI might have on the next generation of talent.
It is studying how this can be addressed, including how curriculum should be updated and what activities students can be exposed to so they are better prepped when they enter the workforce.
“Am I just going to be replaced by a model? These are valid anxieties and we need to address them,” he said. “And it doesn’t help when industry players hype up every new release of the latest model.”
He noted that AI has huge potential, most of which currently remain untapped, but not everything in the world needs AI.
The first question all of us should ask is whether we need AI to solve a particular problem, he said, as he urged everyone not to get swept up by the hype.
The technology should be leveraged to help wherever it’s able to, but this shouldn’t be done blindly.
“You have to understand your business and see where AI actually brings you benefits,” Kankanhalli said. “Do a careful analysis and then adapt AI where it makes sense, and not just implement a blanket adoption of AI.”
“And the whole of society should benefit from it, not just a few,” he added.
Kids, too, should be encouraged to develop deep skills, including in reading, reflection, reasoning, mathematical skills, and the ability to critically analyse information that is thrown at them.
These skills were important a century ago and still are today, and will remain so another century from now, he said.
People who don’t realise they’re starting to rely on AI all the time, and don’t develop any of their skills, will have no future to speak of, Kankanhalli warned.
“Substituting reading with bite-size summaries of everything is dangerous,” he further cautioned.
Humans still the ones held liable
More importantly, if you don’t have the expertise and simply consume whatever the AI model generates, you’re not going to know when it gives you answers that are wrong.
And ultimately, the human will be the one held accountable.
So, too, will organisations.
