Recent studies suggest artificial intelligence is far from replacing humans. Although widely adopted, AI reportedly fails in 96% of professional tasks and often produces corrupted, incomplete, or inaccurate results requiring human correction.
Testing across design, programming, and writing showed only a small portion of outputs usable without modification, while humans met standards far more consistently. Companies investing heavily in AI and reducing staff have sometimes experienced declining quality and increased errors.
AI also lacks real understanding, relying on pattern imitation rather than logic, which increases risks in critical sectors like medicine and engineering. Overall, AI works best as productivity tool.
CHAPTERS: 0:07 AI Job Market 0:32 Tech Jobs 1:05 Automation Employment 1:44 Workforce Reduction 2:18 Human Supervision 3:02 Professional Tasks 4:11 AI Productivity 5:09 Job Displacement 6:21 Workplace Automation 7:28 Future Of Work
Produced by: Samantha Harvey Edited by: Jacob Smith Animation: Charlotte Brown Additional Footage: Getty Images
Despite massive investments and hype, AI currently fails in 96% of professional tasks. While it can improve efficiency in simple drafting and summarization, it lacks real logic and precision, often producing corrupted files or critical errors. The consensus is that AI should serve as a supportive co-worker rather than a human replacement to avoid operational and safety risks.
In recent years, artificial intelligence has been marketed as the technology that will replace humans.
But today, this is still far from happening.
Look, I think what it's going to do is completely transform the industry.
It's going to change the way content is created, produced, and distributed.
But that doesn't mean there isn't going to be a role for the great artists who are in roles today.
Well, a recent study reveals that AI fails in 96% of almost all real professional tasks it performs.
But this was just a reaction to the headline.
Because as you dig deeper into what was actually said in the report,
the picture isn't as clear-cut as AI simply failing everywhere.
In addition, it makes many constant errors,
such as producing corrupt files or incomplete work that forces people to redo them from scratch.
But the real problem is, you don't know what it's made up and you don't know what's accurate.
So you or your staff manually have to go back and check everything.
For this reason, AI cannot work on its own,
since it does not understand rules and does not know what it is doing.
It only imitates language without having real logic.
But sometimes that answer is unwieldy or not what you wanted.
Despite this, many companies are laying off people,
while one in five AI systems makes critical errors in sectors such as medicine or engineering.
Even in the medical field, Reuters just reported that the FDA has received 100 reports of AI malfunctions,
botched surgeries, and misidentified body parts.
Thus, what seemed to be the future of work could become a problem
with much more serious consequences than expected.
So why is AI failing so much?
In recent years, artificial intelligence has gone from being a technological promise
to a tool present in daily life.
More than 300 million people use AI in their work,
which has accelerated the debate about its impact on employment.
However, a recent study shows that AI fails in 96% of professional tasks,
which raises doubts about its ability to replace humans anytime soon.
As companies invest large amounts of money in this technology, expectations also grow.
The six U.S. firms pouring the most cash into AI are projected to spend over $750 billion on it this year alone,
more than the entire GDP of Ireland.
It is expected to take on complex functions autonomously.
More than $400 million are allocated each year to the development of new models,
reflecting market confidence in its potential.
It is estimated that AI investment up to 2025 shows steady growth, reaching $126 trillion.
However, there is a clear gap between what is expected from these systems
and what they can actually deliver with professional quality.
This gap raises a clear issue regarding the true scope of artificial intelligence in today's labor market.
It's only honest for people to speak frankly about the fact this will have a big effect on the job market.
One of the main problems is the low effectiveness of AI in real tasks.
In tests carried out in areas such as design, programming, and writing,
more than 500 different professional tasks were evaluated.
Most systems failed to complete them correctly.
This shows that although AI can generate content,
it still does not reach the level of precision required for human work.
In addition, only 12% of the results were considered acceptable without modifications.
This limitation becomes even more evident when results are compared with human performance under similar conditions.
In work carried out by professionals, 8 out of 10 tasks meet standards from the start.
This shows a clear difference in consistency and reliability.
AI can produce fast responses, but the final quality still depends on human intervention.
Furthermore, more than 200 million work tasks have been analyzed across different sectors to measure AI performance.
Despite this volume of testing, results continue to show that technology does not fully adapt to the demands of professional work.
In one study, Gen AI failed in 95% of cases where companies tried implementing it.
In another study of 25,000 Danish workers, introducing AI meant more work for about 8% of people.
Added to this limitation are the frequent technical errors these systems present.
In many cases, AI generates corrupted or incomplete files that cannot be used directly in a professional environment.
This not only affects work quality, but also causes delays in processes.
For example, in programming tests, around 35% of solutions had failures that prevented execution.
This type of error forces a human to review and correct the work from scratch, reducing the expected benefit of automation.
Additionally, it is estimated that more than 230 million AI-generated files require some type of correction before they can be used.
On top of that, around 35% of results contain formatting errors that make immediate use difficult.
This shows that the technology is still not fully reliable in producing technical content.
And like we told you at the beginning of this, we did use AI to make part of this segment.
And when we got back from the prompt that we put in, there were actually a lot of mistakes.
On the other hand, it is important to understand that AI works better as a tool rather than a complete replacement for human work.
Its real value lies in its ability to support processes and improve efficiency in specific tasks.
In simple activities such as summarizing information or generating drafts, it can reduce work time by approximately 40%.
This allows professionals to focus on more complex tasks.
However, these benefits largely depend on constant human supervision.
When AI is used without control, results tend to lose quality.
In many cases, errors go unnoticed without proper review.
In fact, teams that combine AI with supervision achieve better results, where 7 out of 10 tasks reach an acceptable level.
Additionally, more than 150 million workers use AI as support in their daily activities.
This shows that its current role is more related to assistance than full replacement.
Sure, we're going to have some jobs that change, but really it's going to be about working hand-in-hand with AI,
using AI as our co-worker.
Despite these limitations, many companies have begun making rushed decisions regarding the use of AI.
In some sectors, human teams have been reduced by 15% with the expectation that technology can cover those functions without issues.
However, results have not been as expected.
In several cases, staff reduction has led to a decrease in work quality and an increase in operational errors.
This situation has forced many organizations to rethink their strategies.
It is estimated that more than 110 million workers could be affected by similar decisions in the coming years.
However, it is also expected that 3 out of 10 companies will need to rehire staff to recover lost quality levels.
Additionally, around 45% of companies that implemented AI rapidly have reported problems in their internal processes.
This highlights poorly planned adoption.
The CEO recently walked that policy back and started hiring people again.
Not out of the goodness of his heart, but because AI just couldn't cut it.
And he's not alone.
Another key aspect is the lack of real logic in these systems.
Although they can generate coherent text, they do not truly understand the content they produce.
This limits their ability to make correct decisions in complex situations.
In specific tests, more than 60% of decisions made by these systems were incorrect when practical rule application was required.
This shows that AI does not possess a deep understanding of the world.
This limitation is also reflected in the fact that 5 out of 10 responses contain reasoning errors in complex contexts.
Although the language may appear correct, the content can have significant flaws.
Additionally, more than 170 million analyzed interactions show that AI tends to repeat patterns without evaluating whether they are appropriate for each situation.
All of these disasters make sense.
LLMs predict the next word statistically, but they'll never tell you when it doesn't know the answer.
Finally, risks increase when AI is used in critical areas where errors can have serious consequences.
In sectors such as medicine or engineering, precision is essential.
A 25% increase has been observed in incidents related to the improper use of automated systems in technical environments.
This data reflects the importance of maintaining constant human oversight.
Additionally, more than 60 million automated decisions per year require human review to avoid significant errors.
This shows that technology still cannot assume full responsibility in these fields.
Compared to processes controlled by professionals, 2 out of 5 automated results present risks that cannot be ignored.
And in another study focusing on programmers specifically, AI-made coding take 19% longer on average.
Despite the advancement of artificial intelligence, its inability to perform complex tasks with precision shows that it is still not ready to replace humans.
The gap between expectations and reality remains wide, confirming that its role should remain as support rather than a substitute.