Six in 10 employees worry AI is weakening their skills, while 43% say they get the blame when AI goes wrong, new global research reveals.
A new global study from the IBM Institute for Business Value reveals 60% of employees worry AI is weakening their skills. Artificial intelligence is making critical thinking and human judgement increasingly valuable at work, yet employees are worried greater reliance on the technology could be weakening those very skills, according to the report.
Among those concerned about what IBM describes as “skills erosion”, three in four say AI has already begun to weaken at least some of their skills, with critical thinking cited most frequently as declining. Yet critical thinking is simultaneously becoming one of the capabilities organisations say they need most. Some 49% of employees and 57% of Chief Human Resources Officers (CHROs) identify critical thinking and problem framing as among the most important skills in the AI era.
Meanwhile, 71% of CHROs identify the ability to supervise, validate and override AI outputs as the workforce’s most essential skill, while just 29% of employees rank judgement as important. The findings expose a growing challenge for employers: as AI takes on more of the thinking at work, organisations may need to become much more deliberate about protecting employees’ ability to think for themselves.
CONCERNS AROUND AI WEAKENING SKILLS
The IBM study, based on separate surveys of 1,500 CHROs and equivalent senior workforce executives and 8,800 full-time employees globally, suggests concerns about AI-driven deskilling are no longer confined to employees. Skills erosion also ranks among CHROs’ leading concerns, cited by 46%.
The distinction between skills gaps and skills erosion is particularly important. IBM found 80% of organisations already have a reskilling roadmap to help their workforce collaborate with technology. But learning how to use new AI tools addresses only one side of the problem.
Reskilling can provide employees with capabilities they do not currently possess. Skills erosion, as IBM describes it, occurs when people stop exercising capabilities they already have because AI increasingly performs those tasks for them. That creates a very different workforce challenge.
An organisation could successfully train thousands of employees to use AI while simultaneously weakening their ability to research independently, interrogate information, frame problems, exercise judgement or challenge questionable conclusions. Previous research has already highlighted similar concerns. Nearly half of employees fear AI is making them less intelligent, while 39% said overdependence on the technology was weakening their skills.
That research also found 30% feel unable to function without AI, 28% trust AI more than their own judgement and 41% believe overreliance on AI could damage their long-term career prospects. The latest IBM findings suggest protecting existing human capability could therefore become just as important as teaching employees new AI skills.
CRITICAL THINKING BECOMES MORE VALUABLE AS AI DOES MORE THINKING
IBM’s findings expose an apparent paradox. As AI becomes capable of performing more cognitive work, the human ability to interrogate that work becomes more important, not less. Some 57% of CHROs identify critical thinking and problem framing as important capabilities for the AI era, while 48% identify human judgement.
IBM’s full study also reveals a substantial gap around employees’ ability to evaluate AI output: 71% of executives prioritise the ability to supervise, validate or override AI outputs, compared with 38% of employees. That matters because AI can produce answers that appear authoritative even when they are inaccurate.
As Fair Play Talks recently reported, one in four executives say AI errors have already reached boards or external audiences. The Workiva research revealed 84% of executives were at least somewhat confident in AI output without human review, despite just 11% believing their organisation’s data quality was sufficient for AI use.
Some 26% said internal audits had detected AI errors that had already reached external audiences or board members. If organisations increasingly depend on humans to identify when AI is wrong, they cannot afford to allow the skills required to make that judgement to weaken.
HUMAN OVERSIGHT NEEDED
The IBM findings also challenge the assumption that simply retaining human involvement provides sufficient protection against AI mistakes. The quality of that oversight depends on whether employees have the knowledge, confidence, authority and critical-thinking skills to challenge what the technology produces.
Where judgement is deliberately built into work, 62% of CHROs report growing employee confidence in AI-enabled decisions. Where it is not, 57% report confidence declining. That reinforces another recent study. Four in 10 US workers using AI recently admitted accepting answers they suspected were wrong, while 30% said inaccurate AI output had already caused problems at work and just 22% worked under a written employer policy requiring verification.
The two studies use different populations and methodologies, so the figures should not be directly compared. But together they highlight an important weakness in one of the most frequently cited safeguards for responsible AI. Putting a human in the loop does not automatically create meaningful human oversight. The person needs the skills, time, authority and confidence to challenge the machine.
ACCOUNTABILITY MATTERS
The issue becomes even more complicated when accountability enters the picture. Some 43% of employees surveyed by IBM say that when something goes wrong with AI, the blame falls on them. At the same time, 41% of CHROs believe employees may not feel safe challenging or overriding AI outputs. Another 36% of CHROs say unclear accountability complicates AI deployment. That creates a potentially damaging contradiction.
Employees may be expected to exercise judgement and challenge AI recommendations while simultaneously feeling unsafe doing so – and believing they will be blamed if something goes wrong. As Fair Play Talks has previously explored, AI transparency increasingly depends on psychological safety in the workplace. Employees need to feel able to say “I think the AI is wrong”, disclose when they relied too heavily on a tool or flag an error before it reaches a customer, regulator or board.
IBM’s findings provide further evidence that the way organisations structure responsibility matters. Where the CHRO at least shares responsibility for determining which decisions remain human-led, 76% of employees feel safe questioning or overriding AI recommendations, compared with just 43% where HR is merely advisory. Psychological safety and accountability should therefore reinforce one another rather than being treated as competing priorities.
AI IS CREATING ‘INVISIBLE’ WORK
The research also challenges simplistic assumptions about AI productivity. Some 80% of CHROs believe AI adoption creates “invisible” work for employees, including validating recommendations, correcting mistakes, supplying context and dealing with exceptions.
Meanwhile, 42% of employees say AI increases their work or creates work that goes unrecognised. That is significant because much of the business case for workplace AI rests on its ability to save employees time. But AI can simultaneously eliminate one task while creating others: checking the output, correcting mistakes, adding missing context, handling exceptions, explaining the result and deciding whether the AI should be trusted in the first place.
Those activities are still work. Previous research found two-thirds of employees spend up to six hours each week correcting poor-quality AI-generated work or “workslop”. Some 49% said they fix poor-quality AI work themselves rather than flagging or rejecting it, creating another form of invisible labour.
Taken together, the findings suggest organisations need to be careful about measuring AI productivity solely through the time apparently saved on an individual task. If that saving simply transfers checking, correction or additional work elsewhere, the organisation’s actual productivity gain could be considerably smaller.
WHERE IS THE AI DIVIDEND GOING?
The IBM research raises another important question: what happens to the capacity AI creates? Only 42% of CHROs say AI productivity gains are primarily reinvested in innovation or reskilling. IBM’s wider analysis found other organisations either divide the gains between savings and reinvestment, take them as cost reduction or profit, or lack a clear route for converting productivity into new opportunities. That matters because AI-created capacity can be used in very different ways.
Businesses can convert it into greater output or lower costs. But they can also reinvest it into learning, coaching, career development, innovation, internal mobility, creative work or healthier workloads. If every hour saved by AI simply becomes another hour available for additional tasks, employees may see relatively little of the productivity dividend themselves. It also creates the risk that organisations increase output while reducing the time employees have to exercise and develop the human skills they increasingly need.
The findings add another dimension to the emerging business case for responsible AI. Recent research found organisations with strong trustworthy AI practices are 15 times more likely to report strong or high returns from their AI investments. That SAS research found stronger governance, explainability, data quality and accountability were associated with stronger reported AI returns, suggesting human oversight should not simply be viewed as a brake on AI productivity. Done well, it may be one of the conditions required to achieve it.
CLEARLY DEFINING WHAT HUMANS AND AI SHOULD DO
One of the most important findings in IBM’s full study concerns the design of work itself. Only 26% of organisations clearly define activities as human-led, AI-assisted or AI-executed. Yet organisations that clearly make those distinctions report achieving an 18% reduction in risk and a 20% improvement in quality. The findings suggest businesses could gain more from AI by asking a deceptively simple question: What should the human do – and why?
Not every task should automatically be handed to AI simply because the technology can perform some part of it. Organisations need to consider where human judgement, relationships, accountability, creativity and expertise provide value – and then deliberately protect those capabilities within the design of work.
IBM also found 52% of employees say the tasks they are assigned have changed during the past year as AI reshapes day-to-day responsibilities. But only 25% of organisations dynamically and continuously manage workforce recomposition as work changes. In many businesses, therefore, jobs may already be changing faster than organisations are deliberately redesigning them.
AI STRATEGY STILL DESIGNED WITHOUT HR IN THE ROOM
The research also exposes a leadership gap. Nearly half – 46% – of organisations do not involve the CHRO when AI strategy is being defined. That is striking given that AI increasingly affects recruitment, skills, jobs, career pathways, performance expectations, workload, accountability and employee experience. Coordination across leadership teams also appears weak.
Only 28% of CHROs report having a joint roadmap with IT supported by a shared operating cadence, while 73% say they struggle to coordinate consistently across the C-suite. This reinforces earlier research suggesting businesses are still underestimating the people challenges of AI.
The challenge for leadership is therefore not simply deploying the technology. It is redesigning the organisation around the changing relationship between human and technological capability.
WHY HR IS EXPECTED TO LEAD AI CHANGE
There is another contradiction. HR is increasingly expected to help organisations navigate the workforce consequences of AI, yet 72% of organisations make limited or no use of AI within the HR function itself. CHROs also rate HR’s capabilities particularly poorly in several areas:
- AI literacy across the HR team – 13%
- AI performance measurement – 16%
- Change management for AI adoption – 20%
Yet organisations with mature HR AI capability – including governance, architecture and measurement – are nearly twice as likely to report a higher number of positive business KPIs. For HR leaders, this creates a dual challenge. They need a meaningful role in determining how AI reshapes the wider workforce while simultaneously developing their own function’s ability to use, govern and evaluate the technology.
“AI is changing not only how work gets done, but where people can contribute the greatest value,” said Nickle LaMoreaux, Senior Vice President and Chief Human Resources Officer at IBM. As AI takes on more routine and process-driven tasks, uniquely human capabilities become even more important. CHROs have a critical role to play in redesigning the workplace of the future so people can focus on the areas where they can have the greatest impact.”
AUTOMATING DEVELOPMENT OF FUTURE LEADERS
The question of weakening skills has implications beyond employees’ immediate performance. Junior and mid-level work has traditionally provided opportunities for people to practise judgement, solve problems, make mistakes, learn from experienced colleagues and gradually take responsibility for more difficult decisions.
If AI increasingly performs some of those tasks, organisations need to consider where future employees will acquire the experience required to perform higher-level work. Dr Amit Das, Director and CHRO at Bennett Coleman & Co. Ltd. (The Times of India), warns in IBM’s study: “The organisations obsessing over AI productivity gains are, in many cases, quietly hollowing out the developmental infrastructure that produced their own current leaders,”
That concern extends beyond existing employees. As another study recently revealed, one in five companies surveyed have already stopped hiring entry-level workers because of AI. Entry-level jobs have traditionally acted as training grounds for future managers and specialists.
If organisations automate the tasks through which junior employees learn without creating alternative development pathways, they could inadvertently weaken their future talent pipelines. The challenge is therefore not simply ensuring today’s employees retain critical-thinking skills. Businesses also need to ensure tomorrow’s employees still have opportunities to develop them.
WORKERS ARE ALREADY MOVING FASTER THAN GOVERNANCE
The challenge becomes more complicated as employees adopt AI independently. Previous research found workers are adopting AI faster than many employers can govern it. Across 72 organisations analysed in that research, employees actively used almost 50 different AI applications, highlighting the growth of so-called shadow AI and potential risks around data security, confidential information, inconsistent outputs and oversight.
The IBM findings suggest employers therefore need to address two challenges simultaneously. They must build employees’ ability to use AI. But they must also protect their ability to question it.
WHAT EMPLOYERS & BUSINESS LEADERS SHOULD DO NEXT
Protect critical thinking – don’t just teach AI skills
AI literacy is important, but employers should also identify which existing human capabilities could weaken through lack of use. Training should therefore address skill preservation as well as skill acquisition. Employees need opportunities to research, reason, frame problems, make decisions and exercise judgement independently rather than automatically outsourcing those capabilities to AI.
Define what should remain human
Organisations should map important workflows and clearly define which activities should be human-led, AI-assisted or AI-executed. They should prioritise human oversight for decisions involving significant consequences, ambiguity, ethical judgement, people or accountability.
Recognise AI oversight as work
Checking recommendations, correcting mistakes, providing context and dealing with exceptions are not invisible extras. They consume employee time, so organisations should account for them when measuring AI’s productivity impact.
Create safe routes for challenging AI
Employees should know they have both the authority and responsibility to question AI recommendations. Organisations should create a culture where employees feel confident challenging AI without fearing repercussions or blame when it gets things wrong.
Define accountability before something goes wrong
Organisations should define responsibility for AI-assisted work before problems arise. They should make clear who validates AI outputs, who handles exceptions, when employees should escalate decisions and who ultimately owns the outcome.
Reinvest some of the AI dividend in people
AI-created capacity does not automatically have to become more work. Organisations should consider deliberately reinvesting some of the time saved into learning, coaching, innovation, internal mobility, collaboration and career development.
Measure weaking skills alongside skills gaps
Businesses routinely identify new capabilities employees need. They should also monitor whether existing capabilities – particularly critical thinking, judgement and problem-solving – are weakening as AI assumes more tasks.
Protect the development pipeline
Employers should consider how people will acquire expertise if AI increasingly performs junior-level cognitive work. Work experience, mentoring, apprenticeships, stretch assignments and deliberately protected opportunities to make decisions can help employees continue developing judgement through practice.
Bring HR into AI strategy early
CHROs and people teams should therefore be involved when organisations determine which work will be automated, augmented or retained as human-led – not brought in afterwards to manage the consequences.
Build AI capability within HR
HR cannot credibly lead AI-enabled workforce transformation while remaining inexperienced with the technology itself. People teams need practical AI literacy alongside expertise in governance, performance measurement and change management.
WHY THIS MATTERS FOR RESPONSIBLE BUSINESS
For years, much of the workplace AI debate has focused on whether employees have the new skills required to use the technology. IBM’s findings raise a different question: What happens to the skills people already have when AI starts exercising them instead?
That distinction matters. An organisation can successfully train thousands of employees to use AI while simultaneously weakening their ability to think critically, interrogate information, exercise judgement or make decisions independently. And those are precisely the capabilities businesses may need most when AI gets something wrong.
Recent Fair Play Talks reports illustrate why. Employees have admitted accepting AI answers they suspected were wrong. AI errors have reached boards and external audiences. Workers are spending hours correcting poor-quality AI output. And employees are adopting AI tools faster than many employers can govern them.
The IBM findings also challenge a simplistic interpretation of human oversight. Keeping a person “in the loop” offers limited protection if that person has lost confidence in their own judgement, lacks the skills to identify a questionable output, does not feel safe challenging the technology or is under such pressure that verification becomes another invisible task squeezed into the working day.
Responsible AI therefore cannot simply be about teaching employees how to use increasingly capable technology. It must also be about designing work so people continue to exercise the human capabilities that technology cannot safely replace. That means preserving opportunities to think, question, learn, decide and occasionally disagree with the machine. Because the challenge for employers isn’t simply preparing people to work with increasingly capable AI. It’s making sure that, as AI becomes more capable, people don’t become less so.
Download IBM’s full study here.
RELATED ARTICLES
AI may be transforming workplace productivity, but a growing number of employees fear the technology is making them less intelligent in the process, according to a new survey.
Despite 84% of executives expressing confidence in AI-generated output without human review, just 11% believe their organisation’s data quality is sufficient for AI use, exposing a potentially dangerous gap between confidence, governance and reality.
Some 42% of US workers surveyed who use AI on the job admit they have accepted an AI answer they suspected was wrong, while 30% say inaccurate AI output has already caused problems at work, reveals new research.
Why responsible AI governance must protect workers from blame while helping organisations build trust, accountability and responsible AI governance.
Two-thirds of employees say they spend up to six hours a week fixing low-quality, AI-generated work, according to a new study.
Organisations with strong AI governance, data quality, explainability and accountability are 15 times more likely to report strong returns from artificial intelligence investments, according to new global research.
A growing number of companies are freezing entry-level hiring as artificial intelligence (AI) reshapes the workforce, according to new research.
