Companies face an emerging “AI hangover” after rushing to invest in generative artificial intelligence amid fears of falling behind competitors, according to an analysis of workplace adoption. Global corporate spending on AI is expected to exceed $2.5 trillion in 2026, a 47% increase on the previous year.
The spending surge has been driven partly by businesses giving employees access to tools including Co-Pilot, Gemini and Claude. But after months of attempted roll-outs, many leaders are questioning whether the investment is improving performance.
The analysis identifies three growing problems: stronger-than-expected resistance to AI, limited evidence of a meaningful business impact, and concerns that employees are producing poorer work while feeling more overwhelmed.
Many companies have responded by encouraging staff to use the tools more intensively. But the analysis argues that this risks worsening the problem because businesses have treated AI adoption as a conventional technology project rather than a fundamental change in how people think and work.
Why AI adoption is creating a thinking trap
Generative AI has often been presented as a way to reduce the amount of thinking required for routine tasks. Employees have been encouraged to use it to summarise meetings, write emails, prepare presentations and develop ideas.
It is also being used for marketing plans, product concepts, business problems, weekly planning, difficult customer interactions and workplace disputes. While this can increase the amount of material produced, the analysis warns that higher output does not necessarily mean better work.
Employees may not realise that their critical thinking is being weakened or that the quality of their work is falling. Managers may also become more one-sided in dealing with workplace problems if they rely on AI tools that repeatedly reinforce their existing view.
At the same time, colleagues receiving the increased volume of AI-generated material can become overloaded. Some respond by using AI themselves to process the extra work, while others ignore it or regard it as a stream of average ideas.
The central risk, the analysis argues, is that AI-generated material is treated as finished work. Unless it is used to broaden or deepen human thinking, its output is likely to remain generic and should not be used without scrutiny.
From human in the loop to human in the lead
The proposed alternative is to position generative AI as a tool for improving human judgement rather than replacing it. That means using it to test ideas, identify omissions, offer different perspectives and challenge an initial conclusion.
The analysis refers to this approach as “human in the lead”. Workers remain responsible for the thinking, while AI helps them consider issues more widely, creatively and thoroughly.
It says about 5% of employees with access to generative AI are already using it in this way, often producing better or faster work. These users rarely send out material produced by AI without reviewing it.
Instead, they might ask the tool for several possible responses to an email, write their own version and then seek feedback on how to improve it. The approach is described as a form of “Human-First AI Fluency”, built around metacognition — thinking about how one is thinking.
Under this model, companies should train employees in specific cognitive habits rather than simply measuring whether they are using AI. The aim would be to extend the group of effective users beyond the current 5%.
Employees need greater reassurance
Resistance to AI adoption is not limited to concerns about learning new software. The analysis argues that workplace roll-outs can threaten employees’ sense of status, certainty, autonomy, connection with others and fairness.
Some workers fear that instructions to use AI to “do your work for you” are an indication that their jobs could eventually disappear. Others object to the environmental cost of technology or worry about losing cognitive abilities they value.
Hope about AI among Gen Z has also fallen, according to one study cited in the analysis. While around half of the group uses AI, the proportion feeling hopeful about it dropped to 18% from 27% a year earlier.
Yan Hong Lee, chief human resources officer at DBS bank in Singapore, has banned the word “productivity” when discussing generative AI because of its association with job cuts. Instead, she focuses on what the technology offers each group affected by its introduction.
“My main message to my leadership these days is simple: ‘Can you all please just calm down a little?’” she said.
The analysis argues that businesses should accept a slower and more realistic pace of adoption. Employees were already under pressure before AI was introduced, and forcing new tools on them is unlikely to generate enthusiasm or better results.
Where businesses should set limits
Leaders should be clear not only about where generative AI can be used, but also where it should not be used. A frontline manager, for example, should not rely on AI to give feedback to staff simply because difficult conversations are uncomfortable.
Sales employees should also avoid sending clients emails written by AI without substantial human involvement. Clear guidance linked to individual roles would make it easier for staff to understand what responsible use looks like.
The analysis suggests that businesses may also need to reconsider how work is organised. Routine tasks can fit into a standard eight-hour working day, but deeper thinking may require greater flexibility over where, when and how employees work.
It argues that companies seeking genuine value from AI should focus on improving the quality of work rather than promising that the technology will make work easier. After the initial fear of missing out, the proposed response to the AI hangover is to change the message, reduce the perceived threat and give employees the conditions needed for careful thinking.
