Businesses are reassessing the costs and value of artificial intelligence as executives question whether expensive tools are delivering meaningful productivity gains. The concerns come as LinkedIn chief executive Ryan Roslansky warns that AI-generated documents can add little new thinking to workplace discussions.
The debate reflects a wider corporate challenge: how to control the costs and risks of AI without sacrificing the productivity gains businesses hope to achieve.
Tech Giants Rethink AI Spending
In May, Meta called off a planned second round of layoffs as its AI-led restructuring faced employee opposition and questions over productivity, according to Reuters. The company went ahead with a first round of job cuts, affecting 10 per cent of its workforce, but abandoned plans for further cuts in November.
Reuters reported that internal data suggested AI agents were failing to deliver the expected productivity gains, although it could not establish exactly what prompted the decision.
Meanwhile, LinkedIn chief executive Ryan Roslansky has criticised the spread of low-quality AI-generated material in the workplace.
'AI slop isn't happening just on social networks... it's happening in the day to day of work,' Roslansky wrote on LinkedIn. 'Someone sent me what was clearly an AI generated doc. I read it, realized it was all AI, had my AI summarize it and ended up with nothing that brought new thinking to the idea... This is a doom loop.'
His example highlights a central concern about workplace AI: generating material quickly does not necessarily mean producing useful ideas or improving the quality of work.
Johnny Taylor Jr, president and chief executive of the Society for Human Resource Management, told The Guardian that executives increasingly feel 'we're driving the truck and building it at the same time, and that's a problem'.
'If we could all rewind... we would have introduced [this] in a more thoughtful, slow way,' Taylor added.
AI Workslop Puts Productivity Under Scrutiny
The concern is not limited to unhelpful documents. Businesses are also questioning the cost of running AI tools, particularly as some providers charge according to token usage — the units used to process prompts and generate responses.
A July 2026 survey by Ernst & Young (EY) found that 82 per cent of senior leaders at organisations investing in AI were concerned about token usage and related costs. Among respondents whose organisations both invested in AI and used tools requiring tokens, 98 per cent said these costs had prompted them to reconsider their approach.
However, the findings do not suggest a wholesale retreat from the technology. EY reported that, among this token-using group, 37 per cent were considering expanding their AI roll-outs in response to cost concerns, compared with 15 per cent considering reducing their scope.
Separately, 98 per cent of senior leaders at organisations investing in AI reported a positive return on investment from their initiatives. The findings suggest that, while businesses may see benefits from AI, they are also examining whether those returns justify the expense.
AI Failures Raise Corporate Risks
Beyond the cost of AI tools, companies face potential financial and operational consequences when their systems fail.
A separate EY survey published in September found that 36 per cent of senior AI decision-makers at publicly traded companies with annual revenue of at least $1 billion reported that their organisation had experienced an AI incident or failure with a materially negative impact.
The reported consequences included financial damage, operational disruption, data loss and reputational harm. The finding highlights risks that extend beyond low-quality workplace documents, particularly when AI systems are deployed without sufficient testing and oversight.
AI Audits Expose Gaps in Oversight
The September EY survey also found that 98 per cent of respondents said their organisations conducted formal AI assurance reviews at least annually. These reviews identified recurring problems, including data-quality issues (57 per cent), model drift (48 per cent) and shadow AI (39 per cent) — the use of AI tools without formal organisational approval or oversight.
The findings also showed that organisations were intervening in their AI systems. Among those conducting formal reviews, 64 per cent significantly modified at least a quarter of their systems, 29 per cent paused at least a quarter, and 25 per cent fully stopped at least a quarter.
These figures suggest that businesses are having to scrutinise how their AI systems are developed, deployed and monitored. They do not, however, establish that companies are abandoning AI or that the interventions were all prompted by low-quality output.
The broader challenge is to balance rapid adoption with effective cost controls, reliable systems and demonstrable business value.
