Tech Bosses Reportedly Realise AI Budgets Simply Fund Workers Churning Out Work Slop in a Massive 'Doom Loop'

Tech giants are spending billions on artificial intelligence only to discover much of it is funding low‑value 'work slop' that still needs heavy human checking, with a senior Microsoft executive warning the industry is now trapped in a 'doom loop.'

Some major technology companies are reassessing how they deploy artificial intelligence (AI) as concerns grow over productivity, rising costs and low‑quality workplace content. Microsoft executive vice‑president Ryan Roslansky has described a potential 'doom loop' in which AI‑generated material is summarised by another AI tool without adding meaningful new ideas.

The warning highlights a growing challenge for businesses. Although AI tools can generate documents quickly, their output may still require substantial human review. When employees spend time producing and processing material that adds little value, the promised productivity gains become harder to demonstrate.

The debate is not simply about whether AI can write reports or summarise information. It is about whether those activities improve decisions, reduce workloads or create additional tasks for colleagues who must check the results.

Recent reporting suggests that some businesses are becoming more cautious about their AI strategies. However, tighter controls, additional training and changes to workforce plans should not automatically be interpreted as confirmed cuts to overall AI budgets.

A report published by the Guardian on 13 October 2026 examined the consequences of rapid corporate AI adoption. It described concerns about workplace errors, low‑quality output and the time employees spend correcting problems.

The report also covered changes at Meta and Microsoft, highlighting the difference between introducing AI tools and achieving measurable improvements.

Inside the Office 'Work Slop' Cycle

Roslansky's account offers a practical example of the problem. In a LinkedIn post discussed by the Guardian, he described receiving a document that appeared to have been generated by AI. He then used an AI tool to summarise it, only to find that the process contributed no new thinking.

He wrote: 'AI slop isn't happening just on social networks... it's happening in the day to day of work.'

Roslansky also described the result of processing the document through another AI tool: 'This is a doom loop.'

The example illustrates how quickly automation can become repetitive. One tool produces material, another condenses it, and a human reader must still decide whether it contains useful information. If the final result offers little insight, the apparent efficiency may not translate into meaningful business value.

That does not mean all AI‑generated documents are useless. AI can help workers organise information, draft routine communications and summarise lengthy material. Its value depends on the task, the quality of the output and the level of human oversight.

The distinction matters for Microsoft, which sells AI‑powered products to businesses. Roslansky's comments highlight a challenge facing the wider industry: encouraging adoption is not the same as proving that the technology improves work.

The issue also extends beyond document writing. AI systems can generate plausible‑looking answers that contain errors, while automated agents may introduce additional technical or security problems. Businesses must account for those risks when assessing whether a deployment is genuinely beneficial.

When AI Restructuring Plans Stall

Meta provides a separate example of the difficulties involved in restructuring work around AI. A Reuters investigation published on 26 August 2026 reported that the company abandoned plans for a second wave of workforce restructuring after considering substantial team reductions.

Meta still carried out a 10 per cent workforce reduction in May. Reuters reported that internal information pointed to concerns about AI agents failing to deliver expected productivity gains, alongside technical problems and employee unrest.

The reporting does not establish that Meta cut its AI budget specifically because workers were producing low‑quality documents. Instead, it shows how the practical results of an ambitious AI‑driven restructuring can complicate management plans.

Concerns about the speed of adoption have also emerged among human resources leaders. Johnny C. Taylor Jr., president and CEO of the Society for Human Resource Management, told the Guardian: 'we're driving the truck and building it at the same time, and that's a problem.'

Taylor also said: 'If we could all rewind... we would have introduced [this] in a more thoughtful, slow way.'

His comments point to a management challenge: companies may introduce new technology before establishing suitable training, processes and safeguards. Workers can then be expected to use unfamiliar tools without clear guidance on when they are appropriate.

The Cost Of AI Errors And Oversight Gaps

The financial consequences are also difficult to ignore. EY published an AI Risk and Governance Survey on 15 September 2026, based on responses from 202 senior AI decision‑makers at publicly traded US companies with at least $1 billion (approximately £750 million), in annual revenue.

EY found that 36 per cent of respondents said their organisations had experienced an AI incident or failure with a materially negative impact. Reported consequences included data loss, financial damage, operational disruption and brand damage.

The survey also found that 47 per cent of respondents said their organisations had bypassed AI governance processes for urgent deployments. Among organisations using agentic AI, 26 per cent of respondents said their organisations could not detect unauthorised AI agents operating internally.

These findings indicate weaknesses in oversight, but they do not establish that every company has suffered significant losses. EY also reported a margin of error of plus or minus seven percentage points for the total sample at a 95 per cent confidence level.

For businesses, the challenge is to determine where AI genuinely improves performance and where additional costs, errors or oversight requirements undermine its benefits. That requires measuring outcomes rather than relying on adoption rates or the volume of AI‑generated material.

The evidence does not show that the AI boom is over, nor does it establish that Microsoft or Meta is abandoning the technology. It does, however, highlight why businesses need clearer objectives, appropriate safeguards and credible ways to evaluate results.

The central question is no longer simply how much work AI can produce. It is whether that work helps people make better decisions, serve customers more effectively and achieve results that justify the investment.