Leaders face fresh realities as AI reshapes work.

Even within the tech companies leading the AI charge, internal tensions are coming to light. On 3 October, a long-tenured member of OpenAI’s safety team, David Robinson, went public with why he decided to resign ([1]). In an essay titled, “I Quit OpenAI Because Its Culture Is Broken,” Robinson argues that OpenAI’s ethos of “iterative deployment” - rapidly releasing AI models and adjusting later - has eroded its focus on safety ([2]). He warned that this move-fast approach "guarantees periodic failures" as AI systems grow more powerful ([3]). Robinson, who spent three-and-a-half years at OpenAI, likened the needed precautions to those in nuclear power or aviation, where exhaustive testing and redundancy come before release ([4]). OpenAI responded that it does pause or hold back models when needed ([5]), but the very public nature of this dispute has sharpened a debate over how to balance innovation speed with safety and ethical guardrails.

Robinson’s departure - first reported by Business Insider and The Guardian - underscores a broader cultural challenge. As AI projects accelerate, some employees fear their companies are prioritising rapid deployment over consideration of long-term risks and workforce impacts. Industry-wide, there is also a "transformation paradox" emerging: companies are enthusiastically adopting AI tools, yet often without redesigning roles and workflows to support them ([6]). Microsoft’s latest Work Trend Index found that only 26% of AI-savvy employees believe their leadership is clearly aligned on AI strategy ([7]). This misalignment can fuel frustration inside organisations, where workers see powerful new tools being introduced without the necessary training, ethical frameworks or process changes.

The lesson for leaders is that rushing ahead with AI for competitive edge carries human risks - from eroding employee trust to real safety oversights. Thoughtful change management and open communication are becoming as critical to AI initiatives as the technology itself. As one former insider put it, if companies developing advanced AI can’t model a strong safety culture, it’s no surprise employees elsewhere are growing wary of how AI might be used in their own workplaces.

New data show that while AI adoption in the workplace is surging, turning that investment into tangible gains is proving harder than expected. In many cases the limiting factor isn’t the technology - it’s the organisation. A much-circulated MIT study found that 95% of enterprise generative AI pilot projects in 2025 produced no measurable return on investment ([1]) ([2]). The problem, researchers concluded, was not that AI tools failed, but that companies failed to integrate them effectively - launching dozens of pilots without rethinking workflows, roles or skill sets. The result is a “pilot purgatory” of experiments that boost individual productivity in demos but stall before delivering company-wide transformation ([3]) ([4]). This sobering reality is forcing many executives to rethink their approach, focusing less on chasing the latest algorithms and more on reorganizing teams and processes to actually capture AI’s value.

The gap between AI’s promise and its payoff often comes down to people and skills. PwC’s newly released Global Workforce Hopes and Fears Survey 2026 highlights a growing divide in the workforce’s AI readiness ([5]). Nearly two-thirds of workers worldwide say they’ve used some form of AI on the job in the past year, yet only about half feel they have adequate training or support to develop AI skills ([6]). In fact, the share of employees who say they can access the learning resources they need dropped from 59% last year to just 51% now ([7]). This lack of upskilling is not only leaving 56% of workers “falling behind” in the AI era ([8]) ([9]), but also creating retention risks: the most AI-proficient talent are the most likely to feel confident and seek new opportunities. In the survey, almost 1 in 3 of the most AI-savvy “front-runner” employees (29%) said they are very likely to change employers within a year ([10]) ([11]).

On the other hand, empowering employees with AI appears to pay cultural dividends. Workers who use AI daily report higher job security and greater trust in their leadership than those who rarely use it ([12]) ([13]). They also express more optimism that automation is expanding their opportunities rather than threatening them. Meanwhile, fears of an immediate AI-driven unemployment crisis have not materialized yet at the macro level. The latest U.S. job market data shows no surge in overall unemployment due to AI. In September, U.S. companies announced 4,000 layoffs attributable to AI - a drop from 7,000 in the same month last year ([14]) ([15]). AI-related cuts still make headlines, but they remain a small fraction of total layoffs. Many firms say they are using AI primarily to augment employee productivity, not to shrink headcount ([16]). But entry-level roles may be quietly shrinking as routine tasks get automated - early career hires in highly AI-exposed jobs are down almost 20% from pre-AI trends in some analyses ([17]). This hints at a future where companies value adaptability and human-AI collaboration skills above replacing workers.

The picture that emerges for leaders is complex but clear in one respect: realising AI’s benefits requires investing as much in people and organisational change as in technology. The companies pulling ahead are treating AI as a catalyst to redesign jobs and upskill their workforce, not just a shiny new tool. Those that don’t risk stranded pilots, disengaged employees, and now even regulatory penalties. The mandate for the C-suite is to lead with a balanced strategy - matching AI ambition with training, ethical guardrails, and new structures that empower employees to amplify their impact alongside intelligent machines.