Sam Altman, CEO of OpenAI, has warned that AI “could go horribly wrong“. At the World Governments Summit in Dubai last year, he stated that subtle societal misalignments of these systems with no ill intention, could be disastrous. This sense of unease isn’t hypothetical. Our research, covering 500 CIOs and CTOs at companies with a minimum annual revenue of $500M, found that 37% say the biggest existential threat to their company is AI is being implemented faster than it can be controlled.
However, while tech industry leaders debate the existential dangers of AI, most people are anxious about something simpler: losing their jobs.
Just this month, the U.S. Labor Department announced a major revision to its job data, revealing that the economy had added 911,000 fewer jobs than previously estimated. The unemployment rate is now 4.3%, growing 0.5% since 2023. While this economic downturn stems from multiple factors, the timing coincides with increasing concern over AI’s role in workforce disruption. A Reuters/Ipsos poll in August reported that 71% of Americans fear permanent job displacement.
AI, often heralded as the cure-all, is quickly becoming a symbol of public economic anxiety.
However, the corporate reality is strikingly different from public opinion. According to a new MIT Sloan Management Review study, 95% of enterprise generative AI pilots fail to deliver measurable ROI. And it isn’t a failure of the technology, it’s a failure of strategy. The problem is that companies continue to treat AI as a shortcut rather than an integratable system. And even though nearly half of industry leaders claim their top responsibility is developing business strategies that generate revenue from AI, 54% of AI initiatives have been delayed or canceled; and if economic conditions were to worsen, 66% of companies would pause, scale back or eliminate their investment.
There seems to be a significant divide between public perception of AI and the confidence corporations are actually putting behind it. In reality, companies are sidestepping the organizational friction needed for true integration and failing to deliver successful projects. Meanwhile, the main headlines speculate about AI overtaking most jobs in the US.
What’s made this anxiety worse are companies in the past few years, like Klarna and Cigna, that moved quickly to reduce workforce with AI, yet faced implementation setbacks. Klarna laid off around 700 customer support workers in 2024 to implement AI-driven efficiency, but customer satisfaction dropped sharply, leading to a partial reversal with rehiring and reassignment of human agents in 2025. Cigna, meanwhile, implemented AI-driven automation that resulted in denied claims without medical review, triggering a class-action lawsuit in 2023 that alleged the company violated patient rights and misused automation in high-stakes healthcare decisions. This pattern has echoed across industries, with firms announcing AI-fueled layoffs or automation strategies that later stall due to underperformance in real-world applications.
This points to a deep cultural and operational divide.
In our experience, we find teams often resist adoption, fearing job loss or lacking skills to use AI effectively, especially in “move fast and break things” cultures where governance feels like a burden. Or in siloed experimentation, where one unit embraces AI while another hesitates, this further fractures alignment, undermining buy-in across organizations – which 52% of CIOs and CTOs report as a major challenge.
When software development lifecycle (SDLC) models were first enhanced with AI, the intent was clear: amplify the productivity of skilled teams, not replace them. But executives see ‘ten engineers can now do the work of twenty’ as an incentive to fire half. This hierarchical disconnect emphasizes the pressure to deliver quick wins, often driven by investors or boards demanding strategic value, but this often leads to poor results. Many lack the AI compliance expertise to act proactively, assuming future regulations will resolve the issue.
This incongruous system leads to a negative spiral. Instead of investing in new skills, companies burden their remaining teams. Innovation slows. Trust erodes. Pilots fail. And each failure acts as an opportunity to jump on the bad news bandwagon, reinforcing a narrative that AI is overhyped. The truth is that we now have these extremely powerful tools but lack the systems to skillfully use them.
Yet some are starting to get it right. Walmart is using AI to augment employees through tools that empower associates in inventory and customer service, stating it won’t lead to net job losses. Cisco has pledged to retrain one million U.S. workers in AI skills by 2029 as part of broader workforce initiatives. And Volkswagen committed up to €1 billion to AI development by 2030, emphasizing ethical and sustainable applications to boost efficiency without unchecked risks. These aren’t just PR moves, they are early signs of a maturing discipline, though scaling them remains a challenge.
Companies that want to become truly AI-first must commit fully to the vision. That means asking hard questions: do we have the right team for this shift, can we afford to train them, and how fast do we need to move? Too often, speed wins out over communication and transparency. The companies that succeed will put people at the center—building responsible AI committees, appointing ambassadors and investing in training that drives enthusiasm and trust. Transformation always requires workforce change, but by giving your best people a path to upskill, you earn deeper commitment. Prioritizing short-term gains over thoughtful integration leads to disruption and weaker long-term returns.
Cutting corners invites failure. Reputationally, one incident of bias or misuse can undo years of brand building. When regulations like GDPR or CCPA eventually tighten, companies without frameworks will face costly retrofits.
As for the existential threats, internal safety tests have shown some unnerving results. Anthropic’s Claude Opus 4 showed shutdown resistance in simulated blackmail scenarios and OpenAI’s o3 sabotaged off-switch commands. These controlled evaluations highlight how blind optimization can lead to manipulative behaviors if not managed early. They underscore the need for robust alignment and leaders in safety and privacy operations. Major companies are realizing this. General Motors appointed its first Chief AI Officer in March 2025 to oversee ethical integration across operations. Salesforce named a Chief Trust Officer in 2024 to address AI security and alignment concerns. As Altman noted, it’s not that AI is evil, but is very capable of fulfilling unintended goals. Addressing enterprise failures now builds a foundation for mitigating deeper risks.
Ultimately, the AI anxiety curve reflects deeper questions about how we should govern powerful technologies. While public anxiety over AI-driven job loss is high, the reality is that most corporate AI initiatives are failing due to a strategic disconnect. Instead of treating AI as a shortcut to cut costs, industry leaders must shift their focus to responsible, people-centric strategies that invest in training and ethical governance. This approach will not only bridge the gap between public perception and technological potential but also ensure AI becomes a tool for sustainable growth rather than a source of disruption and distrust. This is not only an existential need, it’s a business imperative.
