AI doesn’t lay people off. Corporate managers do.
The arrow of accountability points to the corporate decision makers, not the AI they blame
When headlines shout that AI threatens jobs, it obfuscates that job losses arising from the uses and abuses of AI are business decisions. Blame the algorithm, the machine, or the inevitable march of technology rather than deliberate attacks on corporate workforces. AI doesn't make decisions. People do. Specifically, corporate managers make decisions about how to deploy AI, and often, that decision is about cutting costs, maximising profits, and consolidating power, not about genuine innovation or societal benefit.
Accountable AI means holding the decision-making individuals to account for the impact of the AI they choose to deploy.
Corporations are pouring unprecedented amounts of capital into AI infrastructure and development. This isn't small-scale innovation; building hyper-scale data centres, acquiring immense computing power, and snarfling vast datasets. These enormous investments serve a dual purpose: to gain a competitive edge and, crucially, to create insurmountable barriers to entry for smaller players. Only the tech giants can truly play this game, further consolidating their market power.
However, these approaches require immense resources with significant ecological and social consequences.1 The energy consumption and water usage for cooling data centres rob local communities of potable water and accelerate the climate crisis. The human toll of relying on exploited ‘ghost workers’ in the Global South for low-paid data annotation and content moderation – the invisible labour2 that trains our "intelligent" machines. This isn't just about efficiency; it's about power, control, and a relentless pursuit of profit, regardless of the actual cost.
The Pressure Cooker of ROI – Human Costs as "Savings"
Investors demand returns on their colossal AI investments. How do corporate managers prove AI's worth? For many enterprise customers, the most straightforward and immediate way to demonstrate Return on Investment (ROI) from AI is by drastically reducing labour costs. The history of automation reveals a complex pattern where technological advancement has often been deployed strategically to reduce labour costs and limit worker power. Since the 1970s, the share of economic growth that goes to labour has steadily fallen, with most benefits trickling up.
The impact of enterprise AI continues this trend:
Headcount Reduction (Layoffs): Corporate managers choose to replace human workers with automated systems, leading to layoffs. It's crucial to understand: AI isn't doing the firing; it's a management decision to prioritise profit over people.
Wage Suppression: Even for jobs that remain, the perceived ability of AI to automate tasks or monitor performance puts significant downward pressure on wages. If a machine *could* do it, why pay a human more?
“Disciplining" the Workforce: Management can use the ever-present threat of automation to exert greater control over employees, discouraging demands for better pay or improved working conditions. This isn't about technological advancement; it's about power dynamics.
The Imperfect Machine – Slop, Sabotage, and Still Layoffs
Despite the hype, AI is far from a magic bullet. It's often imperfect, biased, and susceptible to manipulation. We've already seen the rise of "AI Slop": much of AI output is mediocre, unoriginal, and even inaccurate because it's trained on vast amounts of low-quality, "crappy" data from the internet. This is Sturgeon's Law in action – 90% of everything is rubbish, and our AIs are learning from it. This has predictable results:
Worse, these systems are vulnerable to data poisoning attacks3 and sabotage. Malicious actors can intentionally corrupt training data to cause models to fail, spread misinformation, or produce harmful outputs.
It's ironic that corporate managers are willing to lay off thousands of workers based on the promise of a technology that is often flawed, biased, and even susceptible to external attack. This underscores that the primary driver isn't technological perfection or societal benefit, but rather the single-minded pursuit of cost savings and control. Corporations invest heavily, and workers pay the price. The risk is offloaded onto the human labour force while privatising the potential rewards.
It’s time for Accountability and Justice
The future of AI work is not predetermined by technology. Human choices, corporate priorities, and public policy shape it. We must understand that layoffs aren't an inevitable consequence of AI itself but a deliberate decision by corporate managers to maximise profits at human cost. The arrow of accountability points squarely at boardrooms and executive suites. We need policies that shield workers from arbitrary layoffs and ensure fair wages in an economy increasingly shaped by automation. Let's support the development of publicly funded and managed AI systems that prioritise societal benefit, ethical growth, and shared prosperity over private profit. We must demand greater corporate transparency about their AI deployments, energy consumption, and hidden labour practices. This means insisting that decision-making individuals are held responsible for the societal and individual impacts of the AI systems they choose to implement. You can participate in public conversations about AI governance, ensuring that democratic values, privacy, and human dignity are central to its development.
Let's not let corporations hide behind the myth of autonomous AI. It's time to hold them accountable for their choices and demand an AI future that truly serves humanity, not just the bottom line.
It’s time to understand that, like the skilled weavers who resisted the degradation of their skills and the quality of the cloth they produced rather than the technology itself, modern-day workers need to work together to resist the degradation of AI. The Luddites' ultimate defeat came partly because they lacked broader social support and faced violent state suppression. Modern resistance to exploitative AI deployment might succeed where they failed by building coalitions that include consumers who value quality, professionals who understand the importance of expertise, civil society and local organisation concerned about the environment, and policymakers concerned about economic inequality.
“Limits to AI Growth”, Bhardway et al. https://arxiv.org/pdf/2501.17980
“AIs Hidden Human Cost” https://maniainc.com/technology/ais-hidden-human-cost-the-struggle-of-kenyas-data-workforce/
“Medical large language models are vulnerable to data-poisoning attacks“ https://www.nature.com/articles/s41591-024-03445-1