Author and technology critic Cory Doctorow argues that the most important story may not be what AI can do, but how executives use the technology’s perceived capabilities to change workplace expectations. In his view, generative AI can assist with bounded tasks such as producing code fragments, but it remains poorly suited to the broad organizational reasoning required for complex software engineering.
That distinction matters. If AI tools enhance individual tasks without replacing entire occupations, the technology could still increase productivity. But presenting those tools as substitutes for skilled professionals may give employers more leverage over workers, influence hiring decisions and justify aggressive restructuring before the promised gains have been demonstrated.
Why the AI replacing jobs debate is more complicated
Public discussion often treats a job as if it were a single task. In practice, most professional roles combine technical execution, judgment, communication, institutional knowledge and accountability. AI may automate or accelerate some of those components without being able to assume responsibility for the whole position.
Software engineering illustrates the distinction. Generative AI can suggest functions, explain unfamiliar code, draft tests and help developers troubleshoot. Those abilities can make experienced engineers faster. They do not necessarily mean that an AI system can independently understand business requirements, legacy dependencies, security constraints, user needs and long-term maintenance decisions across a large software environment.
At an October 2025 Brooklyn Public Library event with former Federal Trade Commission Chair Lina Khan, Doctorow summarized his position directly: “I don’t think AI can do your job.” The library’s event centered on his broader critique of declining online platforms and concentrated corporate power. ([Brooklyn Public Library](https://www.bklynlibrary.org/calendar/cory-doctorow-discusses-central-library-dweck-20251006-0700pm))
Doctorow’s argument should be understood as a critical assessment, not a proven forecast. AI capabilities continue to evolve, and organizations are already redesigning workflows around them. The central question is therefore not simply whether AI can perform work. It is whether employers can reliably convert those capabilities into safe, accurate and economical systems that replace complete roles.
AI and software engineering: tasks are not entire jobs
Doctorow draws a line between code generation and software engineering. A model may produce a subroutine from a well-defined prompt, but architecture requires a much broader view: how components interact, what came before, what future changes are likely and how the software fits into adjacent operational and business systems.
This limitation is especially important in high-consequence environments. A generated answer can appear plausible while overlooking security controls, regulatory obligations, performance requirements or undocumented dependencies. Human engineers also negotiate priorities, challenge ambiguous requirements, review tradeoffs and accept accountability for decisions—responsibilities that extend beyond producing syntactically correct code.
None of this makes AI irrelevant. It suggests that the strongest near-term use case may be augmentation rather than autonomous replacement. Companies can benefit when AI removes repetitive work and gives professionals more time for architecture, analysis and collaboration. Results may be weaker when leaders assume that generating output is equivalent to owning an engineering outcome.
How AI narratives can shift power toward employers
Doctorow connects today’s automation rhetoric with a broader change in technology-sector labor conditions. During earlier periods of rapid growth, skilled workers were difficult to replace and could move more easily between employers. Companies often motivated them with a sense of mission: build useful products, expand access to information or change the world.
That mission-driven culture had two sides. It could encourage long hours, but it also gave employees influence when business decisions conflicted with a company’s stated values. Google employees’ opposition to Project Maven, a Pentagon artificial-intelligence initiative, became a prominent example of workers challenging leadership over how technology should be used.
Years of layoffs have altered that balance. Layoff totals vary depending on geography, methodology and whether contractors are counted, so claims about a single two-year figure require caution. Still, independent trackers and news organizations have documented extensive cuts across the technology sector. The continuing uncertainty makes workers less likely to challenge management or leave without another position secured.
Some employers have also promoted demanding schedules resembling China’s “996” model—9 a.m. to 9 p.m., six days a week. Even where that schedule is not formally imposed, pressure to demonstrate productivity can intensify when employees believe automation may eliminate their positions.
From Doctorow’s perspective, that fear has strategic value for management. A workforce that expects replacement may accept heavier workloads, reduced autonomy or weaker bargaining power. Whether or not executives consciously use AI this way in every company, the narrative itself can affect workplace behavior before the technology proves capable of replacing the people involved.
Labor concerns extend beyond office-based tech jobs
Doctorow places the treatment of technology workers within a wider debate about corporate power. He points to Amazon’s warehouse safety record as an example of what can happen when employees lack meaningful leverage.
A 2024 investigation by the U.S. Senate Committee on Health, Education, Labor and Pensions concluded that Amazon’s productivity demands contributed to elevated injury risks. The committee also accused the company of selecting injury statistics that made its safety performance appear stronger. Amazon has disputed characterizations of its safety record and has said it invests in workplace improvements. ([U.S. Senate HELP Committee report](https://www.help.senate.gov/imo/media/doc/amazon_investigation.pdf))
The broader point is that technology does not determine working conditions on its own. Management choices, labor protections, transparency and employee organization shape how productivity gains are distributed. AI could reduce tedious work and create better jobs, or it could be used to increase monitoring and workloads. The outcome depends on governance as much as capability.
Is AI investment creating a bubble?
The labor debate leads to a second question: can AI generate returns large enough to justify the extraordinary capital committed to chips, data centers, energy and foundation models?
Doctorow believes expectations have moved far beyond realistic commercial value. His concern is straightforward: if AI systems cannot replace enough labor or create sufficiently valuable new products, revenue may not cover the enormous capital and operating costs required to build and maintain them.
Khan has raised a related competition concern. During her FTC tenure, the agency examined investments and partnerships connecting Microsoft with OpenAI, Amazon with Anthropic, and Google with Anthropic. A subsequent FTC staff report said these arrangements could affect access to computing resources and engineering talent, raise switching costs and give cloud providers access to sensitive business information. ([Federal Trade Commission](https://www.ftc.gov/news-events/news/press-releases/2025/01/ftc-issues-staff-report-ai-partnerships-investments-study))
These relationships do not by themselves prove that an AI bubble exists. They do, however, complicate the investment picture because money can circulate among a small group of cloud, chip and model providers. Investors must distinguish genuine end-user demand from revenue supported by strategic partnerships and infrastructure commitments within the same ecosystem.
Why financial regulators are watching AI valuations
The Bank of England’s Financial Policy Committee warned in October 2025 that equity valuations appeared stretched, particularly among AI-focused technology companies. It also noted unusually high concentration in major U.S. equity indices and said a shift in expectations about AI’s impact could produce a sharp market correction. ([Bank of England](https://www.bankofengland.co.uk/financial-policy-committee-record/2025/october-2025))
The Bank later explained that a fall in AI-related asset prices would not automatically recreate the 2008 financial crisis. The dot-com crash, for example, contributed to a comparatively mild U.S. recession. However, systemic exposure could grow if future AI infrastructure is financed more heavily with debt, linking banks and private-credit markets more closely to AI valuations. ([Bank of England](https://www.bankofengland.co.uk/bank-overground/2025/all-chips-in-ai-related-asset-valuations-financial-stability-consequences))
This nuance is important. Doctorow’s warning that an AI collapse could be worse than 2008 is his prediction, not a regulatory consensus. The Bank of England has identified plausible vulnerabilities, but it has not declared that an AI crash is inevitable or that its effects would necessarily exceed the global financial crisis.
AI may transform work without eliminating workers
The debate is often framed as a choice between two extremes: AI will replace nearly everyone, or it is useless hype. Reality is likely to be less tidy. AI can be commercially valuable, reshape workflows and reduce demand for certain tasks without becoming a dependable substitute for entire professions.
For employers, the responsible approach is to measure actual outcomes rather than treat automation as an end in itself. That means evaluating accuracy, security, operating cost, failure rates and the human review needed to produce trustworthy results. For workers, it means learning where AI improves performance while protecting the judgment, context and accountability that organizations still need.
Doctorow’s critique ultimately focuses on power. Predictions about **AI replacing jobs** influence budgets, staffing and working conditions today—even if the promised replacement never fully arrives. The future of work will therefore depend not only on what AI systems can accomplish, but also on who controls them, who assumes their risks and who receives the benefits.
By Ece Yildirim
https://gizmodo.com/ai-wont-replace-jobs-tech-bros-want-you-terrified-2000670808
