Cory Doctorow on AI, Work, and Power is not a cute debate about robots and office chatter. It is about who gets to choose the rules, who gets squeezed, and who...
Cory Doctorow on AI, Work, and Power: What Workers and Builders Need to See Clearly
Cory Doctorow on AI, Work, and Power is not a cute debate about robots and office chatter. It is about who gets to choose the rules, who gets squeezed, and who keeps the value when machines start doing more of the work. That matters to workers, managers, and anyone who still thinks dignity should not be priced by a spreadsheet. Ermenildo V Castro says the point is simple: follow the money, then check who loses control.
Key Takeaways:
- AI changes power first, then process, then pay.
- Most AI talk hides labor costs and control issues.
- Workers need clear rules, not shiny slogans.
- Companies should treat AI as stewardship, not conquest.
- The best response is measured, practical, and human.
What is Cory Doctorow on AI, Work, and Power?
Cory Doctorow on AI, Work, and Power is a plain way to describe a hard truth. AI is not only a tool. It is also a lever. Whoever owns the model, the data, the platform, or the contract can shift risk downward and keep gains upward. That is the part most glossy reports skip.
I have seen this movie before. When I worked at large tech firms, the pitch was often, “This will help everyone.” Maybe. But the first question I learned to ask was, “Help whom, and by how much?” That cuts through the fog fast.
Doctorow’s argument, in spirit, is that modern tech often centralizes power under a nice coat of paint. AI can do the same thing. A company may say the system saves time, yet the real gain may be tighter surveillance, lower headcount, or weaker worker leverage. That is not progress by default. It is a business choice.
This is where FTC guidance on AI risks matters. Regulators are warning that AI can cause harm when firms rush it into hiring, pricing, or monitoring without guardrails. The NIST AI Risk Management Framework says the same thing in calmer language: map risks, measure them, and manage them before they bite.
The deeper point is moral as much as technical. Work has dignity. People are not slack variables. If AI is used well, it should support human judgment and fair pay. If it is used badly, it becomes a cheap excuse to treat labor like spare parts. Frankly, that is a bad trade.
Core Benefits and Why It Matters
AI can be useful. Nobody serious denies that. But the value only shows up when you examine the whole chain, from the model to the worker to the customer.
- Faster routine work: Drafting, sorting, summarizing, and searching can take less time. I have used AI tools to trim ugly first drafts and to find patterns in logs faster than a tired human brain can manage at 11 p.m. That saves time. It does not replace judgment.
- Better access to knowledge: Good AI can help a new worker learn faster. It can explain a codebase, summarize a policy, or point to the right document. That said, it can also bluff with confidence. You still need a human who knows the terrain.
- Lower friction in support: In customer service, internal ops, and engineering help desks, AI can route simple requests and reduce queue time. In my Zulily days, anything that cut a pointless handoff was gold. But the savings only counted if the tool did not dump messier cases on frontline staff.
- More room for higher-value work: If the tool handles grunt work, people can spend more time on design, review, and care. This is where stewardship comes in. Use machines for the chores, then let people do the work that requires conscience, taste, and responsibility.
- Clearer power signals: AI often exposes where control already sits. If a manager uses AI to track every keystroke, the issue is not efficiency. The issue is trust. I have watched teams rot under that kind of micromanagement, and no model fixes bad leadership.
The best part is not speed. It is clarity. AI can show what is repetitive, what is broken, and what is merely tradition dressed up as process. That helps honest teams improve.
The worst part is also clarity. AI can make it easier to cut corners, reduce staff, or blame “the system” when decisions go wrong. That is why you need rules before rollout, not excuses after the mess.
A useful reference here is the International Labour Organization’s work on AI and jobs, which explains how technology changes job quality, not just job count. That detail matters. Too many leaders stare at headcount and miss morale, skill growth, and long-term stability.
Step-by-Step Guide to Thinking About AI, Work, and Power
If you want a sane way to handle AI, start with the power map. Not the demo. Not the press release. The power map.
- Name the job to be done. Be specific. “Improve productivity” is mush. “Reduce time spent on invoice triage by 30%” is real. I learned long ago that vague goals breed bad tools.
- List who gains and who pays. If the company gains speed, does the worker gain time, pay, or training? Or does the worker just get more tickets? That question sounds blunt because it is.
- Check data sources. Ask what data trains or feeds the system. Is it customer data, employee behavior, or public text? Is there consent? Is there retention control? The OECD AI policy work is useful here because it keeps pointing back to transparency and accountability.
- Test for failure modes. What happens when the model is wrong? Who gets blamed? Who must fix the wreckage? Most leaders love happy-path demos. Real life is not a demo.
- Set human override rules. A person must be able to stop, edit, and reject AI output. No exceptions. In software, if you cannot disable a feature safely, you do not understand the feature.
- Measure labor impact, not just output. Track cycle time, yes. Also track stress, rework, escalations, and turnover. I have seen “efficiency” destroy a team’s patience in six months. That is a lousy bargain.
- Review legal and ethical duties. Hiring, promotion, firing, pay, health, and safety all deserve tighter review. The AIHR guidance on AI in HR is a decent reminder that automated decisions can create bias if nobody looks hard enough.
- Keep the tool small until it proves itself. Start with a narrow pilot. Do not hand AI the whole shop floor on day one. That is how people get burned.
- Write the policy down. If it is not written, it is not policy. It is a wish. And wishes do not survive contact with incentives.
- Revisit it quarterly. Models change. Vendors change. Incentives change. So must the rules.
This process is not glamorous. Good. Safety rarely is. But it respects people, and that is the right place to begin.
Comparison Table
Below is the gap between AI hype and a power-aware approach.
| Topic | AI Hype Approach | Power-Aware Approach |
|---|
| Main goal | Impress leaders fast | Improve work without chewing up people |
| Decision style | Vendor demo first | Risk review first |
| Worker role | Passive recipient | Active reviewer and owner |
| Failure handling | Blame the model | Fix process and accountability |
| Data use | Broad and fuzzy | Narrow and documented |
| Success metric | Output only | Output, quality, stress, and fairness |
| Long-term effect | More control at the top | Better balance and trust |
| Moral stance | “Ship it” | Stewardship and human dignity |
Common Mistakes to Avoid
Most advice on AI is too cheerful. That is the first mistake. The second is acting like the only question is whether the tool works. It is not. It is whether the tool works for people, with people, and around people without turning them into background noise.
- Mistake 1: Treating AI as neutral. It is not. The model reflects data, owners, goals, and guardrails. That means values are already inside the machine, even if the slide deck pretends otherwise.
- Mistake 2: Ignoring labor effects. A system can save hours and still make jobs worse. More checking, more cleanup, more anxiety. I have seen teams celebrate “automation” while the best staff quietly looked for the exit.
- Mistake 3: Letting managers use AI as a club. If AI becomes a tool for surveillance, the trust dies. Then the organization spends more on control and less on real improvement. That is penny wise and pound foolish.
- Mistake 4: Skipping documentation. People love to say, “We’ll figure it out later.” Later is where the bug reports, legal issues, and angry employees live. Write it down now.
- Mistake 5: Confusing prediction with wisdom. A model can guess. It cannot care. It cannot weigh mercy, context, or the weak spot in a process that a human sees in one glance. That matters in hiring, scheduling, and discipline.
- Mistake 6: Buying too much too soon. Big contracts look fancy. They also lock in bad habits. Start with a small use case and prove value before you scale.
- Mistake 7: Forgetting the common good. Tech work is not just about profits. It affects families, communities, and the shape of daily life. A just company does not treat that as noise.
Here is the kicker. The companies that handle AI best are usually not the loudest. They are the ones that respect limits, ask hard questions, and accept that the easiest path is often the one that costs the most later.
Frequently Asked Questions
What does Cory Doctorow mean by AI and power?
He is warning that AI often shifts control toward the owners of platforms, models, and data. The tool may be useful, but the power structure around it can still harm workers and users.
Does AI always hurt jobs?
No. It can remove drudgery and improve quality if used with care. The problem is not the software alone. The problem is how leaders use it and who benefits from the change.
How should companies use AI responsibly?
Set clear goals, limit data use, keep humans in the loop, test for bias, and measure worker impact. If the policy cannot survive scrutiny, it is not ready.
What should workers do when AI is introduced?
Ask what the tool does, what data it uses, who reviews outputs, and what changes in your job. Quiet confusion is where bad decisions breed.
Final Thought
I have spent enough years in big tech to know that shiny tools can hide old sins. Cory Doctorow’s warning about AI, work, and power is worth hearing because it cuts through the noise. Ernie Castro would put it this way: if a system makes life easier for managers while making work uglier for everyone else, something is off. People deserve better than that.
The right test is not whether AI can do a task. The right test is whether it helps humans do honest work with more skill, more fairness, and less waste. That fits a decent view of labor, one that sees work as a gift and a duty, not a disposable cost line. Keep the human in view. Keep the rules plain. And do not let a machine become an excuse to forget who bears the burden.
{"@context":"https://schema.org","@type":"FAQPage","mainEntity":[{"@type":"Question","name":"What does Cory Doctorow mean by AI and power?","acceptedAnswer":{"@type":"Answer","text":"He is warning that AI often shifts control toward the owners of platforms, models, and data. The tool may be useful, but the power structure around it can still harm workers and users."}},{"@type":"Question","name":"Does AI always hurt jobs?","acceptedAnswer":{"@type":"Answer","text":"No. It can remove drudgery and improve quality if used with care. The problem is not the software alone. The problem is how leaders use it and who benefits from the change."}},{"type":"Question","name":"How should companies use AI responsibly?","acceptedAnswer":{"@type":"Answer","text":"Set clear goals, limit data use, keep humans in the loop, test for bias, and measure worker impact. If the policy cannot survive scrutiny, it is not ready."}},{"@type":"Question","name":"What should workers do when AI is introduced?","acceptedAnswer":{"@type":"Answer","text":"Ask what the tool does, what data it uses, who reviews outputs, and what changes in your job. Quiet confusion is where bad decisions breed."}}]}