Cory Doctorow on AI, Work, and Power comes down to one hard point: AI is not just about smart software, it is about who controls labor, money, and attention. I...
Cory Doctorow on AI, Work, and Power: What He Gets Right About Control, Labor, and the Cost of Cheap Intelligence
Cory Doctorow on AI, Work, and Power comes down to one hard point: AI is not just about smart software, it is about who controls labor, money, and attention. I have seen enough tech hype cycles at Google, Microsoft, Amazon, and Zulily to know the pattern. The shiny part gets the headlines. The power grab hides in the fine print.
Key Takeaways
- AI changes work, but the bigger issue is who gets the gains.
- Most AI harm comes from bad incentives, not bad code alone.
- Workers need a say in how AI is used on the job.
- Cheap AI can still be expensive when it cuts trust, wages, and dignity.
- Good rules, clear limits, and human judgment matter more than slogans.
Cory Doctorow keeps pressing the same nerve. AI is not neutral. It is built, sold, and used inside markets that reward speed, control, and profit, often at the expense of workers and the public. That is the part many glossy AI posts skip. I have sat in rooms where leadership praised automation as if it were a prayer. Then the same teams had to fix the mess when output quality fell, morale tanked, and people felt disposable. That is not progress. That is a bill.
If you want a clean explanation of the real fight, this is it: AI changes work, but power decides who pays and who profits. Cory Doctorow’s criticism lands because it points at the structure, not the marketing. And that matters. Good tools should serve people. People are not raw material.
What is Cory Doctorow on AI, Work, and Power?
Cory Doctorow on AI, Work, and Power is a way of looking at artificial intelligence through labor, ownership, and control. It asks who builds the systems, who trains them, who gets replaced, and who gets richer when the dust settles. That is the plain version. No velvet language needed.
Doctorow is not just talking about code quality or model size. He is talking about market behavior, labor rights, and the old habit of using new machines to squeeze people harder. I have seen this before. At one large tech company, a team wanted to automate a support workflow. The pitch sounded clean. Less waiting. Fewer mistakes. Lower cost. But once we mapped the actual work, we found that half the “automation” was really moving human judgment into a dark corner where nobody would notice the harm. That is the trick. Make the worker invisible, then call the result efficient.
This topic sits at the crossing of AI ethics, labor economics, platform power, and workplace design. It also touches on something deeper: the dignity of work. In Catholic social teaching, work is not just a cost to cut. It is part of human dignity, part of stewardship, part of how a person serves the common good. That does not mean every task must stay manual. It means a system should not treat people as trash once software can do a piece of the job.
Doctorow’s view is useful because it refuses fantasy. AI can help. AI can also concentrate power. Both can be true at once. The question is not whether a model can write text, sort tickets, or summarize a meeting. The real question is who controls the tool, who is watched by it, and who bears the downside when it guesses wrong.
If you want a broader frame on the same issue, see the real-world labor effects described in this ILO future of work overview, the practical AI risk notes from NIST’s AI Risk Management Framework, and the policy concerns in the OECD AI policy work. Those sources do not spin fairy tales. They speak in facts.
Core Benefits and Why This Debate Matters
This debate matters because AI is already changing how work gets done. And no, not always for the better.
- It exposes bad incentives. When companies rush AI into production just to cut headcount, the damage shows up fast. I have watched teams celebrate cost savings, then spend months cleaning up broken customer experiences. Cheap input does not mean cheap output.
- It protects worker dignity. People are not widgets. They carry judgment, memory, and care. A machine can sort records. It cannot own responsibility. That difference matters when customers are upset, when errors hit money, or when a hard decision needs a human face.
- It improves policy thinking. Doctorow’s arguments help leaders ask better questions. Who owns the model? Who audits it? Who gets blamed when it fails? I wish more executives asked those before buying the demo.
- It keeps us honest about productivity. AI may raise output in one slice of work, but that gain can vanish if the system creates more rework, more review, and more noise. I’ve seen this at scale. A workflow that looks faster on a slide can be slower in real life.
- It forces a talk about fairness. If AI boosts profits, who gets the share? Workers? Executives? Shareholders only? The old answer is often “mostly shareholders.” That answer is legal in many places, but it is not automatically just.
- It helps leaders avoid false calm. Most people talk about AI like it is one thing. It is not. It is a bundle of tools, limits, costs, and tradeoffs. The details matter. That is where the truth lives.
- It fits real stewardship. Tools should be used with care, not worshiped like idols. That sounds old-fashioned. Fine. Old-fashioned is useful when the new thing is making everyone act careless.
Here is the kicker. A company can have good AI and still make rotten choices. A company can also have mediocre AI and make fair choices that help workers and customers. The tool is only part of the story. The moral frame matters more.
I learned that the hard way during a product review at Amazon. The team had strong metrics and weak trust. The numbers looked fine until we asked frontline people what the system was doing to their day. The answer was blunt. It was adding friction, stress, and a quiet sense that management did not care. That sort of thing bleeds out over time. People stop bringing problems up. Then quality falls. Then leadership wonders why.
That is why the topic is bigger than “AI will take jobs” or “AI will create jobs.” Those are half-questions. The better question is whether AI will create better work, better judgment, and fairer rules. That is the standard worth keeping.
Step-by-Step Guide to Thinking Clearly About AI, Work, and Power
You do not need a wizard hat to deal with this. You need a process.
- Map the actual work first. Before buying AI, list every step in the task. Who starts it? Who checks it? Where do errors land? I have watched teams buy tools before they understood the job. That is backwards. If you do not know the work, you will automate a mess.
- Find the power shift. Ask who gains leverage. Does AI help a manager watch workers more closely? Does it let one vendor lock the company in? Does it push labor risk downward while profits rise upward? Be plain about it. The truth is usually sitting there, waiting to be named.
- Set human override rules. A human must be able to stop the machine. Not in theory. In practice. That means clear escalation paths, review steps, and a way to fix bad outputs. If nobody can say no, then the system is not a tool. It is a boss.
- Measure harm, not just speed. Teams love speed charts. Fine. But also track rework, complaint rates, error spread, burnout signals, and customer trust. I have seen “efficiency” claims collapse when the hidden cost showed up in support tickets and staff turnover.
- Protect workers from silent replacement. If a company uses AI to remove roles, it should not pretend the move is painless. Give notice. Offer retraining. Share the gains. Treat people like human beings, not line items. That is basic justice.
- Audit data and outputs. AI systems absorb bias from their training data and from the business rules around them. Regular audits matter. So do small tests with real users. NIST has useful guidance here, and it is worth reading instead of guessing.
- Keep the system legible. If nobody can explain why the model made a call, trust will decay. Simple explanations beat fancy charts. I have seen engineers hide behind “the model said so.” That is weak sauce. Responsibility cannot be outsourced to software.
- Review contracts like a skeptic. Vendors often promise safety and flexibility, then bury the limits. Check data rights, retention, indemnity, and exit terms. The cheapest deal can become the most expensive trap.
- Leave room for judgment. A good worker sees context a model misses. A sick customer. A weird case. A fairness issue. People are good at noticing what rules miss. AI should support that, not flatten it.
- Keep asking who benefits. That one question cuts through the smoke. If the answer is “mostly the vendor and the executives,” then you probably have your answer.
For a practical guardrail view, the World Economic Forum’s AI and future of work discussion gives a broad business frame, while the White House Blueprint for an AI Bill of Rights gives a stronger public-interest angle. I do not agree with every line in every policy note, but at least these sources admit that power matters.
Comparison Table: Doctorow’s Power Lens vs. the Usual AI Sales Pitch
| Aspect |
Cory Doctorow’s Lens on AI |
Common AI Sales Pitch |
| Main question | Who gains power and who loses it? | How fast can we deploy? |
| View of workers | People have rights and judgment | Labor is a cost center |
| View of errors | Errors can harm real people | Errors are edge cases |
| View of vendors | Watch lock-in and control | Trust the platform |
| View of regulation | Needed to curb abuse | A hurdle to growth |
| View of productivity | Must include quality and trust | Mostly measured by speed |
| View of ethics | Central to system design | Added later, if at all |
| Likely result | Slower hype, better accountability | Faster rollout, hidden harm |
The gap between these two views is not small. It is the whole argument.
Common Mistakes to Avoid When Talking About AI and Work
Most AI advice is bloated nonsense. There. I said it.
- Mistake 1: Treating AI as magic. It is software. Fancy software, yes. Magic, no. If your plan depends on perfect answers from messy data, the plan is weak.
- Mistake 2: Ignoring labor costs. Companies love to count cloud spend and forget training, review, cleanup, and morale. That is lazy math. I have seen projects fail because leaders counted licenses and ignored people.
- Mistake 3: Calling surveillance productivity. Watching workers more closely is not the same as helping them work better. Sometimes it just makes people nervous and slower.
- Mistake 4: Assuming all automation is good. Some tasks should be automated. Some should not. If a process involves care, judgment, or moral weight, a human should stay in the loop.
- Mistake 5: Letting vendors define the problem. They will always tell you the answer they sell. That is their job. Your job is to ask a cleaner question.
- Mistake 6: Forgetting the customer. A system can look efficient while making users miserable. I’ve watched support queues shrink and complaints rise. That is not a win. That is sleight of hand.
- Mistake 7: Pretending power is neutral. It is not. If a tool gives one side more control over workers, data, or speech, that is a power move. Say it plain.
- Mistake 8: Skipping audits because the demo was smooth. Demos are staged. Reality is not.
- Mistake 9: Letting fear drive policy. Panic leads to bad rules. Better to measure, test, and protect the public good with calm heads.
- Mistake 10: Forgetting the human soul of work. Work is not just output. It is craft, service, and responsibility. That old truth still stands, even if the slide deck says otherwise.
I used to think teams would learn this on their own. They do not, not usually. Pressure makes people cut corners. That is why clear principles matter. The point is not to block AI. The point is to stop it from turning work into a cheaper, meaner version of itself.
Frequently Asked Questions
What does Cory Doctorow mean by AI and power?
He means AI should be judged by who controls it, who profits from it, and who bears the harm when it fails. The tech is not separate from the politics around it.
Does Doctorow think AI is always bad for workers?
No. The point is not that every use is harmful. The point is that many deployments are shaped by cost-cutting, monitoring, and control rather than fair labor practices.
How can companies use AI without hurting workers?
They can set human review steps, share gains, limit surveillance, audit outputs, and involve workers early. That is slower than hype, but it is cleaner and more honest.
Why do critics focus so much on labor and power?
Because those are the parts companies often hide. A tool can look smart while making the workplace harsher. If you ignore labor, you miss the real price.
Final Thought
I keep coming back to one memory. Years ago, in a Seattle meeting room, I watched a smart team celebrate a system that shaved minutes off a workflow. It looked good on paper. Then I spoke with the people doing the work. They were carrying the mess, the retries, and the stress. The machine got the praise. The humans got the burden.
That is why Cory Doctorow’s warning matters. It is not anti-technology. It is pro-reality. It asks us to treat tools as tools, workers as people, and profit as something that must answer to justice. That is a hard line, but a fair one. In the end, good stewardship means using skill for the common good, not just for speed. If I have learned anything from years in big tech and from life here as Ermenildo V Castro, Jr., it is that power always wants a little more room. Our job is to give it boundaries.
That is not cynicism. That is wisdom.