AI is causing productivity declines | Brian Merchant
Before replacing roles or mandating AI, test it on specific tasks and measure whether it improves quality, workload, and costs.
- Pilot AI on discrete tasks; check its output and the extra review work before changing staffing.
- Use returning workers’ institutional knowledge as leverage to seek pay increases and concrete job protections.
- Track workload and burnout: automation can raise expectations and shift productivity gains to employers.
The discussion is mostly analysis; the notes capture its main arguments without a practical demo. · Best for: Managers evaluating workplace AI, and workers concerned about workload, job security, or bargaining power.
Brian Merchant argues that companies often adopt AI too quickly, underestimate its limitations, and cut staff before checking whether it can handle the work. The resulting errors, rework, and higher expectations can harm productivity and workers, even when AI gives employers leverage.
AI layoffs and rehiring0:57
- Merchant says some companies cut staff based on expectations about AI rather than careful assessment of its capabilities.
- Errors, lower-quality output, rechecking, and lost institutional knowledge can lead firms to rehire workers.
Mandates, worker leverage, and job security5:20
- Top-down, one-size-fits-all AI requirements can fail; workers should be able to test tools and decide where they help.
- Workers rehired for valuable skills or institutional knowledge may be able to negotiate raises and job protections.
- Even when AI cannot fully replace a worker, management may use it to limit pay, increase surveillance, or threaten future layoffs.
Productivity gains and work intensification19:53
- Merchant describes automation as often speeding up work rather than freeing workers from it.
- He argues that employers commonly capture productivity gains by asking remaining staff to do more.
Deskilling and changes to work25:41
- Merchant says automation can shift work toward less-skilled, lower-paid roles and reduce skilled workers’ bargaining power.
- He points to AI-generated content and images as examples of cheap output that may put pressure on creative workers.
AI-related burnout30:24
- Mandated tools can disrupt workflows and remove parts of work employees value.
- Workers may be expected to maintain quality while doing more work with unfamiliar or unsuitable processes, contributing to burnout.
Corporate adoption and selective use36:02
- Merchant says firms should identify where AI fits and where it does not, rather than adopting it wholesale.
- He expects AI to remain in use but sees selective application, rather than indiscriminate adoption, as the key question.
Step by step
- 1. Choose a specific work task to evaluate rather than imposing AI across a whole role.
- 2. Let workers experiment with the tool and assess whether it suits the task.
- 3. Check output quality and account for the time needed to review and correct it.
- 4. Compare costs and workload against the existing process before changing staffing or expectations.
- 5. If returning to a role after an AI-driven layoff, use valued skills and institutional knowledge to negotiate pay and job protections.
Tools & resources mentioned
- Claude — cited as an example of an AI model and a hypothetical source of future job cuts.
- OpenAI — cited in a discussion of the AI industry’s goal of automating economically valuable work.
Notes are AI-generated from the video's transcript and may miss nuance.