Work and jobs · A guided reading
Will AI change my job?
Start with the tasks. Then ask what happens to the whole job.
Read this page first for the framework. Follow the three source links when you want to go deeper.
Understand the idea
A job is more than one task.
Imagine a customer-support worker. Writing a reply is part of the job. So is checking the customer’s history, deciding whether a refund is fair, and taking responsibility when something goes wrong.
If AI can draft the reply, the work changes. But that alone does not tell us whether the firm needs fewer people. We still need to know how the rest of the job changes, how much checking is needed, and whether the firm serves more customers.
- Draft a replyAI may help
- Summarise the historyAI may help
- Check the factsStill needs a process
- Resolve an exceptionStill needs a decision
- Own the outcomeStill needs accountability
An illustration, not a forecast. The whole job can be reorganised; the shaded tasks are only a starting point.
Read first · An explanatory essay
The task is not the job
Luis Garicano · Silicon Continent · April 2026
Why I’ve put it here: Garicano asks when a task can be separated from the rest of a job. His answer brings coordination, context and responsibility into the picture.
Look for: the difference between helping someone do a job and breaking that job into separately supplied tasks.
An argument about how work is organised, not a guarantee that jobs will survive.
See the existing library note →Look at firms
Using AI is not the same as cutting staff.
“This firm uses AI” leaves several questions unanswered. Is one employee using it to draft an email? Has the firm changed a whole business process? Has it changed hiring, staffing or output?
Read next · A primary research paper
The Microstructure of AI Diffusion
Bonney and colleagues · U.S. Census Bureau · Working paper, April 2026
Why I’ve put it here: The authors distinguish employees using AI in tasks from firms integrating it across business functions. Those measures have different relationships with reported staff reductions.
Look for: what counts as AI use, whose experience is being measured, and the difference between an association and a demonstrated cause.
U.S. firms, surveyed for November 2025–January 2026. This is evidence about a particular period, not a prediction for every job or for India.
See the existing library note →Carry this forward: ask how the firm reorganised work, not just whether it bought an AI tool.
Look at young workers
A small overall effect can hide a difficult start.
Now complicate the picture. A firm may retain its experienced workers while hiring fewer beginners. The total number of jobs and the opportunity to get a first job are different questions.
Read alongside it · A primary research paper
Canaries in the Coal Mine?
Erik Brynjolfsson, Bharat Chandar & Ruyu Chen · Stanford Digital Economy Lab · Working paper, revised August 2026
Why I’ve put it here: Using U.S. payroll records, the authors find weaker employment among young workers in occupations exposed to AI. They identify reduced hiring as an important part of the pattern.
Look for: the comparison between young and experienced workers, the choice of comparison group, and the authors’ checks for other explanations.
An uneven early pattern in a U.S. payroll dataset. It does not establish one inevitable outcome for all workers, and exposure to AI is not the same as observed replacement by AI.
See the existing library note →Carry this forward: ask who gets to enter a profession and learn its work, as well as who keeps a job.
Use the framework
Try it on a headline.
A made-up example
“AI handles half our support requests. We won’t need half our support team.”
What would you need to know before accepting that conclusion?
- Does “handles” mean resolves the request, or drafts an answer for a person to check?
- Which tasks remain, and how much work do exceptions and checking create?
- Will the firm serve more customers, reduce staff, or hire fewer beginners?
Compare your reasoning
The conclusion does not follow from the first sentence alone. We need evidence about the whole workflow, demand for the service, and the firm’s staffing decisions. The same technology can support more output or fewer workers; the headline has not told us which.
That is the habit to take to the next headline: task → job → firm → who gains or loses.