Indicators · Jobs and pay
Office jobs
Normal Direct evidence
Over the twelve months to Oct 2026 these jobs rose 0.1% and all jobs rose 0.6%, so growth in these jobs was 0.5 percentage points below that of all jobs.
The difference in growth is lower than 80% of the readings since 1991.
US recessions Periods when this signal was on
What it measures
Growth over twelve months in US jobs in information and in professional, scientific and technical services, the work most exposed to AI, compared with growth in all jobs.
Why it matters
If AI replaces office work, these jobs would grow more slowly than other jobs while the economy grows. Temporary help is left out because it falls first in any downturn.
When it triggers
Triggers when growth over twelve months in these jobs is 2.41 percentage points or more below that of all jobs, a shortfall larger than in 95% of months from 1990 to 2019.
Where it fits
Jobs in work AI can already do. The argument predicts: Jobs in work AI can already do fall as a share of all jobs while output grows, and graduates fare worse than non-graduates. At the pace measured so far the change is slow, and for a year or two it may stay within the surveys' margin of error.
Other possible causes. Recoveries, in which physical and less-educated work comes back faster after a recession has cut it (2021); a downturn in the technology industry, which cuts information jobs (2001–02).
The prediction
Open Set 8 Oct 2026
- If we are wrong0.04 points or above
- Now−0.50 pointsOct 2026
- Start−0.58 pointsSep 2026
- If we are right−2.41 points or belowby Sep 2028
If our view is right, twelve-month growth in US jobs in information and professional services minus that in all jobs reaches −2.41 points or below by Sep 2028, the threshold at which its rule triggers. The date is two years after the start, because the argument expects this change to stay within the surveys' margin of error for a year or two. Reaching it counts as evidence only if no US recession signal was on during the period it measures, or in the three months after, and the prices US businesses charge were not rising faster than in 95% of quarters from 1990 to 2019.
If it is at 0.04 points or above on two consecutive readings, that counts against our view. That level is the 25th percentile of its readings from 1990 to 2019; three in four of those readings were higher. The deadline refers to the period the data cover. If it passes without a reading of −2.41 points or below, that counts somewhat against our view.
When the rule was set
The rule was set on 8 Oct 2026, when it was not triggered.
How often it has been on
Applied to past data at the end of each month from Apr 1991, using only what had been published by then, it was on in 6% of months to 2019 and 2% since 2020. More than six months from any US recession as dated by the National Bureau of Economic Research, it triggered once, in May 2021. The rule was set with these years in view, so these figures cannot show how well it predicts.
Past triggers
On the data published at the time, it triggered in Mar 2002 and May 2021.