Great post, Mr. Salathé! Two small additions from my side:
First, have you seen the paper by Hosseini and Lichtinger (2025)? They use LinkedIn résumé and job posting data from the US and actually do find differential effects between junior and senior positions - that might be worth considering alongside the other findings.
Second, something I'm observing in my own work right now that Kane (2017) explained really well: digital disruption is fundamentally a people problem - technology changes faster than individuals can adopt it, and individuals adapt faster than organizations can. What I'm seeing is that only now in 2026 (okay, to be fair, late 2025) companies are slowly starting to understand how to actually deploy these technologies in their workflows. So I think these effects might simply be delayed. The interest rate explanation is certainly correct, but that doesn't rule out AI effects becoming visible as organizational adoption catches up.
Looking forward to more insights on this. thank you for sharing!
PS: Claude helped me write this comment since English is not my native language and while refining my words it also had a fun closing line:
The canary may not have sung yet - but we should keep listening.
References
Hosseini, S. M., & Lichtinger, G. (2025). Generative AI as seniority-biased technological change: Evidence from U.S. résumé and job posting data. SSRN. https://doi.org/10.2139/ssrn.5425555
Thank you, this is interesting. Yes, there are a couple of studies now, the Stanford one was just one of the most talked about. They all suffer from the same problem in that they can't really establish causality, and always hedge accordingly ("this could be AI, but we can't be sure"). Naturally, the same is true for the study I talk about here, but I find the argument much more convincing, for all the reasons outlined. We will have to wait and see, and I do agree with you that the effects may always be delayed. We must keep our minds open. Thanks again!
The interest rate overlay is the most underrated chart in this whole conversation. Everyone wants a clean AI-killed-jobs narrative and the macro timing just doesn't support it. I've been looking at this from the hiring side and the picture is similar; the pie isn't shrinking, it's changing shape. Wrote about how companies need to rethink the whole playbook here: https://reading.sh/the-hiring-playbook-in-2026-looks-nothing-like-2023-9fe9c2a52ffb
Great post, Mr. Salathé! Two small additions from my side:
First, have you seen the paper by Hosseini and Lichtinger (2025)? They use LinkedIn résumé and job posting data from the US and actually do find differential effects between junior and senior positions - that might be worth considering alongside the other findings.
Second, something I'm observing in my own work right now that Kane (2017) explained really well: digital disruption is fundamentally a people problem - technology changes faster than individuals can adopt it, and individuals adapt faster than organizations can. What I'm seeing is that only now in 2026 (okay, to be fair, late 2025) companies are slowly starting to understand how to actually deploy these technologies in their workflows. So I think these effects might simply be delayed. The interest rate explanation is certainly correct, but that doesn't rule out AI effects becoming visible as organizational adoption catches up.
Looking forward to more insights on this. thank you for sharing!
PS: Claude helped me write this comment since English is not my native language and while refining my words it also had a fun closing line:
The canary may not have sung yet - but we should keep listening.
References
Hosseini, S. M., & Lichtinger, G. (2025). Generative AI as seniority-biased technological change: Evidence from U.S. résumé and job posting data. SSRN. https://doi.org/10.2139/ssrn.5425555
Kane, G. C. (2017, September 18).
Digital disruption is a people problem. MIT Sloan Management Review. https://sloanreview.mit.edu/article/digital-disruption-is-a-people-problem/
Thank you, this is interesting. Yes, there are a couple of studies now, the Stanford one was just one of the most talked about. They all suffer from the same problem in that they can't really establish causality, and always hedge accordingly ("this could be AI, but we can't be sure"). Naturally, the same is true for the study I talk about here, but I find the argument much more convincing, for all the reasons outlined. We will have to wait and see, and I do agree with you that the effects may always be delayed. We must keep our minds open. Thanks again!
The interest rate overlay is the most underrated chart in this whole conversation. Everyone wants a clean AI-killed-jobs narrative and the macro timing just doesn't support it. I've been looking at this from the hiring side and the picture is similar; the pie isn't shrinking, it's changing shape. Wrote about how companies need to rethink the whole playbook here: https://reading.sh/the-hiring-playbook-in-2026-looks-nothing-like-2023-9fe9c2a52ffb
Great insights. Really interesting post - thanks for sharing.