The exact text we saved

This is the version exactly as it was saved, character for character. The SHA-256 hash below is a 64-character code worked out from that text. Run the same text through SHA-256 and you get the same code; change a single letter and the code comes out completely different. Each version also stores the previous version’s hash, which chains them together so no version can be swapped out quietly.

TRUVACE RECORD VERSION
record: TRV-2026-2720
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-12T17:14:34.202205Z
status: published
lens: p_space
sector: labor
headline: EFFECTS OF ARTIFICIAL INTELLIGENCE AND ROBOTIC TECHNOLOGIES ON THE FUTURE OF WORK: A META-ANALYSIS STUDY
dek: This meta-analysis examined the impacts of artificial intelligence and robotic technologies on the labor market through 18 empirical studies. Findings were evaluated under three main categories: (1) the average automation susceptibility rate was determined to be 21.3%, with methodological approach and publication year having significant effects on results; (2) while technology had a statistically insignificant slight negative effect on total employment, developed economies showed a decrease in routine manual tas…
gain_title: (none)
problem_title: A meta-analysis of 18 studies found that robot use in developed economies was linked to a significant decrease in employment in routine manual tasks (g = -0.441, p < 0.001).
trace_subject: (none)
gain_reading: (none)
gain_evidence: (none)
problem_reading: A meta-analysis of 18 studies found that robot use in developed economies was linked to a significant decrease in employment in routine manual tasks (g = -0.441, p < 0.001).
problem_evidence: a significant decrease in employment in routine manual tasks (g = -0.441, p < 0.001)
quick_read: A meta-analysis by researchers at Yıldız Technical University pooled 18 empirical studies published between 2015 and 2024 on how AI and robots affect work. Pooling five studies on automation risk gave an average susceptibility rate of 21.3%, well below the figures behind many public warnings about robots taking jobs.

The effect on total employment was slightly negative but not statistically significant (Hedges' g = -0.155, 95% CI -0.323 to 0.013, p = 0.070, seven studies). The shifts showed up by task type: in developed economies, robot use cut employment in routine manual tasks and raised it in non-routine analytical tasks, while developing economies showed no significant effect for any task type.

The authors conclude that technology reshapes jobs more than it removes them and call for policies tailored to each country and sector. The pooled studies differ widely in method and the results varied a lot between them, so the figures describe an average across very different studies.
limitation: The included studies differ in method, which makes comparison difficult, and the literature does not yet adequately cover generative AI or the long-term effects of shocks such as COVID-19.
tag: Evidence-backed problem
key_points: Five studies on automation risk gave a pooled susceptibility rate of 21.3% (log odds ratio -1.312, 95% CI -2.042 to -0.581), with very high heterogeneity (I² = 99.4%). | Studies using a task-based approach reported significantly lower automation susceptibility than occupation-based studies (β = -1.364, p < 0.001); post-2017 studies also gave lower estimates (β = -0.826, p = 0.014). | Across seven employment studies, the overall effect was slightly negative but not significant (g = -0.155, p = 0.070); studies published from 2020 onward found more negative effects (β = -0.279, p = 0.027). | In developed economies, robot use cut employment in routine manual tasks (g = -0.441) and raised it in non-routine analytical tasks (g = 0.501); developing economies showed no significant effect for any task type. | Four studies on skills gave a pooled Fisher's z of 0.382 (r = 0.364, p = 0.016), a positive and significant link between these technologies and skill requirements. | IBM and IFR reports show STEM skills falling from 1st place in 2016 to 12th in 2023, while communication, teamwork and time management rose from 10th to 1st. | IBM (2023) is cited as estimating that 40% of the workforce will need to reskill within three years because of AI and automation.
rundown: The authors searched Web of Science, Scopus and Google Scholar plus grey literature, following PRISMA guidelines. They found 1,874 records, removed 423 duplicates and screened 1,451 titles and abstracts. They then read 142 full texts and kept 18 studies, with two independent reviewers at each stage and a third resolving disagreements. Included studies had to be in English, published 2015 to 2024, and contain quantitative data on automation risk, employment or skills.

Studies were scored 1 to 5 on five quality criteria and analysed with random-effects models in Comprehensive Meta-Analysis software. Heterogeneity was high in every category (I² of 99.4% for automation risk, 82.18% for employment and 98.9% for skills). Funnel plots, Egger tests and trim-and-fill tests found no significant publication bias.

The automation-risk estimates pooled five studies: Frey and Osborne, Arntz and colleagues, Nedelkoska and Quintini, Hawksworth and colleagues, and Manyika and colleagues. Meta-regression with four moderators explained 83.4% of the variation between them. The authors read the task-based results as a sign that technology tends to automate specific tasks rather than whole occupations.

On skills, six studies were reviewed but only four could be pooled, because the IFR report is qualitative and the Fossen and Sorgner study is focused on entrepreneurship. The authors recommend country-specific strategies, training in both technical and social-emotional skills, support for new kinds of jobs and gradual management of workforce transitions.
sources:
- peer_reviewed | Journal of Global Strategic Management | https://doi.org/10.20460/jgsm.2025.347 | 2025-06-01
prev: 0000000000000000000000000000000000000000000000000000000000000000

SHA-256 hash

c16300c0c6e39cb74733faa6f222be9425ef9c6ea960eaa0fb08cf8bc6e1b501

Previous version’s hash

0000000000000000000000000000000000000000000000000000000000000000