Audit Packet: AI Took the Starter Tasks. Who Teaches Judgment Now?

Brief 025 asks how new workers gain judgment after AI takes the codified tasks entry-level jobs used as practice. The closest published neighbor is Brief 018, on what remains of a role for people already in it.

Audit verdict: Supported at the mechanism layer. Payroll, an employer survey, and a literature argument carry the task shift. A leader survey carries the thin design response. Danish earnings and hours, and a hiring-expectation split, bound any headcount reading. No source measures skill formation directly. The join and the 12-month timing are Autonoma synthesis.

• [1] Brynjolfsson, Chandar, and Chen, "No Widespread Displacement, but the AI Employment Gap for Young Workers Has Widened to 19%." Stanford Digital Economy Lab, 12 August 2026. Peer gap, hiring, codified versus tacit, quintile footnote. Descriptive. https://digitaleconomy.stanford.edu/news/canariesaug26
• [2] Strada Institute for the Future of Work, "Entry-Level Hiring in the AI Era: What Employers Are Thinking (and Doing)." 19 May 2026. Artemis Strategy Group, March 3-22, 2026. Holds 42%, 33%, 41% routine, 20%, N=1,387, and 2.7 times. https://cdn.prod.website-files.com/6777c52f82e5471a3732ea25/6a0c949112a573adb34f071e_Strada_EntryLevelHiringAIEraReport_May_2026.pdf Article: https://www.strada.org/news-insights/entry-level-hiring-in-the-ai-era-what-employers-are-thinking-and-doing
• [3] Andre and Schief, "A potential boost from AI in ageing societies. Early insights." OECD Working Paper 1870, landing page 5 July 2026. May hinder learning on the job, and may help knowledge pass. Not a new estimate. https://www.oecd.org/content/dam/oecd/en/publications/reports/2026/07/a-potential-boost-from-ai-in-ageing-societies_144607bb/be002e40-en.pdf
• [4] Deloitte, "2026 Global Human Capital Trends." Press release, 4 March 2026. 85%, 7%, and more than 9,000 leaders in 89 countries. https://www.deloitte.com/us/en/about/press-room/deloitte-report-winning-organizations-will-build-the-human-advantage.html
• [5] Deloitte, "Human-AI interaction design." 2026 Global Human Capital Trends chapter. 66% and 6% leading. No separate chapter date. Dated from 4 March 2026. https://www.deloitte.com/global/en/insights/topics/talent/human-capital-trends/2026/human-ai-interaction-design.html
• [6] Humlum and Vestergaard, "Still waters, rapid currents: Early labor market transformation under generative AI." Brookings, 6 October 2026. Nulls, the 2% bound, task movement, and the adoption line. https://www.brookings.edu/articles/still-waters-rapid-currents-early-labor-market-transformation-under-generative-ai/
• [7] Humlum and Vestergaard, "NBER Working Paper 33777." NBER, issued May 2025, revised March 2026. Recorded hours through December 2024, two years after ChatGPT. https://www.nber.org/papers/w33777

Unused here: a postings note, an enterprise ROI survey, memory-write studies, an agent-evaluation preprint, other Deloitte shares, and a Strada webinar skill ranking.

• A payroll description is not a cause, and a peer gap is not a raw decline. Education controls shrink the Stanford gaps. Pre-trends, the ADP sample, and specification changes are part of the authors' own limit.
• Survey shares that allow more than one answer do not partition the employers. 41% and 33% are different answers.
• An OECD "may," including the possibility that AI helps knowledge pass, is not an employment estimate.
• Leader self-reports are not a census of designed practice paths. The 6% line is "leading."
• Null earnings and hours through December 2024 do not speak to the task mix, which the same research says did change.
• Briefs 018, 016, and 004 stay on their own questions, as do 006 and 008. No legal advice, and no finding about a named employer.

The strong objection is null effects on earnings and hours, and more leaders expect AI to increase entry-level hiring than a decrease. Brookings and NBER rule out average effects larger than 2%. Tasks still moved. Strada's 2.7 times split is a 2026 hiring expectation.

Null effects on earnings and hours sit beside a thinner task list. More seats are not a practice path. The limiting case is a study that finds judgment still forms at the old rate after those tasks are gone. That study is absent.

This Brief is wrong, or has to be rewritten, if:

• the age-and-exposure gap is shown to be only education mix or a pre-trend, with no codified-versus-tacit split left; or
• employers are not raising judgment expectations while cutting skill-building tasks; or
• new workers in codified jobs are shown to gain the same tacit judgment after the starter tasks are removed, with no added practice path.

It gets tighter if a measured study shows slower judgment after that task shift. A headcount prediction is a different question.

The sentence "Over the next 12 months, employers that keep taking codified starter tasks off entry-level jobs will still expect the judgment those tasks used to build, unless they give new workers another way to practice" is Autonoma Intelligence synthesis. The timing is not a fact quoted from any source above.

This audit checks Brief 025 against the public sources in the ledger, and it holds each figure to the page that states it. The join across the employment pattern, the task shift, and the thin design response, plus the 12-month timing, is Autonoma Intelligence synthesis. This draft was prepared with an AI assistant and reviewed and approved by the editor.