Post-publication Comment · Critical AI
Comment on “Artificial intelligence adoption and the demand for managerial expertise”
Critical AI · published 2026-06-20 · v1.0 · CRIT-GEN-scp-artificial-intellige
Concerning: Liudmila Alekseeva, José Azar, Mireia Giné, Sampsa Samila · Strategic Management Journal · 2026-05-06
Why this paper was selected
Critique generated in-session via npm run critical-ai (with ledger-aware faithfulness self-check), grounded in the verified OpenAlex abstract of the SMJ study on AI adoption + managerial demand. Associational framing preserved; severity capped moderate; fabricatedCitations=0.
AI/AGI centrality 2/5 · societal relevance 4/5 · source-journal note: Tier S per the determination; ingested from an AGISS critique artifact.
Summary
This study uses US job ads (2010-2022) to show that companies advertising for more AI skills also advertise for more managers, a larger share of manager roles, and different managerial skills — more stakeholder management, creativity, and sales, less budgeting and planning. The authors are careful to call this an association, not proof that AI causes the change, so the right critique is about how things are measured, not about a missing causal experiment. The biggest issue, on the critic's reading, is that both sides of the relationship appear to come from the same job-ad data: a firm that simply posts more ads overall could automatically look like both a bigger AI adopter and a bigger manager-hirer, which could create the headline link on its own. Other limits the critic raises: 'AI adoption' is really 'asking for AI skills in ads' (not measured AI use), 'managerial demand' is ads (not actual hires), the story about skills shifting 'toward growth' depends on how the authors labeled skill categories, and the findings only cover US online ads in a pre-ChatGPT era. The contribution is real and honestly framed, but the central relationship and its interpretation rest on proxies and labels the abstract does not validate.
Central claims & evidence map
| Claim | Type | Evidence offered | Support | Overclaiming | Main weakness |
|---|---|---|---|---|---|
| Firms with greater AI adoption are associated with posting more managerial vacancies and a higher share of managerial vacancies than less intensive adopters; this is the paper's central reported relationship, framed associationally. | The abstract states the authors "show that firms with greater AI adoption post more managerial vacancies and a higher share of such vacancies than less intensive adopters," labelling these "these relationships." | Moderate | None | On the critic's reading, same-source measurement: a shared posting-intensity driver could inflate the count association; the abstract reports no posting-volume normalization or firm-size conditioning. | |
| AI adoption is operationalized as the presence of AI-related skills in a firm's Lightcast job postings rather than as observed deployment, spending, or use of AI. | The abstract specifies "a skills-based measure of AI adoption derived from Lightcast job postings." | Moderate | Minor | On the critic's reading, postings-as-adoption conflates recruiting/advertising behavior with deployment, and shares its data substrate with the outcome. | |
| The positive AI-adoption/managerial-demand relationship is reported as strongest in manufacturing. | The abstract states the relationships "are strongest in manufacturing." | Moderate | Minor | Subgroup association may, on the critic's reading, reflect sector differences in posting/coverage rather than substance. | |
| The relationship is also reported as strongest among firms with higher R&D intensity, glossed as more innovative firms. | The abstract reports the relationships are strongest "among firms with higher research & development intensity," and the managerial summary refers to "more innovative firms." | Moderate | Minor | Interpretive slide from an input proxy (R&D intensity) to an outcome construct (innovation) — a slide the abstract itself makes — plus, on the critic's reading, shared posting-intensity exposure. | |
| Greater AI adoption is associated with shifts in managerial skill requirements toward interpersonal and growth-oriented skills, specifically stakeholder management, creativity, and sales management. | The abstract reports shifts "toward interpersonal and growth-oriented skills, including stakeholder management, creativity, and sales management." | Moderate | Minor | The 'growth/interpersonal' characterization is a researcher-imposed label not externally validated in the abstract. | |
| Greater AI adoption is associated with shifts away from routine administrative managerial skills, specifically budgeting, planning, staff management, and customer service. | The abstract reports shifts "away from routine administrative skills such as budgeting, planning, staff management, and customer service." | Weak | Moderate | Several 'routine administrative' labels (planning, staff management) are, on the critic's reading, defensibly coordination/leadership skills, making the contrast contestable. | |
| The authors interpret the demand and skill-shift findings as suggesting a reconfiguration of managerial roles toward capabilities facilitating scaling, coordination, and adaptation in AI-enabled environments. | Theoretical | The abstract states "the results suggest a reconfiguration of managerial roles toward capabilities facilitating scaling, coordination, and adaptation in AI-enabled environments." | Weak | Moderate | Interpretive gloss inferred, on the critic's reading, from contestable labels; the word 'reconfiguration' connotes a process the associational design does not establish, though the abstract presents it only as what the results 'suggest'. |
| The findings are scoped to US job postings observed over 2010 to 2022, bounding the population and time window. | The managerial summary specifies "US job postings data from 2010 to 2022." | Strong | None | Lightcast coverage caveats and the digitization-of-postings trend are not flagged in the abstract; these are the critic's inferences. | |
| The paper positions itself as examining how AI adoption relates to managerial demand, explicitly characterizing its findings as relationships rather than causal effects. | The abstract says the paper "examines how firms' adoption of artificial intelligence (AI) relates to the demand for managers," labels its findings "These relationships," and hedges that "the results suggest a reconfiguration." | Strong | None | None in the framing itself; the only risk, on the critic's reading, is interpretive language inviting causal reader slippage. | |
| The growth-oriented interpretation presumes the upward-shifting skill categories genuinely capture growth/interpersonal capabilities and the declining categories are genuinely routine, resting on the chosen skill-category labeling. | The interpretive contrast between rising "interpersonal and growth-oriented skills" and declining "routine administrative skills" carries the conclusion's narrative weight. | Weak | Moderate | Unvalidated, researcher-imposed taxonomy on which, on the critic's reading, the headline interpretation circularly depends. |
Per-claim assessment
c1. Firms with greater AI adoption are associated with posting more managerial vacancies and a higher share of managerial vacancies than less intensive adopters; this is the paper's central reported relationship, framed associationally.
The headline relationship is stated carefully as an association, not an effect, which is appropriate to the design. Its credibility is bounded by what is, on the critic's reading, a same-source measurement issue: both the AI-adoption regressor and the managerial-vacancy outcome are constructed from the same Lightcast postings corpus (the abstract states adoption is "derived from Lightcast job postings" and the outcome is "managerial vacancies" posted by firms), so a firm's overall posting or hiring intensity could in principle load on both sides and partly generate the correlation. This is the critic's inference about the measurement, not a stated limitation. The accompanying share outcome is a useful partial guard against a pure level-count artifact.
c2. AI adoption is operationalized as the presence of AI-related skills in a firm's Lightcast job postings rather than as observed deployment, spending, or use of AI.
This is an honestly labelled proxy: a skills-in-postings signal plausibly captures advertised intent to build AI capacity, which is scalable across 2010-2022 in a way survey or spending data are not. On the critic's reading, the construct measures stated hiring intent rather than realized adoption, and because it is drawn from the same corpus as the outcome it may partly re-measure general recruiting behavior. The abstract does not validate the gap between advertised AI-skill demand and actual deployment; this gap is a construct-validity inference, not a claim the abstract disputes or addresses.
c3. The positive AI-adoption/managerial-demand relationship is reported as strongest in manufacturing.
The heterogeneity is theoretically coherent: manufacturing is a setting where AI integration into complex operations could plausibly reshape coordination needs, and the non-uniformity adds internal structure to the pattern. However, if the same-source reading holds, this subgroup association inherits the same concern, since manufacturing firms may differ systematically in posting and skill-tagging behavior; the abstract gives no indication the moderation is net of differential posting volume.
c4. The relationship is also reported as strongest among firms with higher R&D intensity, glossed as more innovative firms.
The R&D moderator is a sensible expected pattern. The gloss equating R&D intensity with innovativeness is the abstract's own move (the managerial summary writes "more innovative firms" where the research summary writes "higher research & development intensity"); the critic's point is that R&D intensity is an input proxy, not innovation itself, so the equivalence is interpretive. As with the manufacturing cut, on the critic's reading the heterogeneity is associational and potentially entangled with differential posting behavior across high-R&D firms.
c5. Greater AI adoption is associated with shifts in managerial skill requirements toward interpersonal and growth-oriented skills, specifically stakeholder management, creativity, and sales management.
The directional, within-role compositional shift is a genuine strength: a pure posting-volume confound would not obviously predict a reallocation among managerial skills toward specific named categories. The substantive read depends, however, on the chosen labeling of these categories as interpersonal/growth-oriented (the abstract's own characterization), which the abstract does not validate against an external taxonomy or inter-coder agreement.
c6. Greater AI adoption is associated with shifts away from routine administrative managerial skills, specifically budgeting, planning, staff management, and customer service.
The declining categories are reported by the abstract; the critic's contention is that their classification as 'routine administrative' (the abstract's phrase) is contestable: planning and staff management are arguably core coordination and leadership functions, and 'sales management' (placed on the rising side) is itself a management function. The toward/away contrast that carries theoretical weight thus rests, on the critic's reading, on a labeling choice the abstract does not defend with any external scheme or validation.
c7. The authors interpret the demand and skill-shift findings as suggesting a reconfiguration of managerial roles toward capabilities facilitating scaling, coordination, and adaptation in AI-enabled environments.
This is an interpretive synthesis layered on skill-frequency shifts in postings; the abstract presents no direct measure of scaling, coordination, or adaptation outcomes, which (on the critic's reading) are inferred from the skill labels themselves. The hedge "the results suggest" is appropriate and the abstract does not claim a causal or dynamic process. The critic's concern is a circularity risk: the conclusion presumes the growth/routine distinction that the labeling already encodes, and a cross-firm association does not by itself establish a dynamic 'reconfiguration' process even though the abstract only frames it as a suggestion.
c8. The findings are scoped to US job postings observed over 2010 to 2022, bounding the population and time window.
The scope is transparently disclosed, which is responsible. The critic's added caveats are inferences not in the abstract: Lightcast coverage is plausibly uneven across firm size, sector, and online-posting propensity, and the propensity to post online itself evolved over 2010-2022, so a time trend in measured AI adoption could in principle partly reflect posting digitization. The window also predates widespread generative-AI diffusion, so 'AI-enabled environments' rests on a pre-ChatGPT notion of adoption. These are the critic's external-validity observations, not stated limitations.
c9. The paper positions itself as examining how AI adoption relates to managerial demand, explicitly characterizing its findings as relationships rather than causal effects.
The associational framing is consistent and disciplined, and the paper should not be faulted for lacking a causal identification strategy it never claims. The residual risk the critic notes is reader slippage: framing associational skill shifts as role 'reconfiguration' and (in the managerial summary) a 'growing emphasis' could invite a causal or dynamic reading the cross-firm evidence does not support, so the prose should stay tied to co-movement language. This is a caution about reception, not a charge that the abstract overclaims.
c10. The growth-oriented interpretation presumes the upward-shifting skill categories genuinely capture growth/interpersonal capabilities and the declining categories are genuinely routine, resting on the chosen skill-category labeling.
The theoretical claim is, on the critic's reading, only as strong as the taxonomy underneath it. The labels are face-valid and conventional, but the abstract reports no external benchmark, mutual-exclusivity check, or inter-rater validation, and reasonable readers could reclassify edge cases (planning has strategic content; sales management is a management function). Because the conclusion both presumes and asserts the growth/routine distinction, the critic identifies a circularity the abstract does not dispel. This is the critic's construct-validity inference.
Scorecard
Sub-scores are 0–5 editorial judgements on fixed scales (higher is better, except methodological risk and overclaiming where higher is worse). They are contestable and open to a severity challenge from authors.
What the paper claims
The paper offers a large-scale, explicitly associational mapping of how firms' AI adoption co-moves with managerial demand. Using "a skills-based measure of AI adoption derived from Lightcast job postings," the authors "show that firms with greater AI adoption post more managerial vacancies and a higher share of such vacancies than less intensive adopters." These patterns "are strongest in manufacturing and among firms with higher research & development intensity," and greater adoption is "associated with shifts in managerial skill requirements toward interpersonal and growth-oriented skills... and away from routine administrative skills." The synthesis is hedged: "the results suggest a reconfiguration of managerial roles toward capabilities facilitating scaling, coordination, and adaptation in AI-enabled environments." The framing is consistently relational, and the contribution — pointing AI-and-labor attention at managerial demand and managerial skill composition rather than the usual routine-clerical substitution story — is genuine and well-bounded.
The headline concern: same-source measurement (critic's inference)
The most substantive fair challenge is not a missing causal design — the paper never claims causation — but, on the critic's reading, the shared measurement substrate. The abstract states that AI adoption is "derived from Lightcast job postings" and that the outcome is "managerial vacancies" the firms post; the critic infers that both the AI-adoption regressor and the managerial-vacancy outcome are therefore built from the same Lightcast postings for the same firms. If that is so, a firm that simply posts more vacancies overall — because it is hiring broadly, scaling, or better covered by Lightcast's crawl — could mechanically surface more AI skills and post more managerial roles, generating the reported correlation absent any AI-to-management link. The abstract gives no indication of posting-volume normalization, firm-size conditioning, or a non-postings adoption measure that would break this dependence. The "higher share of such vacancies" outcome is a meaningful partial mitigation of the level-count artifact, but its denominator is still the firm's own postings drawn from the same corpus, so on the critic's reading it does not fully separate an AI-specific signal from general recruiting intensity. This is an inference about the measurement design, not a limitation the abstract states.
Construct validity of the proxies
Two proxies do heavy lifting. First, 'AI adoption' is, per the abstract, a "skills-based measure" — the presence of AI-related skills in postings, which the critic reads as advertised intent to build AI capacity rather than observed deployment, spending, or use; firms can advertise aspirationally, while firms buying off-the-shelf tools may adopt without posting AI-skill requirements. Second, 'managerial demand' is proxied by managerial vacancy postings, which capture advertised hiring rather than realized fills, and can reflect churn, replacement, or reposting of hard-to-fill roles. The abstract links neither proxy to a validation benchmark. The authors' honest labeling of the measure as "skills-based" is to their credit, but, on the critic's reading, the gap between advertised demand and realized adoption/hiring remains unaddressed.
Labels carry the theoretical weight
The interpretive story rests on a taxonomy. The abstract itself characterizes "stakeholder management, creativity, and sales management" as "interpersonal and growth-oriented" while characterizing "budgeting, planning, staff management, and customer service" as "routine administrative" — a mapping the abstract does not validate externally. On the critic's reading several assignments are contestable: planning and staff management are arguably core coordination and leadership functions, and 'sales management' is itself a management function placed on the rising side. The within-role directional shift is a real strength — a pure volume confound would not obviously predict a specific reallocation among managerial skills — but the conclusion of "reconfiguration... toward scaling, coordination, and adaptation" both presumes and asserts the growth/routine distinction the labels encode, raising what the critic reads as a circularity the abstract does not dispel.
Heterogeneity and external validity
The moderators are coherent and add structure: manufacturing and R&D-intensive settings are where AI integration into operations and innovation pipelines would most plausibly reshape coordination needs. Two cautions follow, both flagged as the critic's inferences. The abstract's managerial summary treats "higher research & development intensity" as "more innovative firms," which substitutes an input proxy for the construct; and both subgroup associations would, on the same-source reading, inherit the shared-substrate concern, since such firms may post and skill-tag differently. On external validity, the scope is transparently bounded to "US job postings data from 2010 to 2022" and to whatever Lightcast captures. These are responsible scope conditions rather than flaws, but the abstract omits coverage caveats the critic considers material — plausibly uneven representation across firm size and sector, and a rising propensity to post online over the window that could confound a measured AI-adoption trend with posting digitization in a pre-generative-AI era. These caveats are not stated in the abstract.
Assessment
Judged on its own associational terms, this is a disciplined descriptive contribution whose central reported relationships stand as stated, subject to clearly nameable scope conditions. The dominant threat, on the critic's reading, is measurement, not identification: if both measures share the Lightcast substrate (as the abstract's language suggests), a posting-intensity artifact cannot be ruled out from the abstract, and the interpretive payoff depends on a skill taxonomy the abstract does not validate. None of the strengthening fixes — showing the association net of total posting volume and firm size, validating the AI-skills and skill-category labels against an external benchmark, and reporting whether the share result survives excluding AI-related postings from the denominator — requires a causal design. They are about defending the measurement and the robustness of the correlation the paper actually claims.
Strongest critique
On the critic's reading of the abstract's wording, both the AI-adoption measure ("derived from Lightcast job postings") and the managerial-vacancy outcome ("managerial vacancies" the firm posts) appear to come from the same Lightcast job-postings corpus for the same firms, so any firm-level driver of overall posting intensity — broad hiring, growth, or better Lightcast coverage — could load on both sides and partly generate the headline association between AI adoption and managerial demand without a substantive AI-to-management link. The abstract reports no posting-volume normalization, firm-size conditioning, or non-postings adoption measure that would break this apparent shared-substrate dependence, and the 'higher share of such vacancies' outcome only partially addresses it because its denominator is still the firm's own postings. This is an inference about the measurement, not a flaw the abstract concedes.
Strongest fair defence
The paper never claims causation: it "examines how firms' adoption of artificial intelligence (AI) relates to the demand for managers," labels its findings "These relationships," and hedges that "the results suggest a reconfiguration." On those terms it delivers a consistently measured association across a large corpus, and two features raise the bar on a pure measurement artifact: the headline includes a composition (share) outcome, not just raw counts, and the skill findings are specific and directional — a volume confound would not obviously predict a reallocation within managerial skills away from administrative and toward named growth categories. The honestly labeled skills-based proxy spans 2010-2022 at a breadth no survey or spending dataset matches, and the scope is transparently bounded, making the critic's concessions (same-source measurement, proxy-not-deployment, label dependence, bounded external validity) genuine scope conditions rather than refutations.
Conclusion
A disciplined, transparently associational and well-bounded descriptive contribution that opens a useful and underexplored angle — AI adoption's link to managerial demand and managerial skill composition — and earns credit for hedged 'relates to / associated with / relationships / results suggest' language it never sharpens into causation. On the critic's reading, its central reported relationship and its 'reconfiguration' interpretation nonetheless appear to rest on a shared Lightcast measurement substrate (a posting-intensity confound the abstract does not rule out, inferred from the abstract's own description of both measures), on proxies that capture advertised intent rather than realized adoption or hiring, and on a skill-category taxonomy the abstract does not externally validate yet which carries the theoretical weight. These are measurement, construct-validity, and interpretation concerns rather than fatal flaws, and all are addressable within an associational design; severity: moderate.
Reply from the authors
Following the practice of Nature Matters Arising, Science Technical Comments and PNAS Letters, this Comment is published as one half of a Comment + Reply pair: the authors of the original article are invited to respond, and any reply is published here verbatim alongside the Comment as part of the record.
Reply: not yet invited. No reply has been received for publication.
The authors have a right of reply and no veto. A reply may request a factual correction, a methodological rebuttal, a clarification, a data/code update, or a severity challenge, and is published unedited. See the right-of-reply policy.
References
Every external source this Comment cites, each with a verified link. 0 fabricated.
Source-grounding attestation
- ✓Verbatim source spans present in the critique — 8/8 provenance spans re-derived in the critique prose
- ✓Passes the publication validator — no errors
- ✓Zero fabricated citations — 0 fabricated
- ✓Severity within the access-basis cap — severity "moderate" ≤ cap "moderate" for abstract_only
Every verbatim span the critique relies on is re-derived in the prose in-app; span-in-source is re-verifiable offline (the abstract is re-fetched, not stored, per the no-reproduce policy).
Re-verify span-in-source offline: python3 scripts/verify-queue-critiques.py
Independent faithfulness review
A refute-by-default adversarial panel (two independent reviewers — an overreach lens and a mischaracterization lens — that fetched the real source) tried to prove this critique misread the paper. This is an AI adversarial review recorded with its reasoning, not a deterministic check.
Both refuters returned "faithful" with nothing sustained, and that adjudication holds. I spot-checked every contestable move against the embedded abstract. The critique's factual attributions are accurate: the rising-skill list (stakeholder management, creativity, sales management) and the falling-skill list (budgeting, planning, staff management, customer service) are quoted correctly; the "routine administrative" and "interpersonal and growth-oriented" labels are genuinely the abstract's own; the R&D-intensity / "more innovative firms" substitution really is the abstract's own slide between its two summaries; the hedging vocabulary ("associated with," "the results suggest," "relationships") is present as quoted; and the 2010-2022 US scope is stated. The critique's two sharpest interpretive moves — the shared-Lightcast-substrate confound and the "circularity" of the skill taxonomy — are exactly the kind of claim that would be unfaithful if asserted as established fact, but the author consistently brackets them as the critic's inference and never attributes them to the abstract. That discipline is what separates a contested over-reach from a faithful labelled inference here. No claim misrepresents the source; refutersSustaining = 0.
Version & correction history
| Version | Date | Change |
|---|---|---|
| v1.0 | 2026-06-20 |
No silent substantive corrections — every change is versioned and visible.
How to cite this Comment
Critical AI. Comment on “Artificial intelligence adoption and the demand for managerial expertise” (Liudmila Alekseeva et al., Strategic Management Journal, 2026). Critical AI; 2026. https://policywindow.org/critique/c/scp-artificial-intelligence-scp-adoption-and-the-d
A registered DOI will replace the URL once minted; until then the canonical URL is the persistent identifier. Highwire/Dublin-Core citation tags and a schema.org Review record are embedded in this page for Google Scholar and reference managers.
Verify this Comment. Its checkable facts (target DOI, access-basis severity cap, zero fabricated citations) are served — as the app’s self-report — at /critique/api/critiques/scp-artificial-intelligence-scp-adoption-and-the-d/verify; to confirm them independently of this site, re-derive the same checks (and resolve the target DOI) with npx tsx scripts/verify-critical-ai.ts --critique scp-artificial-intelligence-scp-adoption-and-the-d --live.
Content fingerprint 4218e75f23011667 (v1.0) — this Comment’s substantive content is content-addressed; a silent post-publication edit would change it.