AI Task Crossover and the Rise of Skill Based Organisation

Key takeaways

1. Date / source. August 1, 2026, SFEIR. Framing article built on two third-party publications — the P&G experiment and the OpenAI Economic Research report.
2. Key framing. AI doesn't just speed up work, it redistributes who does what. The blurring of professional boundaries is a measurable organizational fact; the conclusion drawn is that the job description can no longer be the unit of work organization. ### The robustness argument Two opposite methods converge: a preregistered field experiment (causal, small N, controlled context) and massive usage trace data (scale, correlational, no context). Neither would suffice alone; together they cover each other's blind spots. It's this combination that matters, more than any single isolated figure. ### P&G — the mechanism 791 experienced professionals (R&D and sales), a full day on real product-innovation challenges drawn from their business units, with the best proposals presented to executives. 2×2 protocol: solo / cross-functional pair × with / without AI. | Result | Value | |---|---| | Individual + AI vs. individuals alone | +0.37 σ | | Team without AI vs. individuals alone | +0.24 σ | | Team + AI, probability of a top-10% solution | roughly ×3 | | Correct selection of one's best idea | ~50% for a human pair without AI, 37% with AI | The effect plays out in the tail of the distribution, not in the average. Without AI, sales people produce sales-oriented output and R&D people produce technical output; with AI, the distinction disappears — solutions balanced across the entire technical-commercial spectrum, without any loss of quality. The authors call the mechanism boundary-spanning: employees far removed from product development, alone with AI, reach the level of teams that include a domain expert. Emotional dimension measured: more enthusiasm and energy, less anxiety and frustration. The nuance that holds up: AI raises the average quality of ideas, but human pairs without AI remain better at identifying their own best idea. Human evaluative judgment retains its place — the same generate/verify asymmetry as in [[sfeir-code-review-anneau-contraintes-2026-07-30]], a different domain. ### OpenAI — the scale 43.5% of profession-specific messages fall outside the user's own profession, once generic tasks are set aside. The reprocessing matters: this is not 43.5% of all usage, but 43.5% of profession-specific usage. | Function | Share out-of-profession | Import | Export | |---|---|---|---| | Customer experience | 77% | — | — | | Design | 75% | 35.2% | 1.7% | | HR | 69% | — | — | | Legal | 56% | — | — | | Marketing | 53% | 24.3% | 8.9% | | Sales / finance | 40% | — | — | | Engineering | 28% | 18.5% | 7.4% | Two tasks appear in the top 3 borrowed tasks across all groups: financial calculation and technology troubleshooting. Claimed status: an early signal, making the recomposition visible before job descriptions and labor-market statistics do. The result the article makes least use of: engineering is the least porous and most export-heavy profession. The porosity described is thus largely one-way, toward the technical side. For a firm whose core business is engineering, this is the most actionable result, and the article draws nothing from it. ### The weak link in the argument The organization-size effect rests on a 2.6-point gap (18.9% for 2-5-role spaces → 16.3% above 100), figures not commensurate with the 43.5%. Verification against the primary source [[openai-work-at-the-frontier-task-crossover-2026-07-27]] shows the gap holds only "among average users," and "among the heaviest users, we do not see the same monotonic pattern." OpenAI's conclusion is conditional ("AI may be especially useful as a generalist tool"), where SFEIR states it as fact. Not to be cited as a strong result. ### Mollick's three-step agenda Co-author of the P&G study, he links the two publications himself, noting that the cross-functional usage measured by OpenAI is more pronounced than the experiment had suggested. (1) Organizational boundaries are becoming porous. (2) Companies will have to rethink the division of labor, and "things are getting messy right now." (3) Properly orchestrated, the recomposition pays off both in employee satisfaction and in performance. ### SFEIR's response: the Skill Based Organisation Under the impetus of Rosalie Zandona (VP People & Culture). If tasks circulate, the fixed job description can no longer be the unit of organization: actually operational skill becomes the base building block — up to 13 skills identified per role — deployed wherever the concrete need is, independent of job title. The underlying shift goes from a status-based identity ("I am a manager") to an operational identity ("I know how to design complex architectures"), made objective and transparent. Claimed benefit: task crossover becomes visible and tooled instead of remaining "informal tinkering in the shadow of the org chart." Closing figure: according to Deloitte, skill-based organizations are 98% more likely to be perceived as excellent places to grow. ### Three caveats to raise in pre-sales conversations 1. Reversed chronology — SFEIR's SBO shift is documented in Zandona's article of February 19, 2026, five months before the two publications it is supposed to "answer point by point." The SBO is reread in their light, not adopted in response to them. 2. The logical leap is not addressed — both studies measure task spillover; the SBO responds with a change in the unit of organization. Nothing in the data says that a skill-based organization absorbs crossover better than a role-based one. It's a design hypothesis, and not the only possible one: the other common response is the merged role (SFEIR's own "Product Engineer," April 2026), which goes in the opposite direction. 3. The Deloitte figure measures perception, not performance. It covers only the satisfaction side of the dual benefit Mollick announced; the performance side of the SBO is quantified nowhere. ### Watch points on the sources
3. P&G freshness. the result has been circulating since March 2025 (NBER w33641 / HBS WP). The article dates it to its publication in Organization Science (2026): true for the peer-reviewed journal, not for the result itself. For a watch note, this counts as a consolidation.
4. Diverging N. 791 professionals stated by the article, 776 in the 2025 working paper. Unexplained discrepancy — to be checked against the Organization Science version.
5. Title variant. "… on Generative AI and Teamwork" (SFEIR) vs. "… on Generative AI Reshaping Teamwork and Expertise" (working paper).
6. OpenAI base. US ChatGPT users, neither a representative sample of the active population nor multi-assistant. And it is OpenAI measuring OpenAI's own effect — a bias SFEIR flags elsewhere, but not noted here.
7. Missing figure. the report gives two headline figures — 16.8% of work-related messages and 43.5% of profession-specific messages, the latter calculated after removing 61.5% of generic usage. SFEIR keeps only the second one. Disambiguation: "skill" in the SBO sense (an HR unit of organization) has no relation to the Agent Skills of the corpus — an FR/EN homonym that must not be allowed to merge in the graph.