All issues

Intelligence Brief

Week of
September 21, 2026

Your weekly edge, the research, frameworks, and signal that separate operators from spectators. Read and summarized so you do not have to.

Published Monday, September 21 at 6:00 AM PT. 14 sources, 6 minute read.

01

The Human Intelligence Edge

What makes humans irreplaceable in the AI era.

WSJ IntelligenceSeptember 10, 2026

New WSJ Intelligence Survey Highlights the Human Skills Companies Must Protect as AI Enters the Workplace

WSJ Intelligence for Philip Morris International, via Business Wire

Why it matters. A survey of more than 2,500 business professionals across five countries puts critical thinking (46%), moral judgment and ethics (42%), and empathy and trust-building (40%) at the top of the capabilities AI is least likely to replicate. The number underneath is the one operators should sit with: only 25% strongly agree their organization has a clear process for verifying AI output. The skill everyone says they value is the one nobody has built a check around.

Boston Consulting GroupSeptember 10, 2026

AGI Timeline 2026: What CEOs Need to Know

Greg Emerson and Ulrich Pidun

Why it matters. BCG's jagged frontier framing cuts against AI hype and human exceptionalism at the same time: capability is uneven across tasks, so the advantage comes from redesigning roles around the places people still add judgment and context. Applied to hiring, that means naming the specific decisions a role owns above the model's frontier before writing the job description, then testing for those.

arXiv preprintSeptember 2, 2026

AI, Practice Style, and Screening in Elite Skill Formation

Song Yao

Why it matters. The study links 10,419 competitive-programming users to their results in AI-prohibited ICPC and IOI contests and separates AI used as a substitute from AI used to support deliberate practice. Frontier skill survives AI when the practice stays effortful, but proctored, AI-free gates are what keep independent capability legible. That is the case for an evaluation a candidate cannot outsource.

02

The Broken System

Why resumes, interviews, and hiring infrastructure are failing.

arXiv preprintSeptember 17, 2026

When Hiring Becomes Agent-Mediated: Evaluating Access and Recurrence in Two-Agent Resume Screening

Jian Gao and Hang Jiang

Why it matters. Across 600 constructed resume and job pairs, two-agent screening advanced different candidates than one-call screening and moved both access rates and recurrence. Who reaches a human now depends on how evidence gets exchanged and reconsidered, not only on which model sits underneath. The procedure is part of the outcome, which makes an unexamined screening pipeline a hiring decision nobody made.

SHRMSeptember 15, 2026

Fraudulent Candidates at Scale: Are AI-Generated Applications Breaking Recruitment?

SHRM Advisor, citing Gartner survey data

Why it matters. A Gartner survey of 3,290 candidates found 39% used AI during the application process, and a separate survey of 3,000 candidates found 6% admitted to interview fraud. When the resume, the writing sample, and sometimes the identity can be generated, polished artifacts are weak evidence and ATS filters are sorting on the wrong signal. Proof has to come from observed work.

arXiv preprintAugust 27, 2026

Counterfactual Bias Testing for Application Tracking System

Sai Yashwant, Shruti Bansal, Anurag Dubey, et al.

Why it matters. The paper runs a concrete audit design over an applicant tracking system: 100 base candidates, five job orders, ten controlled demographic treatments, and nine fairness metrics. Rank stability came out borderline even where the other metrics stayed inside tolerance. That is the argument for multi-metric counterfactual testing instead of the single fairness score on a vendor's one-pager.

03

The SupaHuman Standard

The behavioral science of who actually thrives.

World Economic ForumSeptember 16, 2026

What skills do employers want from young people in the AI age?

Charlotte Edmond, citing WEF and PwC analysis

Why it matters. AI-exposed junior roles are seven times more likely to demand skills previously associated with senior positions, while more than a quarter of entry-level workers expect at most half their current skills to remain relevant in three years. Learning agility, adaptability, resilience, communication, and emotional intelligence are the load-bearing part of a junior hire now, not a soft extra bolted onto the technical screen.

ForbesAugust 29, 2026

What Is Emotional Intelligence? The New Definition For The AI Era

Kevin Kruse, citing Travis Bradberry's assessment data and PwC

Why it matters. More than 500,000 emotional-intelligence assessments put EQ at 58% of performance across job types, with 90% of top performers scoring high in it. As technical execution gets cheaper, self-awareness, self-management, empathy, and relationship skill are the predictors left standing, and all four are observable in how someone handles a real problem with a real team rather than in how they describe themselves.

ForbesSeptember 8, 2026

Why Knowing When To Use AI Is The Next Leadership Challenge

Michael Edmondson, citing a 14,949-student multinational study

Why it matters. The cited study identifies Augmentors, the people who pair heavy AI use with strong cognitive skills, and finds self-testing the strongest predictor of cognitive-skill development while frequency of AI use predicts efficiency. The hiring question falls out of it: does this person keep productive friction in the loop, and do they know when to challenge, verify, or decline what the model produced?

DeloitteJanuary 14, 2026

Human Skills Drive High-Performing Teams in the AI Era

Deloitte US press room

Why it matters. Deloitte's survey of 1,394 workers ties high-performing teams to curiosity, informed agility, resilience, connected teaming, and social and emotional intelligence alongside heavier AI use. Exceptional performance reads as a behavioral pattern of learning, adapting, collaborating, and recovering. A credential cannot certify that pattern. A week of real work can show it.

04

The Future of Work, Without the Clichés

Where AI and human collaboration is actually heading.

arXiv preprintSeptember 15, 2026

Available but Unclaimed: An Empirical Study of Human-AI Synergy

Robin Welsch, Michelle Rausch, Pascal Knierim, et al.

Why it matters. In a 535-person study across 40 items, AI assistance raised accuracy over unaided participants, yet realized synergy was negative on average because people deferred too readily to wrong advice. The gain was available and went unclaimed. Two practical consequences: measure the team against both human-only and model-only baselines, and hire for selective deference rather than enthusiasm.

World Economic ForumSeptember 15, 2026

Why the next AI data race will be for human expertise and knowledge

Reese Wong

Why it matters. A 2026 Bridgewater and Thinking Machines Lab test used expert feedback on six information-filtering tasks to fine-tune a specialized model that beat the alternatives at lower cost. Expert judgment, corrections, edge cases, and institutional feedback are becoming the scarce input that makes a model useful, which raises the market value of people who can explain precisely why a call was wrong.

arXiv preprintAugust 26, 2026

Human learning is an understudied but promising lever for boosting human-AI synergy

Julian Berger, Jason W. Burton, Ralph Hertwig, et al.

Why it matters. A re-analysis of 74 studies and 370 effect sizes found synergy negative without feedback and positive in direction once feedback was present, with explanations helping when paired with feedback and hurting when they were not. The lever is teaching people to calibrate the model against outcomes, not decorating its output with reasons.

MIT SloanAugust 31, 2026

What personality pairings most improve human-AI collaboration?

PNAS study by Sinan Aral and Harang Ju

Why it matters. A randomized experiment with more than 1,200 participants, roughly 1,100 additional ad raters, and about five million real-world impressions found that human and AI personality pairings changed ad quality and ad performance. AI adoption is too blunt a unit of analysis. Fit, task, interface, and interaction design decide whether the collaboration compounds or cancels out.