AI Can Perform 72% of Job Skills. What Should Hiring Teams Actually Hire For?

When candidates can use AI to help produce their work, what should hiring teams actually assess? That is the question Xobin Research's 2026 Mid-Year Edition of the Human vs AI Skills Report sets out to answer, bringing together five findings from the company's first-half 2026 hiring and AI-benchmark records, with matched 2024 comparisons where available.

The report is built on Xobin's own assessment framework and hiring-related records rather than a survey of the broader market.

72% of skill groups can be delegated to AI, under test conditions

Xobin tested all 683 skill groups classified as relevant in its 2024 assessment framework against outputs from OpenAI and Anthropic tool and agent workflows. 72% of those groups met the criteria for full or partial delegation to AI, with at least one tested setup qualifying for each group.

The report is explicit that this describes capability under test conditions, not a claim that these skills stop mattering or that any work was actually replaced in the field. The distinction matters: it tells hiring teams where AI assistance can plausibly show up in candidate work, not where humans have become unnecessary.

Leadership hiring is betting on human judgment

Across 113 distinct leadership scorecard templates from 92 employers, EQ-related criteria averaged 52% of total evaluation weight, with all other criteria combined averaging 48%. The report reads this as a signal that people-facing judgment carries real weight in how organizations plan to evaluate leadership hires, without claiming EQ predicts performance better than any single other capability.

Reviewing 35 distinct technical roles from January to June 2026, the report found that translating business requirements into technical solutions ranked among the five most frequently requested skills, alongside AI integration and analytical problem-solving.

Technical assessments have quietly flipped

Comparing the first half of 2024 with the first half of 2026, traditional coding assessments fell from 75 to 100% of technical assessment requests down to 25% to under 50%, while AI-assisted coding tasks, generation, testing, debugging and explanation, rose from under 25% to 50% to under 75% of requests over the same window.

Within a fixed set of 24 emerging skill groups tracked across both periods, collaboration appeared in 4 groups in 2024 and 12 to 17 in 2026, while non-linear thinking appeared in 0 to 5 groups in 2024 and 6 to 11 in 2026.

How to read these numbers

The report is careful throughout to separate its measured findings from editorial interpretation, and notes that each finding counts a different unit of analysis, skill groups, scorecard templates, job roles, and assessment requests, so the figures cannot be combined into a single sample size.

The full report includes chart-level data tables, a complete methodology section, and citations to external context from Anthropic, Microsoft, and the OECD alongside Xobin's own findings.

Read the full Human vs AI Skills Report 2026 →