New report reveals hiring shifts as AI begins to handle most workplace skills
As artificial intelligence becomes capable of handling nearly three-quarters of workplace skills, a new report is urging employers to rethink what they are actually testing for when they hire.
"The question hiring teams should ask isn't whether AI can do a task; for most tasks now, some version of it can," said Guruprakash Sivabalan, founder and CEO of AI recruitment platform Xobin.
"It's which parts a candidate must do independently, and which they need to direct and check."
The comment accompanies Xobin's Human vs AI Skills Report: 2026 Mid-Year Edition, published this week, which found that 72% of skill groups assessed in its 2024 hiring framework were fully or partially delegable to AI under at least one tested model.
The report, which draws on Xobin's own internal records, covered 683 skill groups, 113 leadership scorecards from 92 employers, and thousands of technical assessment requests through 30 June 2026.
The proportion of skills delegable to AI varied sharply by category, from 90 to 100% for analytical reasoning and problem-solving, down to 28% for operations and execution, with leadership and interpersonal skills below 50%.
The report notes that partial delegation often still requires a person to frame the task, review the output, and handle exceptions, and that the study does not measure jobs eliminated.
How hiring is shifting
Hiring priorities appear to be adjusting accordingly. Among 35 technical roles studied, the ability to turn business requirements into technical solutions, as listed in job descriptions and assessment requests.
It ranked among the five most frequently listed skills, appearing in 10 to 19 of those roles. It trailed AI integration and automation, and analytical problem-solving, which each appeared in 30 to 35 roles.
Leadership hiring is shifting too. Across 113 scorecard templates from 92 employers, emotional intelligence criteria averaged 52% of total scorecard weight, edging out all other criteria combined.
The change is perhaps most visible in how employers are assessing technical candidates, according to the findings.
Traditional, AI-free coding tests made up 75 to 100% of technical assessment requests in the first half of 2024.
By the first half of 2026, that share had fallen to between 25 and 50%, while AI-assisted tasks such as generating, debugging, and explaining code rose from under 25 per cent to between 50 and 75%.
The report cautions that these are shares of requests, not total counts, and that the employer mix across the two periods may differ.
Sivabalan said the ability to catch AI errors depends on candidates retaining core knowledge even as AI takes on more routine work.
"Foundational knowledge hasn't gone away; it's what lets someone catch an AI's mistake," he said. "Separate what a candidate must do independently from what they need to direct and verify with AI, and build that into how roles are assessed."
The 'bad AI hire' problem
The findings come amid broader concerns about AI hiring. More than half of organisations have made a "bad AI hire" in the past year, according to TestGorilla.
This refers to candidates who speak fluently about AI tools during recruitment but cannot apply that knowledge on the job.
TestGorilla CEO Wouter Durville said the issue stems not from prioritising AI fluency over domain expertise, but from failing to test for both together.
"The right framing isn't AI skills vs. domain skills. It's AI skills applied to domain skills," Durville said. "Hire for the combination. The 'bad AI hire' problem is what happens when you optimise for one without testing for both."
