Skill Partnerships Human-AI

McKinsey’s framing for AI’s impact on work: not replacement of humans but a new “skill partnership” where agents and robots handle automatable tasks while humans apply their skills in reconfigured ways — a distinction that critically depends on how organisations design the collaboration.

What It Is

McKinsey Global Institute (November 2025) presents the “Agents, Robots, and Us” framework, which treats AI adoption as a partnership redesign problem rather than a replacement problem. The central insight: more than 70% of the skills sought by employers today can be applied in both automatable and non-automatable work. This means most skills don’t become worthless when AI enters — they get redirected. Workers shift from routine execution of tasks to framing, interpretation, quality assurance, and oversight of AI output.

McKinsey developed a Skill Change Index (SCI) — a time-weighted measure of automation’s potential impact on each skill — to project which skills face the greatest shifts by 2030. Digital and information-processing skills (accounting, coding, data entry) face the most disruption; interpersonal skills (negotiation, coaching, care) change the least; widely applicable skills (problem-solving, communication) evolve as part of the human-AI partnership.

Why It Matters (for Organizations)

The “skill partnership” framing has direct practical implications for org design. Organisations that treat AI as a labour cost reduction tool — replacing workers without redesigning workflows — will likely capture only a fraction of the available value. McKinsey estimates that $2.9 trillion in US economic value could be unlocked annually by 2030, but this figure is contingent on workflow redesign (redesigning entire processes around humans and AI working together), not just task automation (plugging AI into existing processes).

For leaders, the implication is that AI adoption is not primarily a technology implementation challenge but an organisational design challenge: restructuring roles, skills, incentives, culture, and performance metrics around human-AI collaboration.

Evidence & Examples

  • Currently demonstrated technologies could theoretically automate activities accounting for ~57% of US work hours — but this is technical potential, not a job loss forecast (agents-robots-and-us-skill-partnerships-in-the-age-of-ai.pdf)
  • Demand for “AI fluency” in US job postings has grown ~7x in two years — faster than any other skill; approximately 8 million Americans currently work in jobs already calling for at least one AI-related skill (agents-robots-and-us-skill-partnerships-in-the-age-of-ai.pdf)
  • Skill Change Index highlights: Accounting and coding face the greatest automation-driven disruption; negotiation and nursing care face the least; quality assurance, process optimisation, and teaching are growing in demand as AI creates new complementary needs (agents-robots-and-us-skill-partnerships-in-the-age-of-ai.pdf)
  • Job posting mentions are declining for “routine writing and research” — areas where AI already performs well — even though these skills remain essential for much of the workforce (agents-robots-and-us-skill-partnerships-in-the-age-of-ai.pdf)
  • Midpoint adoption scenario by 2030: ~$2.9T in annual US economic value unlocked — contingent on workflow redesign, not just task automation (agents-robots-and-us-skill-partnerships-in-the-age-of-ai.pdf)

Tensions & Open Questions

  • The redesign assumption is optimistic. McKinsey’s $2.9T figure assumes organisations successfully redesign workflows around human-AI collaboration. In practice, change management for AI adoption is among the most cited barriers in enterprise deployments. The “Measuring Agents in Production” study (UC Berkeley, 2025) found production AI agents are typically built using “simple, controllable approaches” — far from the sophisticated workflow redesign McKinsey envisions (see AI Agents in Production).
  • Who bears the cost of transition? Redirecting skills requires retraining. McKinsey is relatively silent on who funds this retraining — firms, governments, or workers themselves. In practice, low-skill workers exposed to AI-automatable Gateway occupations (see AI Career Pathways and Workforce Mobility) may lack access to the retraining that the skill-partnership model assumes.
  • Is “AI fluency” a durable skill or a short-term transitional skill? The 7x growth in AI-fluency job postings may reflect a surge during the implementation phase, after which AI tools become invisible infrastructure that no longer requires explicit skills to use.
  • Interpersonal skills as the new moat: If negotiation, coaching, and care are least disrupted by AI, organisations and individuals that have invested in these capabilities may have durable advantages — but these skills are rarely what corporate training programs emphasise.

Workflow Redesign Around AI · AI Labor Displacement and Augmentation · AI Skill Formation and Deskilling · AI Agents in Production