AI-Wiki

To Thrive Alongside AI, Focus on Mindset—Not Skillset

5 min read

To Thrive Alongside AI, Focus on Mindset—Not Skillset

A short, high-altitude Harvard Business Review essay (12 June 2026) by Marco Argenti, Chief Information Officer at Goldman Sachs. Its argument is a deliberate inversion of the wiki’s durable-skills framing: don’t ask “what 10% of my job can AI never do?” — that question itself is the trap. Reimagine the whole role. It is the wiki’s first first-party Goldman-Sachs-executive source and a clean practitioner companion to the GDPval paper it cites.

TL;DR

  • The wrong question. A senior banker asks Argenti: “what’s the 10% of my job that I should focus on, that the AI will never be able to do?” His surprising answer: let go of that 10%. The horse-rider→car-driver analogy: none of the horse-riding skills transfer, but the reflexes and instincts do. Hang on to the human qualities (instincts, judgement, values), not the specific skills. “Don’t just reskill, reimagine skills and build new habits.”
  • Cites GDPval as the stakes. Argenti opens with OpenAI’s GDPval — “44 professions and 1,320 tasks across the top 9 industries contributing to U.S. GDP” — and reports agents performing as well as or better than humans in ~80% of cases, up from ~50% six months ago, “destined to go up.” (He extrapolates GDPval’s linear-improvement trend forward; the paper’s own Oct-2025 figure was 47.6% wins-or-ties for the best model — see the GDPval page for the trajectory caveat.) Names developers, lawyers, property managers as most-exposed.
  • Operator → supervisor → mentor (“the new 100%”). The mindset shift is learning when to relinquish control — from producing every line of a pitch deck to giving clear instructions and ensuring controls so agents execute safely on your behalf. Banking framing: agents working overnight let you answer client questions at “t-1” — before the client even asks; the banker’s value-add becomes review, judgement, and the human call.
  • Three ingredients for the rewire (rewire the company around a human+agent hybrid workforce as the new normal):
    1. Leadership — letting go doesn’t just happen; needs top-down accountability. Demand radical targets, not incremental ones: “ask developers to be 3× more productive, not 20%”; aim for “90% reduction of manual touchpoints, not 20%.” Interview tasks so hard they require AI mastery (“create a working clone of Excel in 3 hours”). Halfway to a radical target still proves real rethinking happened.
    2. Clarity of objectives and outcomes — “obsess over evaluations and benchmarks.” Real org processes resemble the Garbage Can model (chaotic, nonlinear), not clean SOPs. At Goldman, applying AI to client onboarding meant first codifying what good looks like (quality metrics + experienced-operator decisions), then building evals comparing agent outputs to desired outcomes, with feedback loops so the AI self-improves. Outcome-based, not step-by-step (tell Maps the destination, not every turn).
    3. Mastery of your own data — “AI transformation follows data transformation, not the other way around.” Agents without context revert to chatbots; data is the ground truth. AI has the ultimate garbage-in/garbage-out problem because it makes garbage output look plausible. Leaders should be willing to delay AI-at-scale until data is in order (months/years); data readiness is itself a useful signal for which use cases to prioritize.
  • The habit change. Resist taking AI output at face value — check sources, supervise, verify. “An agentic future requires everyone to turn into a manager of sorts.” The hardest change is personal: the courage to let trusted habits die and adopt a new professional identity.

Why this matters to the wiki

Argenti is the mindset-side counterweight to the wiki’s skills-inventory sources. Where durable-skills (Globerson et al.) and McKinsey enumerate the human skills that survive, Argenti argues the framing of preservation itself is the error — the goal is metamorphosis, not retention. His “3× not 20%” maxim is the sharpest single line in the wiki against the micro-productivity-trap (optimization-as-relief that “massively misses the mark”). The data-before-AI claim and the obsess-over-evals discipline tie him to enterprise-ai-adoption and the agent-harness eval/contract layer. The operator→supervisor shift is the same role transition WP reports across the labor market.

Dynamic-capabilities (W&W) reading

  • digital-transforming/redesigning-internal-structures + improving-digital-maturity — the core prescription is rewiring the company around a human+agent hybrid workforce, redefining roles (operator→supervisor) and processes (step-by-step→outcome-based).
  • strategic-renewal/organizational-culture — the load-bearing claim is a cultural metamorphosis: letting old habits die, reimagining professional identity, leadership holding people accountable for change.
  • digital-seizing/strategic-agility — “reimagine, don’t just reskill”; demanding radical (3×) rather than incremental targets to force genuine rethinking.
  • contextual/internal-barriers — names the human barrier directly: the fear of losing hard-won expertise, the urge to stay in control, and stale/duplicative data as the blocker that should delay AI projects.

Linked entities and concepts

Relationships

Relationships

Supports

Uses

Graph View

Backlinks