AI Is Making Execution Cheap. What Becomes Scarce?

When the cost of doing falls close to zero, the value of deciding, directing, and contextualising rises sharply. This is not a prediction about the distant future — it is already the operative condition of the 2026 labour market.

The cost of execution is collapsing

In November 2022, the public release of ChatGPT marked the point at which AI crossed a practical threshold: it became cheap enough, capable enough, and accessible enough to perform meaningful knowledge work at a cost competitive with human labour. Three years later, the numbers bear this out with unusual clarity.

An MIT study released in late 2025 found that current AI systems could already take over tasks tied to 11.7% of the US labour market — representing approximately 151 million workers and roughly $1.2 trillion in annual wages. Crucially, the study focused not on theoretical exposure to automation but on jobs where AI can perform equivalent tasks at a cost that is already competitive with human labour. The window for treating AI disruption as a distant future issue has closed.

Harvard Business School research tracking US job postings from 2019 through March 2025 found that after ChatGPT's launch, postings for occupations involving structured and repetitive tasks declined by 13%, while demand for analytical, technical, and creative work grew 20%. The labour market is not collapsing — it is bifurcating. Execution is being automated; judgement is being amplified.

The structural shift in one sentence AI does not eliminate work. It eliminates the work that can be precisely specified in advance — and sharply raises the value of the work that cannot.

What is actually being automated — and what is not

The pattern emerging from the data is more nuanced than the headline figures suggest. What AI is replacing is not intelligence in general — it is a specific category of cognitive work: tasks that are structured, repetitive, documentable, and where quality can be evaluated against a clear standard. The World Economic Forum's Future of Jobs Report 2025, surveying over 1,000 employers representing 14 million workers across 55 economies, projects that 92 million jobs will be displaced by 2030 while 170 million new ones will be created — a net gain of 78 million, concentrated in roles requiring judgement, contextual reasoning, and interpersonal capability.

The practical examples are instructive. Amazon eliminated 14,000 corporate roles in 2025, explicitly citing AI-enabled leaner structures. Workday cut 8.5% of its workforce — approximately 1,750 jobs — to reallocate resources toward AI investments. In the first six months of 2025 alone, 77,999 tech job losses were directly attributed to AI, per Challenger, Gray & Christmas data. These are not manufacturing jobs. They are analyst roles, junior developer positions, content functions, and middle management layers — precisely the work that was considered safe from automation a decade ago.

What is not being automated maps to a consistent profile across the research: work that requires genuine contextual judgement, the ability to operate in ambiguous or novel situations, sustained relationships with specific humans, and accountability for outcomes that cannot be specified in advance. The IMF's January 2026 analysis emphasises this explicitly — AI complements human labour particularly in decision-making, pattern recognition across novel contexts, and knowledge retrieval with contextual application.

Being automated	Becoming more valuable
Structured data analysis and reporting	Interpreting ambiguous data in context
First-draft content generation	Editorial judgement and framing
Routine legal and financial research	Strategic advice with client context
Entry-level code and testing	System architecture and problem framing
Middle management coordination	Senior leadership and direction-setting
Customer service at scale	High-stakes client relationships
Translation and transcription	Cross-cultural negotiation and trust

Sources: WEF Future of Jobs Report 2025; Harvard Business School / Srinivasan et al. 2026; IMF 2026; Challenger Gray & Christmas 2025.

The skills that are becoming scarce

The PwC 2025 Global AI Jobs Barometer — the most comprehensive analysis of its kind, covering close to a billion job advertisements across six continents — found that workers with AI skills command a 56% wage premium compared to peers in identical roles without them. But the more instructive finding is what those AI-skilled workers are actually doing: they are not writing code. They are directing AI systems, interpreting outputs, applying judgement to AI-generated work, and taking responsibility for decisions that AI cannot be held accountable for.

The IMF's analysis of millions of job postings across advanced and emerging economies shows that one in ten job postings in advanced economies now require at least one genuinely new skill — skills that did not appear in postings a decade ago. Roles with four or more new skills pay up to 15% more in the UK and 8.5% more in the US. The skills commanding the highest premiums are not technical in the narrow sense. They are the intersection of technical fluency with capabilities that remain fundamentally human.

1. Direction and framing

The ability to define what problem is worth solving, to frame a question precisely enough that AI can produce a useful answer, and to recognise when AI output is subtly wrong — these are capabilities that compound in value as AI execution capacity grows. A consultant who can direct an AI system to produce a 40-page market analysis in four hours has not lost value — but the value now resides in knowing what analysis to commission, what assumptions to challenge, and what the client actually needs to decide. The execution cost has fallen; the direction cost has not.

2. Contextual judgement

AI performs well on well-defined problems with clear evaluation criteria. It performs poorly — sometimes invisibly poorly — on problems where the relevant context is tacit, the evaluation criteria are contested, or where the stakes of being subtly wrong are high. A fractional CFO advising a founder on whether to raise a Series B is not performing a task that AI can replicate. The judgement required draws on pattern recognition across dozens of similar situations, an understanding of specific investor relationships, and a read of the founder's psychology. None of this is documentable in advance.

3. Accountability and trust

AI can produce a recommendation. It cannot be held responsible for it. As AI penetrates professional services, the value of a human professional who will stake their reputation on a specific recommendation — and who will be available to manage the consequences — rises significantly. This is not a soft benefit. It is a structural market premium. Law firms, consultancies, and advisory practices that can credibly attach human accountability to AI-assisted work product are positioned at a premium relative to those that simply deliver AI outputs.

4. Relationship capital

The 2025 LinkedIn Work Change Report identified relationship-building and stakeholder management as among the fastest-growing skill demands across all industries. This is counterintuitive in a world of remote work and digital tools — but the logic is clear. As execution commoditises, access to the right people, the right problems, and the right opportunities becomes the binding constraint on professional income. Relationship capital — the accumulated trust and access built over a career — is not replicable by AI, not transferable, and not subject to cost deflation.

The optionality imperative

The professionals who will navigate this transition well are not necessarily those with the highest technical skills or the most AI fluency. They are those who have structured their professional lives to maintain options — multiple income streams, portable skills, independent distribution channels, and a reputation that is not entirely dependent on a single employer's continued existence.

Consider the difference between two mid-career professionals with equivalent experience. The first has spent a decade building deep expertise within a single organisation, with a salary, a job title, and an implicit assumption that the role will persist. The second has built equivalent expertise but distributed across clients, published independent analysis, maintained a professional network outside their employer, and developed independent income streams. When AI eliminates a layer of their organisation's middle management, the first professional is exposed. The second has options.

This is not a hypothetical. The WEF projects that by the end of 2026, 20% of organisations will use AI to flatten their hierarchy, eliminating over 50% of current middle management positions. The professionals most exposed are those who manage coordination and information flow — precisely the functions that AI handles most efficiently. The professionals least exposed are those whose value is in the judgement they apply to novel situations, the relationships they hold, and the accountability they can credibly offer.

The core optionality insight Single-employer dependence is a structural vulnerability in an AI-accelerated labour market. The professionals who build career optionality — portable expertise, independent distribution, multiple income sources — are not just more financially resilient. They are more valuable, because optionality itself signals the kind of scarce, high-judgement capability that AI cannot replicate.

How to prepare — five concrete moves

The research is consistent on what distinguishes professionals who are building value in an AI-transformed labour market from those who are not. The following five moves reflect both the empirical evidence and the practical logic of the transition underway.

1. Develop a point of view, not just expertise. Expertise — knowing things — is being commoditised at a rate that most professionals have not fully absorbed. What AI cannot replicate is a specific, developed, defensible perspective on a domain: the kind of opinion that a senior practitioner has earned through repeated exposure to failure and success in context. Publishing analysis, taking positions, and building a track record of specific judgements is now a professional asset with compounding value.
2. Build distribution independent of your employer. The single most important career infrastructure investment in an AI-accelerated market is owning your audience. A newsletter, a body of public writing, a professional reputation that precedes you — these are the distribution channels that make your expertise findable and referable independent of any organisational affiliation. PwC's analysis notes that demand for formal degrees is declining for most jobs, including AI-related ones. Credentials are becoming less valuable; demonstrated judgement, evidenced in public work, is becoming more so.
3. Develop multiple income streams before you need them. The fractional work model — selling expertise to multiple clients on a retained basis — is not just a financial strategy. It is a hedge against the concentration risk of single-employer dependence. The IMF's 2026 analysis finds that employment in AI-vulnerable occupations is already 3.6% lower after five years in regions with high AI adoption. Professionals who have already developed fractional or consulting income when disruption arrives are far better positioned than those who begin the search under pressure.
4. Invest in the skills that compound with AI, not against it. The PwC barometer identifies a 56% wage premium for workers with AI skills — but the relevant competency is not technical AI development. It is the ability to direct, evaluate, and take responsibility for AI-assisted work. A professional who can commission, review, and improve AI-generated analysis with genuine domain expertise is worth more than one who can either do the analysis manually or operate the AI tool without domain understanding. The combination is where the premium lives.
5. Narrow your positioning deliberately. The professionals least exposed to AI displacement are those with a specific, hard-to-replicate combination of expertise, context, and relationships. A generalist consultant is more exposed than a sector-specific adviser with ten years of client relationships in a specific industry. Narrowing is counterintuitive when it feels like reducing your market — but in a world where AI can approximate generalist knowledge instantly, specificity is the source of scarcity.

The question to ask yourself

The useful frame is not 'will AI take my job?' — it is 'which parts of my work could be precisely specified in advance, and which parts require something that cannot?' The former is vulnerable. The latter is where your professional value is increasingly concentrated.

The professionals who are building career optionality now — diversifying their income, owning their distribution, developing a defensible point of view, and positioning themselves as the human layer of accountability above AI execution — are not hedging against disruption. They are positioning for the labour market that is already arriving.

Execution is becoming cheap. Direction, judgement, and accountability are not. The scarcity that AI creates is the same scarcity that has always been at the centre of the most valuable professional careers — it is simply becoming more visible, and more urgent, than it has ever been.

Check out our Optionality Scorecard to check your readiness to go independent!

Optionality Lab publishes analysis on career structures, independent income, and professional autonomy for mid-career professionals.