Where the value of AI really goes — and 6 ways to capture a share

The cheap-intelligence paradox

The working headline matches the core story: where the value of AI really goes, and six ways to capture a share.

The paradox of the current AI economy is that consumers can be much better off while many workers feel no richer. A monthly subscription can now provide something that resembles basic legal guidance, tutoring, coding help, writing support, research assistance and first-pass medical information at any hour, at a price closer to a streaming bundle than to a professional invoice [1]. That is a real increase in lived welfare. It is also hard to see in conventional growth statistics, because a large part of the gain arrives as consumer surplus: people receive cheaper help, faster answers and more options, but not necessarily higher pay [1].

This is why the public debate keeps sounding confused. One side looks at the usefulness of the tools and sees abundance. The other looks at entry-level hiring, bargaining power and ownership concentration and sees fragility. Both can be right. The new Australian speech by Andrew Leigh, published on October 13, frames the issue cleanly: AI can compress productivity differences between workers while widening the gap between people who earn mainly from labour and people who own productive capital [2]. Capital Brief’s report on the same speech made the political point sharper: AI may lift some less-experienced workers, yet still concentrate wealth among a powerful few [3].

Consumers get the surplus first

The first destination of AI value is the consumer. If a student receives unlimited explanations, a founder gets decent contract drafts, a developer gets debugging help and a parent gets a simplified medical explanation before calling a doctor, the immediate beneficiary is not necessarily the worker selling expertise. It is the person who suddenly receives a slice of expertise almost for free.

That matters. Cheap digital expertise can reduce friction in daily life, especially for people who previously could not afford accountants, tutors, analysts or consultants. In that sense, AI is egalitarian at the point of use. It gives ordinary users access to capabilities that were once rationed by price, geography or professional gatekeeping [1].

But this is not the same as distributing income. A consumer can gain value without receiving cash. A worker can use AI and become faster without gaining bargaining power. A firm can reduce costs while passing some savings to customers and some to shareholders. Leigh’s speech stresses exactly this distinction: productivity forecasts alone say little about inequality; the distribution depends on hiring, ownership, competition and market power [2].

Workers may gain capability but lose leverage

The second destination is more complicated: workers gain tools, but not necessarily power. AI can help a junior employee perform closer to a more experienced one on bounded tasks. That sounds good, and in some contexts it is. It can make training faster, reduce errors and open doors for people without elite credentials.

Yet if a firm needs fewer juniors because each remaining worker is augmented, the wage story changes. The productivity gap inside a team may narrow while the total number of early-career opportunities shrinks. HelloBro’s source brief points to the vulnerability of entry-level knowledge tasks such as first drafts, research, simple coding, proofreading and administrative work [1]. The newest freeCodeCamp engineering report makes a related point from the software side: code generation is cheaper, but judgment, verification and expertise become more valuable precisely because generated work still has to be directed and checked [4].

That creates a new hierarchy. People who can supervise AI, define the problem, own the client relationship, sign the work or carry liability remain scarce. People whose role is mainly routine execution face pressure. AI does not erase expertise; it changes where expertise must sit.

The durable rents sit below and around the model

The third destination is infrastructure. Models attract attention, but the most durable rents may sit around them: chips, GPUs, data centers, power, cooling, cloud platforms, distribution channels and identity layers. HelloBro’s brief argues that model capability is being commoditized quickly, while scarce assets around the model remain harder to replicate [1].

This is the deep reason the AI economy feels less like a labour revolution and more like an ownership revolution. If intelligence becomes cheap, the price of “having an answer” falls. Value migrates to whatever remains scarce: compute, proprietary data, trusted workflows, regulated authority, customer access and physical infrastructure. Race to AGI’s analysis of Leigh’s remarks reaches the same conclusion: widely deployed models can boost output for average employees, but the main financial upside can accrue to owners of models, chips and platforms [5].

The result is a stack economy. At the top, users get cheap help. In the middle, many workers become more productive but more replaceable. At the bottom, owners of compute, energy, distribution and capital collect tolls.

GDP misses part of the story

AI also exposes a measurement problem. When a €20 tool replaces occasional paid help worth far more to the user, conventional GDP may record only the subscription fee. The saved time, avoided anxiety and improved decision-making are real, but they do not appear as neatly as a lawyer’s invoice or a consulting contract [1].

This is why households can simultaneously enjoy better digital services and feel poorer. AI can make information, drafting and software assistance cheaper while housing, healthcare, childcare, education and energy remain scarce. If the money saved on digital help is redirected into rent or insurance, the user may experience AI as useful but not liberating.

That gap between welfare and income is politically explosive. People do not pay mortgages with consumer surplus. They pay them with wages, dividends, business income or public transfers.

Six ways to capture a share

1. Own scarce assets, directly or indirectly

The cleanest way to capture AI value is to own part of the scarce layer. That can mean public equities, diversified funds, infrastructure vehicles, data-center suppliers, energy assets or businesses with defensible distribution. It does not have to mean betting on one frontier lab. The principle is broader: if capital captures more of the surplus, households need some exposure to capital. Leigh explicitly highlights broad participation in capital ownership as one way workers may share in gains [2].

2. Sell outcomes, not hours

Hourly billing becomes weaker when clients believe AI has reduced the time required. Outcome pricing is more resilient. A consultant, lawyer, designer, marketer or developer should package results: a working system, a compliant filing, a qualified lead pipeline, a signed customer, a deployment, a cost reduction. HelloBro’s brief makes this point directly: fixed-fee contracts, revenue shares and results-based work protect better than reselling time [1].

3. Attach yourself to trust and liability

The signature becomes valuable. So does certification, insurance, fiduciary duty, medical responsibility, safety review and professional accountability. AI can draft a memo; it cannot fully absorb legal responsibility for the advice. Workers who remain close to trust, audit and liability can keep pricing power because the buyer is not only purchasing output. The buyer is purchasing someone who can stand behind it.

4. Build proprietary workflows and data

Generic prompting is fragile. Proprietary workflow is stronger. A small firm that embeds AI into its own customer data, operating procedures, measurement loops and distribution can create value that a generic chatbot cannot easily copy. The durable advantage is not “we use AI.” It is “we use AI inside a system only we can see.”

5. Become the human who verifies the machine

The freeCodeCamp report’s software evidence points toward a general rule: the more AI generates, the more valuable high-quality verification becomes [4]. The winning worker is not the person who blindly accepts output. It is the person who knows what to ask, how to test, when to reject and where the hidden failure modes live.

6. Move toward physical scarcity

AI makes digital work abundant before it makes housing, nursing, logistics, energy repair, construction, advanced manufacturing or local trust abundant. Hybrid roles that combine AI fluency with physical-world execution may be safer than roles that live entirely in generic digital production. The plumber, nurse, electrician, robotics technician, lab operator or field engineer using AI may have a better moat than the prompt-only knowledge worker.

The real divide

The divide is not between people who use AI and people who refuse it. The deeper divide is between those who consume cheap intelligence and those who own, direct, verify or insure the scarce systems around it.

That is the uncomfortable conclusion. AI can make everyday life more convenient and still widen wealth gaps. It can help novices and still weaken the ladder that turns novices into experts. It can lower prices and still concentrate rents. The question is no longer whether AI creates value. It does. The question is who captures it.

Sources from the last 72 hours

1. [1]Where the value of AI really goes (and 6 ways to capture a share)Oct 12, 2026, 5:30 PM
2. [2]Speech: Will AI Increase Inequality? - 13 October 2026Oct 12, 2026, 3:00 PM
3. [3]AI may have complex effects on productivity, inequality: Andrew LeighOct 13, 2026, 1:39 AM
4. [4]The State of AI and Software Engineering: Code is Cheaper but Expertise are More Valuable [Full Book]Oct 12, 2026, 2:00 AM
5. [5]Andrew Leigh warns AI may lift wages yet widen wealth gapOct 12, 2026, 2:00 AM

AI-generated article based on recent web research, then preserved as a dated editorial snapshot.