The Capability Layer Frontline labour markets · No. 01 · 2026

Professional identity · Frontline work

The world's largest professional network is still offline

Not because its members lack phones. Because the products built so far assume a shape of career that most of the world's workers do not have.

A professional profile is not one thing. It is four assumptions stacked together, and a frontline career satisfies none of them. That is why the gap between LinkedIn's 1.3 billion members and the world's 3.5 billion workers does not close with cheaper data or better translation. It closes only by changing what the record is made of, and who owns it.

If you have one minute

Four assumptions, and a career that meets none of them

S1The muddle

Ask why frontline workers are not on LinkedIn, and you will be handed the wrong answer

LinkedIn passed roughly 1.3 billion members in October 2025.1 The global labour force is around 3.5 billion. Put those two numbers beside each other and the industry reaches, almost reflexively, for a story about reach: the platform is large and growing, the remaining workers are poorer and harder to serve, and the fix is cheaper data, lighter pages and more languages.

That story is comfortable because it makes the problem logistical. It is also wrong, and the evidence against it is sitting in plain view.

Pew Research found in January 2024 that 53 per cent of American adults holding a bachelor's degree use LinkedIn. Among those with a secondary education or less, the figure is 10 per cent.2 That is a 43-point spread, and it is the steepest education gradient Pew records across any platform it tracks. Note where those respondents live. This is the United States, where smartphone ownership is near-universal, data is cheap and the interface is already in the user's first language. Every logistical barrier the reach story depends on has been removed, and the gradient is still there.

So the sorting is not being done by infrastructure. It is being done by the product.

Two objections arrive at this point, and both are worth taking at full strength. The first is that 10 per cent is not zero, so the product plainly works for some workers without degrees. That is correct, and it is the framework below doing its job rather than a hole in it: the 10 per cent are overwhelmingly people in formal employment with transferable titles and named employers, which is to say the ones who satisfy the assumptions despite not holding a degree. The gradient tracks the assumptions, not the credential.

The second objection is that the gradient reflects income and job-search custom rather than design. Graduates change employers more often, in markets where online search is the norm, so of course they populate a professional network. This is the stronger objection, and it explains part of the spread. It does not explain the rest. A worker who changes employer frequently has more reason to carry a portable record, not less, and the populations with the highest job mobility on earth are the informal and migrant workforces described later in this paper. If search frequency drove adoption, they would be the platform's heaviest users. They are absent instead, which points back at what the product requires before it will hold anything.

Exhibit 01 / 04
Strip out every access barrier and the gradient survives, which means the product is doing the sorting
LinkedIn use among United States adults by education level, in a market where device access, data cost and language are not constraints.
SHARE OF US ADULTS WHO USE LINKEDIN, 2024 Bachelor's degree or higher 53% Some college 28% Secondary education or less 10% 0% 53% A 43-point spread, the widest education gradient Pew records for any platform.
Source: Pew Research Center, Americans' Social Media Use, January 2024. United States respondents only; the education gradient is measured, the inference about product design is the argument of this paper.
S2The framework

A professional profile is four assumptions stacked together, and a frontline career satisfies none of them

Take the profile apart and look at what it silently requires before it can hold anything at all.

It requires a job title that travels. "Senior analyst" means roughly the same thing at the next employer, which is what allows a one-line entry to carry information. It requires a record the worker maintains, authored in the first person and updated over a career. It requires colleagues who are findable and will vouch in public, which is what converts a set of claims into a network with something at stake. And it requires employers that outlast the tenure, so that naming one transfers its reputation to the person who worked there.

Each assumption is doing real work. Remove one and the profile degrades but survives: a consultant between firms still has the title and the network. Remove all four and there is nothing left for the product to hold, which is the position a domestic worker in Riyadh or a welder in Gauteng occupies from the first day of their working life.

This is the difference between an access problem and an architecture problem. An access problem is solved by lowering a cost. An architecture problem is solved only by changing what the thing is made of.

Exhibit 02 / 04
Remove one assumption and the profile degrades; remove all four and there is nothing to hold
What a professional profile requires before it can carry information, set against what a frontline career supplies.
WHAT THE PRODUCT ASSUMES WHAT A FRONTLINE CAREER SUPPLIES A job title that travels One line carries meaning at the next employer "Housemaid" describes a station, not a capability A record the worker maintains Authored in the first person, updated over years Nothing written down that she controls Colleagues who vouch in public Claims become a network with something at stake Worked alone, in a private household Employers that outlast the tenure Naming one transfers its reputation A family, whose name signals nothing to a stranger Four of four absent. The product has nothing to hold.
Schematic. The four requirements are derived from the product's own field structure; the right-hand column describes the Saudi household-work case.
S3The scale

This is not an edge case at the margin of the labour market. It is most of the people in it

The ILO's Employment and Social Trends 2026 puts 2.1 billion people in informal employment.3 Its earlier statistical update, covering 2019, gives the more useful framing: roughly six in ten of everyone employed on earth.4

Two groups inside that population are where this work starts. The ILO counts 167.7 million international migrant workers as of 2022, more than 30 million above its previous estimate.5 It counts at least 75.6 million domestic workers worldwide as of 2019.6

Hold those proportions against the framing in the first section. A product that excludes six in ten workers by design is not serving a market with a long tail. It is serving the minority and calling it the profession.

Exhibit 03 / 04
The excluded population is not a tail, it is the majority of everyone who works
Workers in informal employment set against total global employment, with the two populations this work begins in.
GLOBAL EMPLOYMENT, BY VISIBILITY TO THE PROFESSIONAL RECORD 2.1bn informal Remainder of global employment Roughly six in ten of everyone employed, on the ILO's 2019 statistical update. WITHIN THAT POPULATION, WHERE THIS WORK BEGINS 167.7m International migrant workers, 2022 75.6m Domestic workers worldwide, 2019
Sources: ILO, Employment and Social Trends 2026; ILO, Women and Men in the Informal Economy, 2023; ILO Global Estimates on International Migrant Workers, 4th edition, 2024; ILO, The Road to Decent Work for Domestic Workers, 2023. Data years differ by series and are labelled.
2.1bn
In informal employment, on the ILO's most recent count
167.7m
International migrant workers, up 30m on the prior estimate
75.6m
Domestic workers, superseding the widely quoted 52.6m
S4The lens applied

Follow one worker for ten years and watch capability accumulate while the signal keeps resetting to zero

I spent two years running worker matches by hand in Riyadh before UpSkillMe existed as software. A hundred and twenty of them, paid, for households with real briefs.7 What that work taught me was not that matching is hard. It is that the same information gets destroyed and rebuilt every single time.

A woman with ten years in one household has learned a great deal. She has run a home to a standard, managed a budget, worked across a language barrier, handled a child through illness, and built the judgement that only shows itself when something goes wrong at eleven at night. The United States Department of Labor's own occupational database records twenty distinct tasks for her job classification.8 None of them is written down anywhere she controls.

Three consequences follow, and they follow structurally rather than through anyone's ill intent.

Her only available reference is her current employer, which is a conflict of interest at precisely the moment she most needs an honest signal. The person best placed to confirm her capability is the person with the most to lose when she leaves. Her record does not survive a border, so migration resets references, networks and recognised experience even though the underlying capability crosses intact. And nothing compounds: in a professional career each role makes the next more legible, while here each role starts from zero.

Exhibit 04 / 04
Capability compounds in a straight line while the hiring signal is zeroed at every transition
The divergence between what a worker has actually learned and what a prospective employer can see, across ten years and three employers.
ACCUMULATED CAPABILITY VERSUS VISIBLE HIRING SIGNAL What she can actually do What a new employer can see Job change Border crossing Year 0 Year 10 The gap at year ten is the mispricing. It is also, precisely, the thing a portable record closes.
Analytical model of the mechanism described above. The transitions are real features of the career; the curve shapes are not measured quantities.
Worked example, illustrative

How much of a ten-year career survives one border crossing?

Years worked in household roles10
Distinct tasks recorded for the occupation820
Capabilities a former employer could confirm, if asked and willingup to 20
Former employers reachable from the destination country0 to 1
Claims a new employer can check before hiring0
Effective hiring signal, expressed as years0

The worker arrives with identity verified, contract registered and wage protected, and with nothing at all recorded about what she can do. Figures show the method and the shape of the loss, not a specific case.

S5Where the lens breaks

The framework has two real limits, and the second one is the more interesting

Every framework earns trust by naming where it stops working, so here are both places this one does.

The first limit is that the four assumptions are not equally binding everywhere. A unionised construction worker in Germany has a portable qualification, findable colleagues and a durable employer, and lacks only the maintained record. The framework predicts, correctly, that such a worker is far better served by existing products than a Filipina housekeeper in Riyadh. The argument is strongest where all four fail together, which is the informal economy and the migration corridors running into it, and it weakens in proportion as any assumption is restored.

The second limit is that the framework assumes the worker would benefit from being visible, and visibility is not free. A record that follows a worker between employers also follows her into a dispute, and a market where employers can see more is not automatically a market where workers do better. Anyone building this has to hold that objection rather than wave at it. The design answer is worker-held control over disclosure rather than a public profile, but I would rather state the tension than claim it is fully solved.

There is also something the framework cannot reach, and it is worth naming because the absence is itself a finding. In preparing this piece I went looking for a basic figure: what share of the global workforce holds a verifiable, portable record of what they can do. The ILO, the World Bank, UNESCO, the World Economic Forum's skills-first work. It is not measured. There are market-size estimates for credentialing technology and employer surveys about hiring intentions, but no count of how many workers actually hold portable proof of their own capability.

We count employment carefully. We count unemployment carefully. We do not count whether the people employed can demonstrate what they are able to do, because the question has never been treated as an economic one.

The instrument does not exist because nobody decided the measurement mattered.

S6The bottom line

If the exclusion is architectural, then translation is the wrong investment and ownership is the right one

The practical consequence of this framework is a reallocation, and it cuts against where most effort currently goes.

If the barrier were access, the right investments would be device subsidies, data zero-rating, interface translation and literacy support. Those are all worth doing on their own merits, and none of them puts a single checkable claim into a worker's hands. The Pew gradient is the evidence: in a market where all four have effectively been achieved, the exclusion persists.

If the barrier is architecture, the investment is different. It means building a record whose unit is a demonstrated capability rather than a job title, whose provenance is attached to each claim so a stranger can check it, and whose owner is the worker rather than the employer, the agency or the platform. Ownership is the load-bearing part, because the three consequences in section four all trace back to the same root: the record belongs to the party with an interest in it staying quiet.

For employers and agencies in the Gulf, the test is narrower and more immediate. If your hiring decision currently rests on a job title, a number of years and a reference from the person who is losing the worker, you are not making a judgement about capability. You are making a judgement about paperwork, and the paperwork was never measuring the thing you care about.

The world's largest professional network is neither LinkedIn nor anything I am building. It is the two billion people who already work, already carry real capability, and whose professional reputations exist entirely in other people's memories. That network is offline because of a design decision about what a record is made of, and design decisions can be made differently.

Disclosure

UpSkillMe

The case above stands on its own evidence; nothing in it depends on what follows. I am the founder of UpSkillMe, which builds a worker-owned capability record for household work in Saudi Arabia and the trades in South Africa. The company has a direct commercial interest in the conclusion that ownership of the record is the decisive variable, and readers should weigh the argument accordingly.

Next in this series: what "five years' experience" actually predicts, and why the answer is almost nothing.

Sources

  1. Microsoft, Q1 FY26 earnings call, October 2025, reported by GeekWire (2026). LinkedIn's own corporate site states the rounder figure of more than one billion members.
  2. Pew Research Center (2024). Americans' Social Media Use. Survey of United States adults, fielded 2023, published 31 January 2024.
  3. International Labour Organization (2026). Employment and Social Trends 2026, published 14 January 2026.
  4. International Labour Organization (2023). Women and Men in the Informal Economy: A Statistical Update. Reference year 2019.
  5. International Labour Organization (2024). ILO Global Estimates on International Migrant Workers, 4th edition. Reference year 2022.
  6. International Labour Organization (2023). The Road to Decent Work for Domestic Workers. Reference year 2019. Supersedes the 52.6 million estimate from the 2013 report.
  7. Founder transaction records, Riyadh, 2025 to 2026. Unaudited.
  8. O*NET OnLine, United States Department of Labor, occupation 37-2012.00, Maids and Housekeeping Cleaners. Task count parsed from the database, 2026.

Written in British English. Figures are quoted at the reference year of their source, which differs between series; where a widely circulated figure could not be traced to a primary source it has been excluded rather than repeated. Claims about informal employment, migrant workers and domestic workers are measured estimates; the four-assumption framework is an analytical claim and is argued, not measured.