Rails carry claims; the network earns its keep by making them trusted.
Frommarketplace→Toopen coordination layer
Open protocols can already carry learning offers, work opportunities and verifiable credentials between systems, and ONEST, built on Beckn, was designed to do exactly that for education, skilling and work. What protocols can’t supply is agreement on what evidence means, accountability for distributed work, enforceable worker rights and someone to pay for verification. That is where an open human-capital network has to create its value and recover its costs, without owning the worker. Its network effect isn’t guaranteed: it exists only if each verified transaction makes the next one cheaper. This last shift decides whether the other eight add up.
Isn’t this what ONEST does?
Partly, and that is good news, because no new rail is needed. ONEST, the Open Network for Education and Skilling Transactions, adapts Beckn’s decentralised skilling and education protocol, known as DSEP [2]. At launch, ONDC was named as its network facilitator, EkStep as incubator and FIDE as protocol partner [3]. Its published design covers four categories: learning experiences, financial aid, work opportunities and mentorship [4]. Its use cases include course discovery and enrolment, issuing and verifying credentials, matching learners to opportunities and hiring on the basis of verifiable credentials, and feedback and ratings [1]. Its trust layer is built on verifiable credentials and electronic registries [4]. Its stated objective is to break up a siloed ecosystem in which education providers, learners and job providers work in isolation from one another [5]. That is the same diagnosis this series started from. Public documentation describes it as moving from pilots toward ONDC’s network infrastructure and governance [1, 4].
Other rails exist too. Open standards for badges, verifiable credentials and learner records, credential registries and state programs for learning-and-employment records already move evidence between systems [6]. Malta runs its job matching through the engine of Flanders’ public employment service, VDAB, as a shared service. VDAB publishes its competency taxonomy through an open interface, so vacancies and profiles can use the same vocabulary [7].
So this series builds on those rails. The machinery to discover, enrol, credential and match increasingly exists. The question this series adds is economic: which object moves across these transactions, and who pays for it to be trusted?
Rails can carry a claim. Trust in the claim has to be built on top of them.
What is actually new
Almost every component exists somewhere. The review separates what already exists, what exists but only in fragments, and what exists only experimentally.
| Already exists | Exists, but in fragments | Experimental |
|---|---|---|
| Skills frameworks and competency-based education | Learning linked to later demonstrated performance | AI-generated career and learning pathways |
| Registered apprenticeship and intermediaries | Portable evidence connected to outside employer demand | Agent-based job search and automated recruiting workflows |
| Work samples and project-based assessment | Apprenticeship shared across independent trainers and employers | Skills wallets that work across platforms |
| Badges, verifiable credentials, wallets and learner records | Worker-controlled identity connected to continuous discovery | Repayment tied to outcomes for short training |
| Skills-first hiring and internal talent marketplaces | Assessor reputation tied to later job performance | State and sector skills marketplaces |
| Algorithmic matching in public and private services | Work outcomes repaying capital for future learners | |
| Income-contingent training finance |
That supports three different claims of novelty, and they differ a great deal in strength. Structural novelty is about how the components connect. Here the hypothesis is reasonably strong, since no full precedent turned up in the sources checked. Economic novelty is about whether the connection creates surplus that existing arrangements don’t. Commercial novelty is about whether anyone pays for that surplus. The second and third are unproven.
The review names only two elements as potentially new at the level of economic design. One is a portable capability record that outcomes keep updating, with provenance and later validation contributed by assessors, trainers, employers and workers. The other is a revolving, worker-protective capital mechanism that finances formation across providers. Both are hypotheses. The most defensible statement of what is new is therefore a narrow one.
The components are not new. The bet is on evidence that stays useful after the transaction that created it.
The thesis
Open commerce for human capital is an open economic architecture in which people can continuously form, demonstrate, carry, discover and exchange capability across organisational boundaries. Its foundational mechanism is to turn job-relevant work into trusted, reusable evidence that travels with the person and lowers the uncertainty and repeated cost of later transactions.
Around that evidence layer, distributed knowledge providers and trainers contribute to capability formation. A worker-held living identity carries evolving evidence. A personal agent orchestrates each person’s pathway according to their intent. Demand increasingly discovers capability rather than waiting for applications. And over time, accumulated evidence of capability and income may support new forms of human-capital finance.
The original concept set out to make human capability as discoverable, composable and economically accessible as other forms of networked commerce [CN]. The evidence review reached a narrower formulation. In it, a sector-specific network may create incremental value if it turns real work into portable, trusted evidence and uses that evidence to reduce employer screening and the friction between training and work. The open identity, agent and revolving-finance layers remain unvalidated and could add substantial privacy, governance and financial risk.
That is where this series ends. Not with we have built the future of human capital, but with a narrower and more useful claim. There is a credible economic primitive to test, with clear mechanisms around it and identifiable points of failure. And there is a path from one capability family to the larger network.
The worker doesn’t need another platform. They need proof that travels with them.
Go deeper
What the rails don’t solve
Interoperability isn’t the same as agreement. Two systems can exchange a credential and still disagree about what the skill means. They can disagree about the level of mastery it shows, the context of the performance, how old the evidence is and how good the issuer is. Settling that takes semantic governance, versioning, dispute resolution and expiry rules, not more transport.
Employers add their own constraint. HR systems were largely built around degrees and full qualifications, and many can’t yet register smaller credentials or skills data [6, 8, 9]. Beyond meaning, somebody still has to take responsibility for distributed apprenticeship (Shift 4), enforce the worker’s rights (Shift 5) and pay for verification (Shift 2).
A division of responsibilities
| Layer | Responsibility |
|---|---|
| The person | Intent, consent and every final decision |
| Personal agent | Pathway, representation and decision support (Shift 7) |
| Living identity | Worker-held evidence, trajectory and terms (Shift 5) |
| Evidence and verification | Rubrics, independent verification, provenance and expiry (Shift 2) |
| Accountable intermediary | Supervision, contracts, IP, worker protection and disputes within a capability family (Shift 4) |
| Knowledge providers, trainers and employers | Learning, practice and real work |
| Open rails such as ONEST and Beckn | Discovery, ordering, fulfilment and carrying credentials between participants |
| Public and philanthropic funders | Anchoring, first-loss capital, standards and enforcement of rights |
Don’t rebuild the rails. Build the trust and coordination above them.
The network, assembled
The loop that matters runs from work to evidence to trust to discovery and back to work. The network doesn’t just list human capital. Every completed transaction can produce evidence of capability that was previously invisible outside the organisation that saw it.
The flywheel has to be earned
There are two loops. The individual loop runs work → evidence → trust → discovery → work. The ecosystem loop runs demand → capability gaps → formation → work → evidence → better pathways. Only the loops that run through reusable evidence create distinctive surplus, and the others mostly create activity. More profiles can mean more noise, and more assessments can mean credential inflation. More data can mean more bias and surveillance, and more providers can mean a heavier curation burden.
The review therefore treats the network effect as unknown until reuse savings are shown to exceed the costs of infrastructure, verification, governance, privacy and coordination. Signals of demand across the network carry a further risk of herding many people into the same capability (Shift 7).
The unit of the network is not the profile. It is the verified transaction.
The platform trap
Profiles plus search plus ranking plus fees add up to a platform, even on open rails. The slide is predictable. First the network ranks workers. Then placements get sponsored. Then the operator starts owning the evidence, because that is where the value sits, and the result is another platform.
If identity and evidence are genuinely portable, the network can’t rely on owning the worker’s data. It has to earn its keep through functions participants will pay for. Those include verified evidence and provenance, assessment standards, and integration with employers’ hiring systems. They also include accountable coordination and compliance, dispute resolution and correction, privacy-preserving identity services, measurement of outcomes and, later, risk management for finance.
Who pays?
Every participant needs an economic reason to join, and the network needs someone with a budget for coordination, verification, privacy, compliance and disputes. The review is clear about the order. The employer is the most plausible commercial payer, because it captures the savings in screening, vacancies, mismatch and retention. Public or philanthropic funders are likely to be needed for early experimentation, equity, first-loss capital and socially valuable pathways. Workers shouldn’t pay for unproven evidence or for being discoverable.
| Flow | Payer | Recipient | When | Condition |
|---|---|---|---|---|
| Work | Employer or client | Worker | On delivery or through payroll | Always paid, with explicit consent for any evidence produced |
| Apprenticeship | Employer, consortium or public co-funding | Apprentice (wage), and trainer or intermediary | During training | Output plus retention covers supervision and coordination |
| Verification | Employer or consortium | Assessor | At assessment | Replaces some repeated testing |
| Discovery and coordination | Employer subscription or success fee | Intermediary or network operator | At shortlist, hire or retention | The fee is lower than the savings in screening and mismatch |
| Learning | Employer, public funder or worker | Knowledge provider | Before or during learning | Fee for service now; outcome-linked only once contribution can be measured |
| The worker’s agent | Not yet decided (Shift 7) | Agent operator | Ongoing | Must not depend on placement fees |
| Finance | Public or philanthropic first-loss capital, blended later | Fund | Before formation | A later layer (Shift 8) |
Money should come from whoever receives measurable value.
Structural defences
Six operating rules keep the layer open. Evidence and identity can be exported, and workers can use more than one channel. No place on a shortlist is for sale. A published share of every shortlist goes to exploration and new entrants (Shift 6). Rubrics are open and verification is independent.
Matching outcomes are also audited across groups; New York City began requiring bias audits of automated hiring tools in 2023 [10]. And workers are represented in governance. There is precedent for that last rule. One of the sector programs that produced large earnings gains is run by a partnership in which worker representatives play a central part in design, administration and operation [11].
What not to build
The first product shouldn’t be a universal capability graph, another credential wallet, another job board or another course marketplace. It shouldn’t use autonomous agents that sign contracts on anyone’s behalf. It shouldn’t make everyone discoverable to everyone, and it shouldn’t start with a revolving fund. The first mission has to be narrow enough to measure (Shift 1).
Ten ways this breaks
This list merges the review’s falsification tests and the points at which the thesis would break. Each is tested in the shift noted.
| Failure mode | How it breaks the thesis | Tested in |
|---|---|---|
| 1. Thin or unpaid demand | Too few employers hire from the family repeatedly, or none will pay for evidence | Shifts 1 and 9 |
| 2. No reuse | Second employers re-test anyway, because evidence is stale, gamed or hard to read | Shift 2 |
| 3. Illusory composition | Assembled micro-skills don’t add up to capability anyone will hire | Shift 3 |
| 4. Negative net contribution | Apprentice output never covers supervision in knowledge work | Shift 4 |
| 5. No accountable party | Nobody takes responsibility for supervision, liability, IP and disputes | Shift 4 |
| 6. Forced disclosure | Useful discovery requires giving up privacy or correction rights | Shift 5 |
| 7. Automated applications | Matching multiplies contacts, tests and bids rather than transactions | Shift 6 |
| 8. Misaligned agent | The agent serves whoever pays for it | Shift 7 |
| 9. Steering capital | Finance needs permanent subsidy or pushes careers toward what repays | Shift 8 |
| 10. Platform drift | The operator recovers its costs by owning worker data | Shift 9 |
Not every failure is fatal. If revolving finance fails, the network can still run as an evidence and matching system on conventional funding. If apprenticeship fails, paid projects and work samples can generate evidence instead. But if reuse or worker-controlled discovery fails, the claim to be open commerce for human capital fails with it, and what remains is a better training-and-credential product.
The equation and the kill test
Surplus from reuse (screening, testing, mismatch, time to productive work) − (evidence + verification + coordination + governance + privacy) = net network value
If the result is negative, interoperability is just another layer of cost. If it is positive in one capability family, the network has a reason to exist there, and a template for the next family.
| Kill-test question | Answered in |
|---|---|
| Does verified evidence of work beat a résumé, reference, structured interview and work sample, once its cost is counted? | Shift 2 |
| Will a second employer reuse the evidence, and pay for it? | Shifts 1, 2 and 9 |
| Can learning inside paid knowledge work cover its own supervision? | Shift 4 |
| Can demand start transactions without costing workers their privacy or pay? | Shifts 5 and 6 |
| Does the whole system create surplus after coordination and governance? | Shifts 1 and 9 |
The review names the single strongest finding that would kill the thesis. It would come from a controlled employer experiment in which verified evidence of work plus network discovery does no better than a conventional résumé, reference, structured interview and work sample, once the costs of producing the evidence and coordinating the network are counted.
If the tests fail in particular ways, the honest move is to rename the product.
| If the evidence shows | What it really is |
|---|---|
| A résumé, reference, structured interview and work sample capture all the value | An improved assessment product |
| Employer-linked training does as well at lower coordination cost | An apprenticeship intermediary |
| Portable credentials cut verification effort but change nothing else | Credential infrastructure |
| Matching raises applications but not accepted work or pay | Recruiting automation |
| Gains come only from selection, placement and support | A training and placement program |
| Finance needs permanent subsidy | A subsidised financing product |
Each of these is a legitimate thing to build. None of them is an open network for human capital.
How we’d test it
| The test | What we would measure | What would end the idea |
|---|---|---|
| Network surplus (Experiment 7) | For one regional or sector network of 20–40 employers, against matched employers: duplicated assessments, screening hours, vacancy duration, output, training cost, retention, earnings, governance cost, disputes and public spending | Benefits appear only through subsidy, or activity rises without lower total cost or better outcomes |
| Portability audit | Whether employers accept evidence from independent issuers without the original platform involved | Acceptance depends on the issuing platform |
| Platform drift | Share of revenue from placement or data; evidence exports; use of more than one channel | Revenue depends on placement or on control of data |
| Recurring payment | Whether any participant keeps paying fees after the pilot | No one pays |
How strong is the evidence?
- Established
- Open protocols and credential standards can carry learning offers, work opportunities and credentials between systems [1, 2, 6]
- Evidence-supported
- Technical interoperability doesn’t produce semantic agreement, employer acceptance or liquidity [8, 9]; a multi-sided human-capital network needs an anchor to start
- Hypothesis
- An open coordination layer can recover the cost of trust through verification, coordination and integration, without owning worker data
- Unknown
- Whether network effects emerge across the sides, whether governance can fund itself, and who pays for the worker’s agent
Still to confirm
The evidence review did not examine ONEST. Confirm its current status and scale, and whether its specifications now cover evidence of work or apprenticeship flows. The review also did not examine India’s related public infrastructure, including DigiLocker, the Academic Bank of Credits, APAAR IDs, the National Credit Framework and the Skill India Digital Hub. These are leads for the next evidence pass, not evidence.
Sources
- [1]ONDC, ONEST specifications repositoryExcerpt
- [2]ONEST Network, ONEST specification repositoryExcerpt; its work-opportunities schema was checked for the Hyperlocal Commerce series
- [3]YourStory, article on ONEST’s launch as a decentralised network for education, skilling and jobs (2023)Excerpt
- [4]Innovation Hub for Prosperity, ONEST use caseExcerpt
- [5]ONEST, starter packExcerpt
- [6]National Governors Association, resource page on learning and employment recordsChecked
- [7]World Bank, Solutions for Youth Employment, draft note on digital job-matching platforms in public agencies (2023)Checked (draft discussion note)
- [8]Cedefop, Microcredentials for labour market education and training: the added value for end users (2023)Checked (executive summary and chapters 1–3)
- [9]Aspen Institute, UpSkill America, report on employer perspectives on digital credentials and skills profiles (2024)Checked
- [10]Broecke, Artificial Intelligence and Labour Market Matching, OECD Social, Employment and Migration Working Papers (2023)Checked
- [11]Katz, Roth, Hendra and Schaberg, Why Do Sectoral Employment Programs Work? Lessons from WorkAdvance, Journal of Labor Economics (2022)Checked
- [CN]Open Commerce for Human Capital: concept document consolidated from the discussion (internal)Provided
Spotted a source we missed or a rule we misread? Tell us what we got wrong.