Verifiable Credentials for the Global Workforce

Almost every hiring decision, contract award and freelance engagement rests on a document the candidate wrote about themselves. The CV, the portfolio, the profile, the reference supplied on request: none of these is evidence. Each is a claim. For most of the world’s workers it is the only claim available, because there is no employer to call and no payroll record to check. And generative AI has now made a polished, tailored, entirely plausible version of that claim free to produce in seconds.

This paper makes the case for work provenance: proof that comes from the party who received the work, captured at the time it happened, checkable by anyone, and owned by the person who earned it. It sets out the problem with data from the International Labour Organization, the World Bank, Harvard Business School and Stanford University; explains why detecting AI-written applications cannot repair the hiring signal; defines the principles a verified work history must satisfy to be worth anything; and describes how Can0nical applies them.

Contents

1. 01The problem: work that leaves no record
2. 02Why AI detection cannot fix the hiring signal
3. 03What work provenance is
4. 04The verification stack: identity, history, and work
5. 05Ten principles for a verified work history
6. 06How verified work history works
7. 07What changes for professionals, employers and managers
8. 08Use cases
9. 09Where this goes next
10. 10References

The problem: work that leaves no record

It is tempting to describe verification as a fraud problem. That framing is defensive, and it misdiagnoses the cause.

The real condition is that the systems around work do not remember it. A contract ends and the record of it ends too. You leave a company and your track record stays inside their systems. You leave a marketplace and the reviews stay behind. You move country and start from nothing. Nothing in the way work is organised is designed to remember what a specific person delivered, so the entire burden of remembering, and of re-proving, falls on the individual, permanently.

Fraud is a symptom of that condition. When nothing is reliably remembered, everything becomes a claim, and claims are cheap to fabricate. Fix the memory and the fraud problem shrinks without being attacked directly.

Why AI detection cannot fix the hiring signal

The industry’s current instinct is to detect: scan applications for AI, score writing style, flag anomalies in a document. This fails for three structural reasons.

1. 1

   It is an arms race against a falling cost curve. Generation improves faster than detection, and every detector becomes training signal for the next generator.

2. 2

   Its errors are asymmetric and invisible. A fabricated document that passes is silent. An honest candidate auto-rejected for phrasing is also silent, and the employer never learns what they lost. A system whose failures are both invisible cannot be improved by the people operating it.

3. 3

   It examines the wrong object. Detection interrogates the document. But the document was never the evidence; it was always a summary of events that happened somewhere else. Analysing the summary more aggressively cannot recover information the summary never contained.

Provenance changes the object under examination. It does not ask whether a description of work looks authentic. It asks whether the party who received that work confirmed it, at the time, in an attributable way. That is a categorically stronger claim, because it is anchored to something outside the candidate’s control. And it degrades gracefully: even a partial record of three confirmations from three named organisations carries more information than a flawless, unverifiable document.

What work provenance is

Proof of work is a phrase most people now associate with cryptocurrency. It had a plain English meaning first, and that is the one we mean: evidence that the work was actually done.

The unit of that evidence is not the task, the project or the CV line.

The atomic unit of work provenance is the confirmation: a dated, attributable, tamper-evident attestation by an organisation that a specific person delivered specific work.

It is also the canonical unit. Everything else (a CV, a public profile, a reputation score) is a rendering of a set of confirmations, and where two renderings disagree the confirmations settle it. That is what makes the record authoritative rather than merely stored. The confirmations are the asset. A verified work history is what you get when they accumulate.

The verification stack: identity, history, and work

Most people have no mental slot for work provenance and will file it under either background checks or identity verification. It is neither. There are three layers, and almost all existing investment and competition sits in the bottom two.

Layers 1 and 2 both answer questions about a person’s status. Neither answers a single question about their work. Identity verification can establish with high confidence that the person holding the passport is real, alive and not sanctioned. It cannot indicate whether they can do the job. A background check confirms someone held a title between two dates. It does not confirm they delivered anything.

Both are necessary. Neither is sufficient, and the gap between them is exactly where hiring and contracting decisions fail. Layer 3 is the canonical layer for the work itself, and it is the one nobody occupies. The three layers are complements, not competitors. Stated plainly: identity verification proves you are a real person; work provenance proves you did the work.

Why the third layer behaves differently

Identity is becoming verify-once-reuse-everywhere; that is the explicit design of the EU wallet framework, and the W3C data model gives it a portable format. As identity becomes something a person already carries rather than something each service buys per user, the natural commercial move for identity providers is to climb the stack, and the climb from “who are you” is “what have you done.”

Work provenance should therefore be expected to become contested. The defence is not secrecy or a proprietary format. It is that a network of real confirmations between real organisations cannot be purchased, only accumulated, which makes starting early the whole of the strategy.

Ten principles for a verified work history

A record is worth exactly what its weakest guarantee allows. These are the principles Can0nical is built on. Several of them foreclose revenue, and that is the point: a commitment that costs nothing to keep is not a commitment.

1. 01

   Confirmed by the party who received the work, never by us

   A record’s value comes entirely from who attested to it. We are a registrar, not a rating agency. We record that a client confirmed the work; we do not grade how good it was, rank people against each other, or sell placement. A verifier that grades has become an opinion, and opinions can be bought.

2. 02

   Captured close to the event

   Retrospective attestation degrades quickly: memories fade, contacts move, companies dissolve. Provenance is built continuously, as work happens, which is why logging is designed to take two minutes at the end of a day, not an afternoon at the end of a job search.

3. 03

   Independently checkable, without our permission

   Anyone holding a record can verify it themselves: no account, no key, no request to us. A record only we can vouch for is a record that depends on our survival.

4. 04

   Owned by the person who earned it, and paid for by them

   Nobody should lose access to the evidence of their own working life because a third party stopped paying. The individual side of this business stays subject-paid, permanently, with any later revenue layered on top of that floor rather than replacing it.

5. 05

   Never for sale

   Verification outcomes, reputation scores, ranking, badge placement and preferential display are not, and will not be, purchasable. This forecloses real revenue. Stating it publicly is what makes it credible.

6. 06

   Only verified information is displayed

   Nothing appears on a Can0nical profile that Can0nical has not verified through its own mechanism: no self-declared certifications, no endorsements, no aspirational fields. The only self-asserted items permitted are identity presentation: name, picture, and social links. This costs us a richer-looking profile, permanently.

7. 07

   Never surveillance

   Records are voluntary and authored by the professional. We will not build or sell keystroke tracking, screen capture, activity scoring or productivity ranking, including when an enterprise buyer asks and offers to pay for it. A record someone is compelled to produce is not evidence of trust.

8. 08

   Minimum permanent data

   Permanence and privacy are in genuine tension. The resolution is that only a cryptographic fingerprint of a record is permanent, never a name, an email address, a narrative or a company name. Personal data stays deletable on request without breaking anyone’s ability to verify the record.

9. 09

   Fraud resistance is our cost, not the user’s problem

   Per-event checks catch naive cases; they will not catch a reciprocal ring of people confirming each other’s work over months across plausible domains. We intend to fund network-level detection ahead of the growth that requires it, because the value of every record in the system is capped by the weakest fraud it tolerates.

10. 10

    Standards over lock-in

    We intend to align with W3C Verifiable Credentials and wallet interoperability, to publish the verification method as an open specification, and to let people carry their records elsewhere. The aim is to win because the network is here, not because the exit is blocked.

How verified work history works

Four steps, from a day’s work to evidence that outlives the contract.

Two properties of that loop are worth stating explicitly, because they are what make the resulting record credible rather than merely convenient.

Nothing is recorded as proof that a counterparty has not confirmed. Unconfirmed entries are drafts. They are never presented as verified, and they never reach a public profile.

The professional and the client are the authors; software assists. Can0nical uses AI to suggest the skills evidenced by confirmed work and to draft documents from records that already exist. It never invents a claim, and it never confirms one. People attest; software organises.

From an accumulating set of confirmations, a professional can generate a blockchain-anchored CV, a public profile at a permanent identifier, and an integrity score derived from confirmed work rather than from self-assessment. Each one is a rendering of the same underlying evidence, and any of them can be checked against the chain by a stranger holding nothing but the identifier.

What changes for professionals, employers and managers

For professionals, including the two billion people whose work currently leaves no formal record, the change is the end of the re-proving tax. Evidence accumulates instead of resetting, and it belongs to the person who earned it rather than to whichever counterparty happened to keep the records. A career becomes cumulative.

For companies, the change is the recovery of a signal that has degraded into noise. Confirmed delivery is a different class of input from a self-authored document: attributable, dated, and checkable at the moment of decision rather than days later through a reference call. For an employer who removed the degree requirement and reached fewer than one hire in 700, it is the input that makes the intent executable.

For managers, the change is that the judgement they already exercise stops evaporating. Sign-off happens anyway, in email, in stand-ups, in meetings. Provenance makes that act durable and attributable, for the person receiving it and for the manager who gave it.

For the market as a whole, the implications run past hiring. A verified, counterparty-confirmed history of delivery is genuinely novel input for decisions that currently exclude independent and informally employed workers outright: credit, insurance, tenancy, cross-border compliance. The population least served by conventional verification infrastructure is the one for which provenance is most consequential.

The professional trust ecosystem

A verified work history proves one person to one outside party. That is the useful thing about it on day one, and it is not the interesting thing about it in year five.

Every confirmation adds an edge to a network: a named organisation on one side, a named professional on the other, and a date attached to the work between them. As those edges accumulate, the people inside the network come to hold something no individual profile can offer on its own: a pool of professionals whose work has already been confirmed by someone else in the pool. That is what we mean by a professional trust ecosystem: not a directory of members, but a body of evidence its members generated for each other.

Our intention is to make that pool usable by the people who built it. In time, members should be able to find, vouch for and hire each other on the strength of confirmed delivery rather than reputation by assertion: a consultancy staffing a project from professionals whose work its peers have already verified, a founder engaging a contractor whose record was confirmed by a company they know, an agency subcontracting with the evidence already in hand rather than a portfolio to interpret. Cross-hiring inside the ecosystem turns a trust signal that members curated collectively into opportunity that returns to them.

Two commitments govern how we build it, and they are the reason it can work at all. Placement will never be for sale. No ranking, no promoted position and no paid visibility, ever, because the moment introductions can be bought the pool stops being evidence and becomes advertising. And the signal has to stay earned: cross-hiring within the ecosystem is worth something only while every confirmation inside it remains a considered attestation rather than a favour traded between members. The value of the whole pool is capped by the honesty of the smallest confirmation in it, which is why nothing in this product will ever reward volume over truth.

Use cases

Where this goes next

Can0nical today records confirmed work, anchors it so that anyone can verify it independently, extracts the skills that confirmed work evidences, and turns the result into a professional CV, a public profile and an integrity score. The free tier includes logging and client confirmation, together with invoice and timesheet tools that need no account at all. Four directions follow from there.

The self-report is failing as a signal at the exact moment work without an employer becomes the majority condition, and regulation standardises verifiable credentials. Detection cannot repair it, because detection examines a summary rather than the events the summary describes. The response that holds is provenance: proof that comes from the party who received the work, captured when it happened, checkable by anyone, and owned by the person who earned it.

That is what Can0nical is building, starting with the professionals nobody vouches for.

Good work shouldn’t expire. Your reputation shouldn’t reset.

Logging and client confirmation are free. No card to start.