Job seekers, employers, and AI hiring systems are colliding with the same missing layer: capability infrastructure.

Publication Note

This article is part of the OSI-PIA public article series. It describes an active research and development direction. It does not claim that OSI-PIA, OSI, PIA, PCA, or any related prototype is a finished hiring product, validated employment-screening system, legal compliance tool, psychological assessment, or production decision system.

This article uses the term capability infrastructure to describe the systems, records, standards, workflows, translation layers, and governance practices that allow human capability to be represented, interpreted, trusted, routed, developed, and used without reducing people to brittle proxies.

Article

You cannot find work. Companies cannot find workers.

Workers say they are applying, learning, earning credentials, building skills, and still disappearing into silence. Employers say they have openings, but cannot find qualified people. Recruiters face volume they cannot reasonably interpret. AI tools promise efficiency, but often accelerate the same old assumptions. Internal employees hold skills their organizations do not see. Job seekers are told to optimize resumes while companies complain that the talent they need does not exist.

These sound like separate problems. They are not.

That does not mean every worker is qualified for every job. It does not mean every employer is careless, every applicant tracking system is broken, or every credential is useless. It means something more structural: the systems connecting people, evidence, work, and opportunity are failing at interpretation.

The labor market cannot read people.

Everyone Is Naming a Piece of the Same Failure

The public conversation already has many names for this problem. Some call it a talent shortage myth. Some call it a screening problem. Some call it the AI hiring gap. Some call it a skills visibility gap. Some call it hidden talent. Some call it faulty signals. Some call it resume translation failure. Some call it outdated hiring. Some call it a credential problem. Some call it a workforce planning problem.

They are not wrong. They are incomplete.

Each phrase names one part of a larger infrastructure failure. When companies say there is “no talent available” while people remain unemployed, underemployed, or stuck in jobs that do not match their capability but pay the bills, the issue is not necessarily a simple supply problem. It may be a coordination problem between how capability is presented and how it is evaluated.1 When companies rely on rigid filters, employment-gap penalties, keyword screens, degree requirements, or knockout questions, they may reject people they later claim they cannot find.2

When organizations say they need skills but cannot see the skills already inside their workforce, the issue is not only hiring. It is internal capability visibility. TalentLMS’s skills visibility research frames the problem around skills that exist but remain underused, unrecognized, or disconnected from opportunity inside organizations.3

When AI changes work faster than old hiring patterns can interpret, the issue is not only technical skills. It is whether the system can recognize contextual reasoning, judgment, communication, domain knowledge, and adaptive learning. The World Economic Forum’s 2025 Future of Jobs work projects significant labor-market churn by 2030 and identifies skills gaps as a major barrier to business transformation, while LinkedIn’s 2025 Work Change reporting says the skills needed for jobs may change substantially by 2030 under AI pressure.45 Open Data Science’s discussion of the AI hiring gap similarly emphasizes transferable capability, domain expertise, judgment, and the limits of hiring systems that search for already-perfect candidates instead of developing people.6

When employers use degrees, job titles, brand-name institutions, or exact prior experience as substitutes for evidence, those proxies may become gates that look efficient while hiding capable people. Stand Together’s discussion of hidden talent and degree requirements frames this as a faulty-signal problem: employers may rely on degree screens even when capacity to learn, practical skill, and aligned training are more relevant to the work.7

The problem is larger than any one tool. It is capability infrastructure.

The Old Career Model Is the Deeper Problem

The labor market is not only struggling because resumes are weak, applicant tracking systems are brittle, or AI is moving quickly. It is struggling because many of its assumptions were built for a different working world.

The old model imagined a relatively linear path: education, credential, career identity, stable employer, upward ladder, retirement. That model still shapes how schools advise students, how employers write job descriptions, how hiring systems screen candidates, how people explain themselves, and how organizations decide what counts as legitimate experience.

But working lives no longer move that way for many people. Fast Company recently argued that education and career preparation remain too oriented around a “one-job career” model, even as modern work increasingly requires reinvention, flexibility, lateral thinking, and multiple forms of contribution.8

Labor-market data supports the broader point. In January 2026, BLS reported median employee tenure of 4.1 years overall, 3.0 years for workers ages 25 to 34, and 20.6% of wage and salary workers with one year or less at their current employer.9 BLS longitudinal data also reports that younger baby boomers held an average of 12.7 jobs from ages 18 to 56, while older Millennials held an average of 9.0 jobs from ages 18 to 36.10

The exact number of “careers” someone has is harder to measure than the number of jobs, and BLS is careful about that distinction. But the pattern is clear enough for this argument: many people do not move through work as a single fixed professional identity attached to one employer or one ladder.

People change employers, roles, tools, industries, locations, responsibilities, and identities across longer working lives. They learn formally and informally. They build capability through paid work, unpaid responsibility, caregiving, military service, side projects, community work, self-directed study, contract work, operational repair, and repeated transitions.

The system still asks an older question: “What are you?”

The better question is: “What capability can you demonstrate, where is the evidence, and how should that evidence be interpreted now?”

That shift matters. A labor market built around fixed professional identities will misread nonlinear people. An education system built around one-job careers will underprepare people for reinvention. A hiring system built around static proxies will confuse unfamiliar paths with lack of capability.

This is why capability infrastructure matters. It is not only a hiring fix. It is the missing architecture for working lives that are longer, more transitional, and more evidence-rich.

What Capability Infrastructure Means

Capability infrastructure is the missing layer between human capability and institutional decision-making. It includes the ways people preserve evidence of what they have done, the ways organizations interpret that evidence, the standards used to compare evidence to requirements, the systems that route people toward opportunity, and the governance boundaries that prevent those systems from becoming extractive.

A labor market with weak capability infrastructure asks people to prove themselves through brittle signals: the right job title, degree, keywords, uninterrupted work history, platform profile, network connection, timing, formatting, and guess about what a system wants.

Some of those signals can be useful. None of them is the person.

A stronger capability infrastructure would ask better questions. What evidence supports this claim? What capability does that evidence demonstrate? What context shaped the work? What is directly proven? What is inferred? What is uncertain? What has expired? What can the person correct? What should the system refuse to assume?

That is a different kind of labor-market architecture.

The Resume Is Carrying Too Much

The resume has become the overloaded object at the center of the labor market. It is expected to summarize a person, satisfy a parser, match a job description, persuade a recruiter, orient a hiring manager, survive keyword filters, explain nonlinear experience, compress years of learning, and prepare the ground for an interview.

That is too much work for one document.

A resume can be useful. It can summarize, orient attention, and help someone enter a process. But a resume is not a person. It is not the full evidence record. It is not the history of capability development. It is a projection.

The problem is not that resumes exist. The problem is that modern systems often treat the projection as if it were the underlying reality.

That is how people become misread.

The Labor Market Is Using Proxies as If They Were Proof

A proxy is not always wrong. Degrees can matter. Certifications can matter. Prior roles can matter. Tools can matter. Experience can matter. Keywords can help systems route information.

The danger comes when proxies become substitutes for interpretation.

A degree may be a real requirement in one role and an inherited filter in another. A certification may show exposure but not execution. A job title may hide work that was much broader than the title suggests. A career gap may signal risk, or it may signal caregiving, illness, relocation, study, recovery, self-employment, or unpaid responsibility.

A keyword match may show familiarity, or it may show that someone copied the language of the posting. A polished AI-generated profile may read well without proving much. A messy work history may contain exactly the adaptive capability the organization needs.

The question is not whether signals matter. The question is whether the system knows what kind of signal it is reading.

Without that distinction, the labor market confuses evidence, proxies, interpretation, and decision.

Capability Can Exist and Still Remain Invisible

Capability does not disappear just because a system fails to see it.

A job seeker may have the ability to do the work but fail to pass the intake shape. An employee may have skills the organization never maps. A frontline worker may understand a broken process better than the people authorized to change it. A caregiver returning to paid work may have developed logistics, crisis management, teaching, scheduling, emotional regulation, advocacy, and administrative skill that no conventional job-title ladder knows how to read.

A veteran may hold operational, training, safety, planning, leadership, and decision-making capability that fails to translate into civilian keywords. A self-taught builder may have real artifacts but lack the credential proxy. A cross-functional worker may look scattered on paper because the organization has no category for the pattern.

A company may then say: “We cannot find talent.”

But the more accurate statement may be: “We cannot see, trust, route, or interpret the capability available to us.”

That is capability leakage.

The Skills Gap Is Also a Visibility Gap

The skills-gap conversation often assumes that the missing skill is somewhere outside the organization. Sometimes that is true. Sometimes the capability already exists but is not visible, trusted, or deployed.

TalentLMS’s skills visibility report puts numbers on that internal visibility problem: 49% of employees said their company underutilizes their skills, 50% of respondents said their company hires externally for skills current employees already have, and only 12% said their organization does not have skills visibility issues.3

Skills can be trapped inside roles that do not name them. They can be hidden inside informal work, mentoring, troubleshooting, documentation, process repair, community responsibility, or self-directed learning. They can be known to coworkers but invisible to leadership. They can be captured in training records but disconnected from work assignments. They can be present in employees who are never asked the right questions.

That is why skills visibility matters.

If an organization cannot see what its people know, it will hire externally for capability it already has, underuse employees who are ready to grow, and misread turnover as a talent-supply issue rather than a system-design issue.

The organization does not only need more data. It needs better interpretation.

AI Can Help the Labor Market Read, or Help It Misread Faster

AI is entering a labor market that already has weak capability infrastructure. That is dangerous.

AI can help organize evidence, compare language, summarize experience, identify gaps, support career navigation, map skills, and improve access to information. Used carefully, it can help people and organizations see patterns they would otherwise miss.

But AI can also accelerate old mistakes. It can turn weak proxies into confident rankings. It can summarize people badly. It can amplify inherited filters. It can reward polish over proof. It can generate resumes that look impressive but detach representation from evidence. It can help employers process more applicants without understanding more people.

The question is not whether AI belongs in hiring, workforce planning, or career development. The question is what role AI is allowed to play.

AI should support evidence organization, translation, comparison, and human review. It should not become the authority over a person’s capability.

The Governance Question

Capability infrastructure is not only a technical problem. It is a governance problem.

A system that represents human capability can help people become visible, but it can also become extractive. It can clarify evidence, but it can also create hidden profiles. It can support better decisions, but it can also harden weak assumptions into automated authority.

That is why the architecture matters.

Who owns the evidence? Who has authority to interpret it? Who can correct it? Who can export it? Who can see it? How long does it remain valid? What should the system refuse to infer? What remains human judgment?

Without those boundaries, capability infrastructure becomes surveillance infrastructure. With those boundaries, it can become a way to make human capability more visible, portable, reviewable, and usable without taking ownership away from the person.

The Person Owns the Output

The central OSI-PIA principle is that personal capability outputs should belong to the person they describe.

A resume, capability profile, evidence packet, professional identity report, portfolio brief, or interview proof pack should not become trapped inside an employer system, vendor platform, or opaque scoring model. The person should be able to review it, correct it, export it, reuse it, withhold it, update it, and understand how it was produced.

That does not mean employers must accept every claim. It means the person should not lose control over the evidence system that represents them.

The organization can evaluate fit. The system can help translate evidence. AI can help organize and compare. But the person owns the capability output.

Personal Professional Packages Are the Bridge

If working lives no longer follow a one-job, one-career, one-credential path, then people need a way to carry their capability across transitions.

A resume alone cannot do that. A profile alone cannot do that. A credential alone cannot do that. A platform account alone cannot do that.

People need personal professional packages: owned, portable, updateable collections of capability evidence that can generate different outputs for different contexts.

A useful package might include a resume, capability profile, evidence packet, portfolio links, credential records, project summaries, interview proof pack, learning history, role-specific crosswalks, and provenance notes. Some parts may be human-readable. Some parts may eventually become machine-readable. Some may be public. Some may remain private until shared.

The point is not to create a larger resume. The point is to give the person a governed source structure from which trustworthy outputs can be produced.

That matters because current hiring systems are not going away overnight. Applicant tracking systems, job boards, employer portals, credential platforms, professional networks, and AI screening tools already exist. A 21st-century capability system has to meet those systems where they are while pushing them toward something better.

The bridge is linkage.

A person-owned professional package should be able to produce conventional outputs for current systems while also carrying richer evidence for future systems that know how to read it. It should support a standard resume when needed, a structured capability packet when possible, and an evidence-backed profile when a reviewer needs more than keywords.

That is also where public workforce conversations about skills passports, recognition of prior learning, common skills languages, modular learning, and micro-credentials become practical. OECD’s skills-first labor-market work points toward systems that connect learning, work, and mobility through more portable and recognizable structures.11

That is how the labor market begins to move: not by waiting for every employer system to modernize at once, and not by forcing people to rebuild themselves separately inside every platform, but by giving people portable, governed capability packages that can connect to the systems that exist now and prepare for the systems that should come next.

What the Missing Layer Needs to Do

A useful capability infrastructure needs at least five layers: evidence, translation, recognition, deployment, and governance.

1. Evidence Layer

People need ways to preserve proof of what they have done, learned, built, solved, repaired, led, translated, or developed. That evidence should include more than job titles and dates. It should include context, artifacts, constraints, outcomes, tools, learning, review status, and provenance.

The future is not the perfect resume. The future is a personally owned evidence system.

2. Translation Layer

A job requirement should not be treated as a flat keyword. It may be a true requirement, a proxy, a credential constraint, a tool signal, a risk signal, a compliance need, or an inherited filter. The translation layer should ask what the requirement is really trying to identify.

Is this a real gap, a tool gap, an evidence gap, a credential constraint, a proxy mismatch, a transferable-capability match, or a partial match needing review?

This is where an individual capability graph can meet a job requirements graph.

3. Recognition Layer

Hiring systems, managers, reviewers, and AI tools need a way to read structured capability evidence without collapsing it into resume noise.

The goal is not automatic acceptance. It is preserving enough signal for interpretation.

A stronger recognition layer would distinguish evidence from proxy, proxy from interpretation, and interpretation from decision.

4. Deployment Layer

Organizations need to see the capability they already have. That means internal skill visibility, better role design, better development pathways, better knowledge preservation, and better routing of people toward work where their capability can be used.

This is the OSI side of the problem.

Where is capability created? Where is it visible? Where is it trusted? Where is it blocked? Where is it underused? Where is it leaving? Where is the organization buying capability it already has?

5. Governance Layer

Capability infrastructure needs rules before it becomes power. It needs consent, provenance, correction, review, temporal authority, individual ownership, and limits on inference.

The system should not silently decide what a person is. It should help evidence become legible while preserving human agency.

What This Is Not

This is not an argument that every candidate is qualified. It is not an argument that employers should lower standards. It is not an argument that resumes, degrees, credentials, experience, or screening systems have no value. It is not an argument that AI should be removed from hiring or workforce systems.

It is an argument that the current system uses too many weak signals as if they were strong evidence, while failing to build the infrastructure needed to interpret capability responsibly.

Standards matter. Evidence matters. Judgment matters. Governance matters.

That is why capability infrastructure matters.

The Better Question

The labor market keeps asking partial questions. How do we make resumes better? How do we fix ATS filters? How do we close the skills gap? How do we make AI hiring fairer? How do we find hidden workers? How do we improve internal mobility? How do we make credentials more useful?

Those are real questions, but they sit inside a larger one:

How should human capability be represented, governed, interpreted, trusted, routed, developed, and owned?

That is the capability infrastructure question.

Conclusion

You cannot find work, and companies cannot find workers, because the labor market cannot read people.

Not because people lack all capability. Not because employers have no real needs. Not because every tool is broken. Because the systems connecting people, evidence, work, and opportunity are failing at interpretation.

The next step is not just better resume advice, more credentials, more automation, or another platform promising perfect matches. The next step is capability infrastructure.

People need evidence systems they own. Organizations need better ways to read capability responsibly. AI needs to support interpretation without becoming the authority. Governance must be built into the architecture before the architecture governs people.

That is the turn. That is what OSI-PIA is building. We have proofs of concept already built. We continue to build more.

Come help build the solution.

The labor market does not only need more talent. It needs better ways to see the talent already there.

Limitations and uncertainty

This article synthesizes public reporting, labor-market research, and the project’s working concepts. It does not establish that any single source explains the labor market as a whole, that capability infrastructure will improve a particular hiring outcome, or that every mismatch is caused by translation failure. The proposed layers remain a research and development direction; their effectiveness, interoperability, and consequences require further testing.

References and Source Notes

  1. Wadhwani Foundation, “Why Companies Say ‘No Talent Available’ While Millions Remain Unemployed,” April 1, 2026. This source frames the contradiction between unemployment and employer claims of no available talent as a coordination problem between how talent is presented and how it is evaluated. https://wadhwanifoundation.org/why-companies-arent-hiring/ ↩

  2. The Interview Guys, “Companies Can’t Find Workers They Need… Because They Keep Rejecting Them: Inside the Self-Inflicted Hiring Crisis,” May 4, 2026. This source is used as recruiting-industry commentary on rigid screening, employment-gap penalties, keyword matching, knockout questions, and applicant-volume bottlenecks. https://blog.theinterviewguys.com/companies-cant-find-workers/ ↩

  3. TalentLMS, “Skills Visibility Report 2026.” This source supports the internal skills-visibility argument, including reported findings that 49% of employees say their company underutilizes their skills, 50% of respondents say their company hires externally for skills current employees already have, and only 12% say their organization does not have skills visibility issues. https://www.talentlms.com/research/skills-visibility-report ↩ ↩2

  4. World Economic Forum, “Future of Jobs Report 2025: 78 Million New Job Opportunities by 2030 but Urgent Upskilling Needed to Prepare Workforces,” January 8, 2025. This source supports the labor-market churn and skills-gap claims, including projected job disruption by 2030. https://www.weforum.org/press/2025/01/future-of-jobs-report-2025-78-million-new-job-opportunities-by-2030-but-urgent-upskilling-needed-to-prepare-workforces/ ↩

  5. LinkedIn, “Work Change Report: Skills for jobs set to change by 70% by 2030,” January 15, 2025. This source supports the claim that skills used in jobs are expected to change substantially by 2030, with AI acting as a catalyst. https://news.linkedin.com/2025/work-change-report-2025 ↩

  6. Open Data Science, “The AI Hiring Gap: Why Companies Can’t Find Talent.” This source supports the argument that AI-era work increases the importance of contextual reasoning, communication and collaboration, decision-making under uncertainty, domain expertise, and internal talent development rather than only spot-market hiring. https://opendatascience.com/the-ai-hiring-gap-why-companies-cant-find-talent-and-workers-cant-find-jobs/ ↩

  7. Stand Together, “America’s labor shortage is a screening problem.” This source supports the faulty-signal and degree-screening discussion, especially the argument that some employers rely too heavily on four-year degree requirements when practical skill, learning capacity, and aligned training may be more relevant. https://standtogether.org/stories/future-of-work/a-hidden-talent-pool-that-employers-cant-ignore ↩

  8. Fast Company, “Stop training kids for a one-job career,” September 25, 2026. This source supports the broader pivot from hiring failure to outdated career assumptions, especially the idea that education and career preparation often still assume a single professional identity while modern working lives require reinvention, flexibility, and broader forms of contribution. https://www.fastcompany.com/91610113/stop-training-kids-for-a-one-job-career ↩

  9. U.S. Bureau of Labor Statistics, “Employee Tenure Summary - 2025 A01 Results.” This source supports the employee-tenure claims: median tenure of 4.1 years in January 2026, median tenure of 3.0 years for workers ages 25 to 34, and 20.6% of wage and salary workers with one year or less of tenure with their current employer. https://www.bls.gov/news.release/tenure.nr0.htm ↩

  10. U.S. Bureau of Labor Statistics, “NLS FAQs.” This source supports the longitudinal job-history claims: younger baby boomers held an average of 12.7 jobs from ages 18 to 56, and older Millennials held an average of 9.0 jobs from ages 18 to 36. https://www.bls.gov/nls/questions-and-answers.htm ↩

  11. OECD, A Skills-First Labour Market. This source supports the move toward common skills languages, modular learning pathways, micro-credentials, skills-first HR practices, career guidance, skills passports, and recognition of prior learning as structures linking learning to jobs. https://www.oecd.org/en/publications/a-skills-first-labour-market_2e1b85f0-en.html ↩