The portable future is (not quite) here
What Anthropic, OpenAI, the Trump Administration, and the Gates Foundation are missing about unlocking economic opporutnity
In the past year, Anthropic, OpenAI, the Trump Administration, and the Gates Foundation have all released policy agendas that center one of the newest buzzwords in the world of economic opportunity: portability.
This term often comes up related to benefits (tax-advantaged savings accounts like retirement with 401ks and IRAs, health insurance and HSAs, and Skill Savings Accounts featured in legislation that was introduced earlier this month), but AI has put the focus specifically on data.
The ED and DOL’s vision for portability with education and training
As we’ve digitized the world over the past 30 years, we’ve created a bit of a mess. There’s more data than ever, but it can’t flow across systems in a reliable way.
In 2018, Credential Engine reported that there were over 300,000 different types of credentials in the US (college degrees, licenses, certifications, certificate, badges, apprenticeships, microcredentials, bootcamps). Today, we have almost 2,000,000.
There’s nothing necessarily wrong with more education and training options, especially if people are actually pursuing them, but without a centralized system to manage all of these non-college credentials, employees have a hard time signaling their skills, and employers have a hard time trusting that those signals are legitimate. Even the marginal credential issuer can’t break through the noise to prove the value of their program compared to their competitors.
As part of America’s Talent Strategy, the ED launched the Connecting Talent to Opportunity Challenge earlier this year in order to help states build and scale talent marketplace platforms where all credentials and skills can become verified via Learning and Employment Records (LERs), and portable across employers and state lines.
The technology to make interoperability possible has existed for a long time, but the social and political pressure to solve the coordination challenge that unlocks portability has hit a new peak, thanks in large part to people like Nick Moore at the ED who led a similar initiative in Alabama and Jason Tyszko at the U.S. Chamber of Commerce Foundation who has been pushing this movement across their initiatives for years.
Historically, the U.S. government has cared a lot about helping solve labor market inefficiencies and failures. During the Great Depression, the Wagner-Peyser Act created job exchanges that were the predecessor for the thousands of job centers across the country today. In the early days of the internet, the DOL created “America’s Job Bank (AJB),” the country’s first national online job board that reached millions of job postings and users years before LinkedIn and Indeed.
However, with the mixed ROI data on today’s job centers and projects like AJB being ultimately sunset due to funding shortages, the government’s inconsistent track record with bold projects like this begs the question: will this actually work?
As is the case for most government projects, execution and continuity across administrations is difficult. As is the case for most coordination challenges, aligning all stakeholders around a common standard or system is hard. And as is the case for most massive data transfers, some packet losses and errors are likely.
However, this movement that has the support from the highest levels of government, philanthropy, and tech still hasn’t addressed the real reason why our labor market underperforms on matching efficiency, and especially how digitization and automation not only won’t fix it, but might even make it worse.
Solving portable identity and history aren’t enough
Before we even get to a hiring decision, verifying that people actually are who they say they are, and that people have actually done what they say they have done is a lot harder than it seems.
Think about all the systems we have just to confirm identity: REAL IDs, passports, fingerprints, passwords, MFA, TSA Pre-Check Touchless ID, CLEAR, KYC, cryptography, and background checks. Even with these advancements, fraud and security are still major concerns and huge and growing industries, especially when we’ll need all of this for agents, too.
Then, consider the significantly fewer and weaker systems we have to confirm education and work history: transcripts, credentials, and employment verification. It’s only been a few years since LinkedIn started allowing users to voluntarily verify their identity, education, and employment. Building the “Checkr for education/employment” where verifications are quick and cheap would be immensely valuable.
However, we’ve long known that while this information is necessary, it is nowhere near sufficient to make hiring decisions, especially not at scale. Knowing where someone has been done doesn’t tell employers much about what they can actually do.
This is the gap that the skills-based hiring movement has tried to close.
We can’t afford to have “dark fiber” marketplaces
In a blog post published by The Aspen Institute earlier this year titled “Missing Demand May Miss the Opportunity: Research Insights for States Building Talent Marketplaces That Work”, the authors emphasize an important reminder about the ED’s CTO Challenge: none of this matters if employers won’t use the platforms. If all we do is focus on making something sound great for employees and neglect the needs of the employers, we’ll end up losing them both.
State talent marketplaces are most effective when they support the seamless exchange of skills data as the currency of a functioning market. That data, often packaged as digital credentials, should make it easier to understand what skills learners have and how those skills align with what employers need. Strong platforms also surface clear signals about which skills are in demand, allowing both learners and employers to adjust and respond to one another in real time.
Many existing playbooks for building talent marketplace platforms have focused heavily on capturing and organizing the supply of skills data. Far less attention has been paid to fully engaging employers.
…
Too often, new state platforms replicate existing approaches without meaningfully strengthening feedback loops between learner skills and employer demand. States typically start on the supply side by developing expansive skills taxonomies and registries, “skillifying” resumes, and creating digital wallets where learners can store information about their achievements. But there are few ways to test whether that output actually reflects what employers are looking for. The result is an explosion of skills data and credentials that can be difficult for employers to interpret or differentiate.
States can build stronger feedback loops by designing platforms that track how employers engage with skills data. This means looking not only at the skills employers hire for, but also at the types of information they rely on to make decisions and the level of detail that is most useful in real hiring contexts.
- The Aspen Institute, “Missing Demand May Miss the Opportunity: Research Insights for States Building Talent Marketplaces That Work”
While I agree with all of the above and acknowledge all of the value that the skills-based hiring movement can and will continue to bring to the labor market, I’m not sure I see digital wallets like LERs with validated credentials and skills data “as the currency of a functioning market.”
Skills are the wrong currency for the labor market
If a lack of verifiable skills and skill-based job descriptions were truly the unlock we needed, that assumes that the vast majority of learners and employees have been flat out lying in their job applications about their skills, and I don’t think that’s true.
For example, is the #1 problem with hiring in the trades that a bunch of people are claiming they went to HVAC school when they really didn’t? I’m sure that happens sometimes, but I’m not confident that’s the real problem worth a new multi-million dollar, multi-state government program.
Inherent in the definition of a skill is that it’s something that can be learned, including on the job. A small minority of jobs require a specific license prior to starting on day #1 (especially among the margins of the labor market in need of the most intervention), and a small minority of firms expect to get a fully-skilled employee out of the box that requires no training or skill development. What exactly are all of the skills that are essential enough to need verification but not assessable enough as part of the hiring process so that this pre-application infrastructure is necessary?
If we assess it as a strong market “currency”, it doesn’t pass the test. Skills mean different things to different people, are valued differently by different stakeholders, and are non-binary and therefore extremely subjective. Just because a PDF certificate turns into a checkmark in a digital wallet doesn’t make employers trust “skills” any more than they did before. Marketplaces built on skills are technically feasible and politically popular, but might be distracting us from the real market failure: information assymetry that causes a lack of trust.
The labor market does need a new currency, but skills aren’t it.
Work performance data, or the system I’ve been referring to as Open Work, could be what we’re looking for.
Knowing how someone handled responsibilities that were fully within their control with prior employers is perhaps the best indicator of their character, potential, and effectiveness in a new role. Rather than talking about skills in the abstract, we can root them in real data.
Which candidate would you take?
Applicant A: no direct skills or experience within the industry, but a verified performance history that showcases that they are extremely dependable and adaptable
Applicant B: AI-generated resume and cover letter full of skills jargon with an online credential in a digital wallet on a government-run website
You can teach the former candidate to do anything and have the confidence that they’re a hard worker and fast learner; HR has been long aware of the shortcomings of only hiring the latter.
Jevon’s paradox, but for human labor
If we made effective hiring easy, cheap, and trustworthy by leveraging past work performance data, what if we actually saw hiring increase substantially just like we’re seeing with AI?
What if, even amidst the job disruption with AI and robotics, we see a Jevon’s paradox, not for agentic or robotic labor, but for human labor because underwriting is stronger, faster, cheaper, and easier than ever before?
What if labor demand wasn’t weak, but justifiably risk-averse given how the labor market currently functions, and Open Work was able to unlock that latency?
What if our labor supply, even amidst shortages in crucial industries, was never under-skilled, but just underrevealed?
If these types of ideas keep you up at night, too, I’d love to connect.