It’s Skills, Stupid
Why this upcoming labor market revolution needs a skills engine
By: Jomayra Herrera
Since the AI craze, I have been writing and re-writing an empowered economy 2.0 article to highlight how to expect workforce dynamics to change in the face of AI (see Part I and Part II here). After hours of framework building, deck editing, and re-reading my prior my pieces, I just kept coming back to a simple truth: it’s all about skills.
Let me elaborate.
In May 2020, at the height of the pandemic, I took a stroll down memory lane and pulled out a 100-slide deck that I created when I was working at Emerson Collective. I wrote a TLDR where I basically said: 1) the education to employment continuum consists of 4 categories: workforce signaling (what are the skills the workforce needs), learning and training, education finance, and employee signaling (what are the skills the workforce has), 2) there has never been great integration among any of those components and, as a result, major technological changes create a lot of pain, and, most importantly, 3) as the rate of change increases the pain will be felt more than ever unless we have a major overhaul of the continuum itself.
In January 2021, I went micro to focus on the worker and discussed the new career stack in an empowered economy - an economy where workers have both the independence they crave and the stability they need. As a worker, you explore career paths, find a job, and ultimately apply your skills - and that process continues. The enablers to help in the career phases are learning (primary learning, reskilling & upskilling), a place to showcase your professional identity and skills, and a community to help you along the way. I suggested there were improvements to be made in each of these categories for workers (e.g. better predictive data for new skills that will be needed).
On May 11th, 2023 at 7:02pm (yes that’s when I felt compelled to start writing this and why my graphics look the way they do), I’m bringing these together and arguing the single biggest opportunity to make this messy labor market work towards an empowered economy is to have a verifiable and adaptive skills engine that powers the overall ecosystem.
Why?
The truth is the rise of generative AI only accelerates the trends we’ve already been seeing - the half life of a skill is shrinking and the need to reskill and redeploy talent is increasing. Will we lose jobs? Yes. Will we gain jobs? Also yes. Will it be painful? Yes, but the severity will depend.
In April, Goldman Sachs released a report on the labor market impact of AI. Their economists analyzed a database of task content for over 900 occupations and estimated that roughly two-thirds of U.S. occupations were vulnerable to automation by AI.
But in this same report, they remind us that 60% of today’s workers are in jobs that didn’t even exist in 1940, implying that more than 85% of employment growth in the last 80 years is driven by innovation-driven positions.
So the question becomes do we have the infrastructure to support the biggest and fastest reallocation of talent we will have ever seen? It took 75 years for 100 million people to get a telephone, 16 years to reach the same adoption for mobile phones, and only 2 months to hit 100 million users for ChatGPT. We need a system that can identity the skills people have, understand the skills the workforce needs, and close the delta between those two as quickly as possible. And I believe the answer to how painful this process will be in whether or not we can have a powerful skills engine to power this transformation.
Well this has always been true, why is now the right time for it to happen?
AI skills inference technology is rapidly improving: The ability for AI to review work, infer the set of skills that are being demonstrated, assess those skills, and create an estimate of the type of work that makes up that role is better than ever and improving everyday.
Which is great, because there are more units of work for a skills engine to assess: As work is increasingly online, it’s much easier to capture an employee’s skill in the flow of their work.
The pain of not being a skills-driven organization is increasing: The pain for both workers and employers will only get worse as the delta between skills workers have and the skills that are needed increases. This will manifest in lower productivity, challenges in hiring, and increasing unemployment. In fact, in a recent Deloitte survey of over 1,200 business leaders, sixty-one percent of them said technologies such as automation and AI that require new skills will be a primary driver of their organization adopting a skills-based approach.
So the barriers to organizational change are falling apart: Workers are demanding a skills-based approach to work and executives are increasingly prioritizing it. See more from that Deloitte survey below👇🏾
So what would a good skills engine have?
First party data with high signal: A proper skills engine can’t be powered by scraping online job descriptions, O*NET data, and pulling from linkedin. Or rely on self-reporting. There is little validity and signal in that data and, more often than not, it’s a lagging indicator for skills demand. The data likely needs to be first-party and/or tied to a true assessment of what someone can do, not a certificate of completion that will age over time.
In the flow of work: When someone takes an assessment at a point in time, it’s directionally helpful and may give granular enough data. But that data becomes dated pretty quickly. The most powerful engine will be integrated in every place where someone is showcasing their skills. This includes in meetings, coding environments, slack, emails, figma, and more. A reflection of someone’s skills will be real-time and less about whether they can answer a question on a test right and more about whether they can actually do the work.
Diversity in skill assessment: A skills engine can only be as powerful to a company as the amount of its workforce it can cover. Many assessments currently in market started initially with technical skills because those are arguably the easiest skills to assess (it’s more clear what is right or wrong). As the need for more human skills only increases in the age AI, the need to more reliably assess outside of technical skills is critical.
Adaptive and shared taxonomy: The challenge in companies today is that skills data sits in disparate systems and doesn’t speak or play nicely with one another. There is a need for a shared taxonomy that sits across a range of tools, but it also needs to smartly adapt as employees change and also the needs of the business change. And that taxonomy should sit across the entire talent management cycle from hiring to promotion to internal mobility, and beyond.
Leverages AI in assessing: AI is increasingly getting better at skills inference and building assessments. With the partnership of strong psychometricians and a strong taxonomy, AI can rapidly increase the velocity of a company in this category.
Worker ownership of data: Skills data on what an employee knows and how they’ve evolved should live with the employee over time. They should be able to use it to power their professional identity and also bring it with them to new employers. This is how the benefits to start translate outside of a traditional corporate employer to the broader labor workforce.
Start in companies and expand outwards to the labor market: After a certain amount of density, a powerful skills engine could share data publicly (aggregated and anonymized) that would be a lot more helpful to young people who are thinking about their career paths, learning solutions who are deciding what content to prioritize, and reskillers who are deciding where to focus their efforts.
This was a long way of saying that after months of sitting on this, I ended up writing it in a night because well… it’s skills, stupid (Referring to the famous “it’s the economy, stupid” phrase from James Carville, I promise i’m not calling anyone stupid). If you’re building something that sounds like this, ping me at jomayra@reachcapital.com!