Every Workforce Plan Is a Bet on Skills. But What Can Your People Actually Do?

The supply side of strategic workforce planning has been running on inference.

Strategic workforce planning has two halves. Demand asks which roles and skills the business strategy will require, at what volume, where, and over what horizon. Supply asks what talent and skills we will have when we get there.

Over the past few years, the demand side has been rapidly maturing with significant advances in key capabilities like demand drivers, scenario modelling, and the breadth and depth of external labor market signals. And with the upcoming release of our demand driver agent, this work is going to get even easier. We can now estimate not only which skills a strategy will need, but how much of each, in which roles, in which locations, across a 3-, 5-, 7- or even 10-year horizon.

The supply side has not kept pace. Most enterprise skills profiles are assembled from CVs, LinkedIn profiles, job titles, and keyword extraction. Useful signals for sure, and as a broad mechanism to understand the shape of skill supply in the business, it’s been an adequate solution, just not a great solution. The problem is that these approaches describe what someone claims or what a model infers based on keyword matching.

At a deep technological level, all these approaches are based on linguistic analysis of the skill expressed as words – whether it’s in a CV or a work artifact. None of these approaches are based on what someone can demonstrate or any kind of psychometrically valid assessment of work outputs.  Which means every solution today relies on some level of human validation – employee or manager – to confirm that the inference got it right.

The problem with that is we think humans are adding value in this equation.  At two different clients over the last few weeks, I’ve heard an almost identical concern:

“We’re basing critical workforce decisions on what skills we have, and we have zero confidence in our data – we say it’s validated by managers, but let’s be honest, most managers have zero clue what skills their employees have so they just rubber stamp whatever the employee claims. Which means we’re ultimately reliant on the worker’s self-attestation – which works in some parts of the world, but not in others; in some roles, but not in others; in some people, but not in others.  So what we know about our skills, like really know – with confidence and some level of defensibility – is basically nothing.”

Skill and task live at the heart of future workforce planning – it’s a critical data layer to get right, but even the company that claims to be the “data layer for the AI era of work” is basically just doing a fancy version of inference and still relying on human validation.
TechWolf asks whether a person’s HR and work records are semantically linked to a skill, and estimates a proficiency level from how recently and how often that work appears – that’s an experience proxy, not an actual evaluation of demonstrated skill, and technologically, it’s not dissimilar to how every labor market intelligence vendors extracts skills from job postings.

While all of this is useful, we think skills data is too critical to rely on inference – in any of its forms. What is required for truly defensible outcomes is “planning grade skills data.”  Skills evidence strong enough that you would use it to make critical workforce strategy decisions on it, argue for a 3x increase to your reskilling budgets, or base a multi-million-dollar site-location decision using your skills data as a base. Most organizations have skills data. Far fewer have skills data that is planning grade.

Which is why we’re more deeply partnering to bring CodeSignal to our workforce planning clients.  Earlier this year, we launched our partnership with the ability to link their skills validation and skill assessment outcomes to our skills intelligence engine, enabling clients to get defensible skills intelligence via objective assessments of demonstrated skills.  

CodeSignal has now applied this same assessment rigor to the work artifacts that an employee produces.  Unlike TechWolf’s inference strategy, CodeSignal evaluates actual performance: AI agents score actual work artifacts (code, sales calls, documents, AI-tool usage) against structured rubrics grounded in industrial-organizational (I/O) psychology, anchored by its validated assessment and simulation business.  In other words, validated assessments against known rubrics of what “good” looks like vs. “Johnny seems to be involved in this kind of work, does it a lot, and therefore is probably pretty good at it.”

What CodeSignal just made possible

On October 28, CodeSignal announced the general availability of Agentic Skills Intelligence. The premise is simple and powerful: evaluate skills from the work people already do.

Specialized agents assess work artifacts from the systems employees use every day, including Gong, Salesforce, GitHub, Cursor, Claude, ChatGPT, Workday, Greenhouse and Google Workspace. That work is scored against structured rubrics grounded in industrial-organizational psychology. The output is a living skills profile, updated as new work is produced, with a view of proficiency and not just presence.

What makes this credible is the foundation underneath it. CodeSignal has spent more than a decade building simulation-based assessments, and its platform has completed more than 21 million skill evaluations. That same assessment science is now being applied to real work. It is a meaningfully different starting point than asking a general-purpose model to infer proficiency from a set of words.

CodeSignal’s research shows that 6 in 10 managers say they make critical talent decisions based on guesswork. More than two-thirds of employees say their skills have never been formally assessed. This tracks with every conversation I’ve had in the last two years with leaders struggling with the quality of their skills data.

Alongside work-artifact evaluation, CodeSignal continues to offer the two capabilities that built its reputation: rigorous skills validation through assessments and simulations, and personalized practice through Cosmo, its AI tutor.  The former remains critical in situations where agentic skills detection may hit roadblocks or limitations – Worker’s Council approval challenges, data privacy restrictions, or skills that are harder to measure directly via digital work artifacts alone – for example judgement, discernment, or coaching.

Until this latest release from CodeSignal, there was no practical way to observe proficiency at scale without stopping people from working to test them. Inference was the best tool we had. The opportunity now is to stop treating it as the only one, and leverage a more advanced, defensible approach.

Four kinds of skills evidence, one view of the workforce

The real power of this partnership is not any single capability. It is what happens when four distinct kinds of skills evidence sit on the same foundation, with the same role, task and skill definitions.

1. External evidence: what the market says. This is where TalentNeuron has always been strong. We see how skills are evolving within roles, which are rising or falling in demand, and how skills cluster and connect through adjacency. We can see which skills competitors are hiring for, and how their definition of a role differs from yours. Our skill evolution framework, combined with the human insight of your own leaders, lets you build a future job profile, not just a current one.

2. Declared and inferred evidence: what people and their records say. Our platform already builds employee skills profiles by extracting skills from CVs, LinkedIn profiles, project documentation and other assets. Employees can refine their profiles with adjacency-based recommendations, and add skills built outside their current role, including through volunteer work and other experiences. For workers whose work rarely leaves a digital trail, people can enter skills directly and managers can validate them.

3. Demonstrated evidence: what people show under structured conditions. CodeSignal’s psychometrically grounded assessments and simulations measure proficiency directly, with a rigor that self-reporting cannot match.

4. Evidence in the work: what people actually do every day. CodeSignal’s agentic evaluation of real work artifacts adds a continuous signal of which skills are being applied, and how well.

Each layer has blind spots on its own. Together, they move a skills profile from claimed toward proven. And because they connect to the market view in layer one, you can see not only what your people can do today, but how that compares to where the market and your competitors are heading.

Intelligence delivers value when it influences decision-making

A skills profile is an input. Skill-driven talent decisions are the outputs. Skills intelligence that only lives as a data layer can be a good or bad thing depending on the sophistication, staffing, and bandwidth of your HR Tech team. We act as that data layer for some very large, very technically mature teams. They take our skills data and plug it into every system where they need it: HRIS, Learning Platform, SWP engine, Talent Marketplace, Org Design, Sourcing, Applicant Tracking, Onboarding, Performance…

But that’s not the norm, and most HR teams have no idea what they are signing up for when they agree to take in a “skills data layer.” For most teams, the result is a lot of skills intelligence, no idea what to do with it, no idea how to drive ROI, and no capability to make it actionable in their systems or record or systems of execution.  There is a growing chorus of “great, we know stuff about what skills we have, now what?”

Our view has always been that intelligence should not stop at a data layer. We do the heavy work of making it actionable. We don’t just live in the ivory tower of a clean data layer, we walk into the weeds with you – hand in hand – to get the actual real work done. Skills evidence must be put to work across the decisions that matter. Fundamentally, these are:

• Strategic workforce planning - Supply forecasts grounded in validated proficiency, so the gap between future demand and forecasted supply reflects what your people can actually do.

• Skill gap analysis and gap filling recommendations - Gaps identified at the skill level against current state and across 3- to 10-year horizons, then matched to the right response: internal mobility, reskilling, external hiring or some combination of all three.

• Reskilling and upskilling - Adjacency models that show which people are closest to a future skill, or already have it, so reskilling targets the right people first; support for personalized learning plans against current skill gaps or gaps linked to future role aspirations.

• Job and work architecture - Role, task and skill definitions kept current with evidence from the work itself, not just job descriptions, and from market and competitor’s changing expectations for roles and skills, including nuanced insights like “skill worth” in any given job profile.

• Competitive analysis - Your validated internal skills set against how competitors or exemplars define the same roles and where and at what rate they are hiring for them; compare to specific named business competitors, talent competitors, or top-tier firms outside your space that you want to emulate.

• Labor market analysis – Skill evolution over time and velocity of skill change, skill adjacency models, and differences in skill availability across locations and roles, supported by trend data.  

• Hiring and location strategy - Where to hire, or where to place a new plant or center of excellence, based on where the skills you need are available; hiring difficulty scores based on skill expectations as a driver of supply and demand estimates and competitors hiring for similar profiles.

This is the critical “now what?” work that needs to happen once you have a skills intelligence data layer. Our TalentNeuron platform can help you deliver on any or all of these critical next steps. We believe that evidence of what your people can do becomes far more valuable when it sits beside evidence of what your strategy will require and what the market can supply. And more valuable still when you can actually put that intelligence to work in all the critical places above without spending another 6-18 months getting it all to talk to each other.  

From claimed skills to proven skills

We are proud to partner with CodeSignal. Their assessment science and agentic work evaluation, joined with TalentNeuron’s labor market intelligence, work architecture and strategic workforce planning, give our clients something they have long needed: an employee skills intelligence foundation they can confidently and defensibly plan on, and the help to turn it into outcomes for the business.

If you’d like to see what planning-grade skills data looks like inside a real workforce plan, we’d be glad to show you.