Has AI Fundamentally Changed Skill Gap Analysis?
How artificial intelligence can spark a new plan of attack to solve the skill gap problem in an innovative way.
"If I had asked people what they wanted, they would have said faster horses."
Henry Ford's apocryphal quote not only tells us something about innovation, but it also shows a surprising parallel to one of the most fundamental challenges in HR - closing skill gaps. Skill gaps today are more complex than ever, mainly driven by the rate at which they transform. New skills enter the market, old ones become irrelevant, and people remain in the job market with an outdated skill set. There have been attempts to identify and close these gaps using surveys and skill inventories, but these no longer do.
With an ever-changing skill landscape, HR today doesn't require a faster horse, but a car - an entirely new way of solving the skill gap problem.
The Old Way of Skill Gap Analysis
While there are many ways to map skill gaps in an organisation, most of them share a few key components. Typically a different flavour of the following steps [1, 2]:
- Create an inventory of skills for each role
- Create an inventory of skills for each employee
- Perform skills-gap analysis
While this plan of attack might look straight-forward at first, it involves substantial complexity. As the organisation gets larger, this complexity grows exponentially. As long as the workforce is relatively small, these approaches make sense and are sufficient. However, this doesn’t scale well past a couple of hundred employees, as these manual methods are slow and error-prone.
Mapping Skills to Roles
Before the mapping exercise can start, a listing of all roles in the organisation is needed. Here's where the trouble often starts. Job titles are scattered across many systems, usually containing noise, paraphrases and plain duplicates. Before you can even begin with any skill mapping, you need to get these duplicates out of the way, usually through a tedious manual task of mapping the required skills.
Obtaining a skills-mapping for roles is typically done by interviewing individual contributors and line-managers. As people in the field know best what it takes to be successful on the job, interviewing them is a reliable way to figure out the associated skills. However, this approach only works for the roles you have today, not the roles you need to succeed in the future. As the number of roles increases, these interviews stand in the way of an efficient process and don’t even cover the full job-spectrum.
Interviewing individual contributors and line-managers to map skill to roles only works for the roles you have today, not the roles you need to succeed in the future.
Our customers report that these interviewing processes might take anywhere from months to years, with multiple roles being outdated by the time of completion. In a world with a skill half-life of fewer than five years, an approach solely based on interviewing may set you up with a disadvantage at the start.
Mapping Skills to Employees
After mapping skills to roles, and defining the skills needed to be successful, you need to know your employees’ skills. This next step is even more daunting than the previous one, as the quality of employee skill data is too low. It is too generic, outdated, or there's simply no data at all. Many HCM platforms today have functionality for skill profiles, but 75-80% of employees don’t complete their talent profile.
Other sources of employee skill data mainly revolve around surveying the workforce. Apart from survey fatigue, surveys can let us fall prey to recency bias. HR leaders often tell us that employees tend to report skills they’ve been using recently, and thus, the survey misses out on the skills that employees haven’t been using and might have value too.
Not only is this step handicapped by insufficient data, enriching it with survey-data doesn't necessarily give additional benefits. When detecting skill gaps, knowing where you stand today is crucial: if the AS-IS data is incomplete, how can you expect to find a way to the desired TO-BE situation?
Performing Gap Analysis
In an ideal world, each person’s skillset would be aligned precisely with their current position. This often isn’t the case, and this gap gets more significant over time. In a knowledge economy, knowing and closing these gaps quickly can mean the difference between winning or losing the race for sustainable growth.
To determine this gap, you compare an individual’s skills to the skills required by a job. This job can be their current occupation or another future role, but the concept remains the same; identifying which skills are lacking. The main issue with this paradigm is that people tend to look at it in a very binary way: you either have a skill or you don’t, with no other options in between.
This binary approach wouldn’t be a big issue if there were a universal way of describing a skill set. This isn’t the case. Try asking two co-workers with the same job to assign skills to their role, and you will get vastly different results. In literature, this is called: 'inter-annotator agreement', and when describing skill sets, this agreement is very low.
This phenomenon is one of the root drivers behind the obsolescence of 'binary' gap analysis. The presence or absence of a particular skill might be due to a difference in wording. Without profound background knowledge of adjacent skills, manual gap detection might even assign employees the wrong skill gaps.
The Deeper Issue
While all three steps have their respective inefficiencies with an increasing headcount, the approach itself, traditionally accompanied by the excel sheet, is flawed as well. They all rely on manual input or subjective data. The bottleneck of the entire process is, in fact, manual labour. This makes it slow and reduces its frequency to yearly or three-yearly cycles. Whereas in today’s ever rapidly changing knowledge economy, keeping your competitive advantage means continuously monitoring your weaknesses; your skill gaps.
The New Way of Skill Gap Analysis
Today, artificial intelligence is adopted at an increasing rate throughout the organisation to optimise processes and reduce administrative overhead. While AI has seen success in talent acquisition and powers some newer HR-tech platforms, it is still not applied enough for strategic HR decision support. Let's revisit the steps we discussed in the traditional approach and see where they can be improved.
Mapping Skills to Roles and Employees
Just as AI can look at a picture and tell what is in it, it can also take a look at positions and job titles and tell you what underlying occupation can be recognised. This removes the need for long survey campaigns to a straightforward way of knowing what your organisation’s roles entail.
Employees generate hundreds of data points every day, creating a wealth of information that can be used for skill inference. As some HCM vendors started doing already, you can combine structured and unstructured data into a skill profile. This approach’s main advantages are that AI (1) can surface hidden skills or skills you might have forgotten you have (2) the sheer amount of data that can be considered is a lot higher.
Performing Skill Gap Analysis
Given the thousands of skills and millions of synonyms, it's not easy to determine an accurate skill gap: is a skill lacking or simply mentioned differently? Without full knowledge of skill-adjacencies - how skills relate, overlap, and influence each other - it’s hard to determine a gap with certainty. It’s even harder to do this at scale, and that’s where AI comes in.
The main advantage of using AI for gap analysis is there are no more trade-offs. Gaps can be determined on the employee-level and with enormous granularity. Each employee brings a different set of skills to the table, even when they temporarily occupy the same function. Using AI means that we’ll use everyone’s unique context and skills to look at gaps with specific roles. This unleashes hyper-personalised career-guidance for every employee, without the associated costs. Hyper-personalisation means that L&D resources can be spent more efficiently, straying away from one-size-fits-all approaches.
As gap analyses in the traditional sense are expensive and time-consuming, their frequency is limited, only done when it has to be. Given a more cost-efficient way to analyse gaps, the frequency can be increased as well. With AI, it’s possible to do continuous gap analysis, taking into account all the changes in your organisational landscape, and resulting in an up-to-date view at all times.
This doesn’t mean the human aspect is lost in any sense; one could even argue that it is improved. With individual gap data present, HR can help close these gaps more effectively, even proactively.
Artificial Intelligence to Tackle Skill Transformation
Going back to our horse and car analogy; the horse no longer does the job. Manually detecting skill gaps for a large organisation is an error-prone process not suited for today's changing nature of work. A faster horse will no longer do: instead of marginal improvements on the traditional approach, we need a paradigm shift to tackle the skill transformations in the next years.
In this sense, AI is the proverbial car suited to get us to our destination: a future-proof workforce. Using AI doesn't just improve the process; it expands and enriches it without sacrificing granularity. Ultimately, this leads to a better employee experience by providing better personalisation, taking into account individual capabilities holistically. For the organisation, this means better alignment and a workforce ready to tackle the future, head-on.