With the demand for talent development increasing, leaders are expected but often struggle to develop employee skills and stay updated on the relevant content needed to upskill. But thanks to artificial intelligence (AI), they now have a tool that empowers them to be effective talent managers.

Integrating large language models into skills development software can result in remarkable improvements for measuring skills, offering insights and curating the best content available. It’s a game-changer that enables managers to excel in their role and nurture their team’s potential by offering effective mentorship — while saving business leaders time, money and effort.

The Importance of Mentorship

I’ve co-created and taught AI classes at Stanford University and online to millions of students. One day a student said to me: “The gap between me and a student from a top university like Stanford isn’t about course content, which I can access online. It’s about mentorship. I can’t gauge my skills against a Google engineer, a Meta product manager or a Silicon Valley startup founder because I don’t have anyone around me to offer advice, support and insight. Without anyone to guide me through this kind of career, I feel lost.” In other words, content wasn’t the limiting factor to develop a career anymore — it was mentorship.

But mentorship is scarce. Students at top universities often have direct access to mentorship through their professors and peers, so they’re able to know what skills they need to develop and how their skill development compares to others. But most people — particularly working professionals — don’t benefit from the same guidance, especially in cutting-edge domains like AI.

Scaling Mentorship With Technology

Mentorship today is like a rare gem — those who find a mentor consider themselves fortunate. But it shouldn’t be this way. Mentorship should be as available as the air we breathe, and accessible to all. This is where the innovation of AI mentorship makes such a difference.

Imagine AI mentorship as a grand machine, needing fine-tuning to scale up. The first cog in this machine is understanding — a mentor can’t guide a journey without knowing the traveler’s current location and destination. So mentorship begins with a map and compass, a skills assessment and a clear set of goals.

The second cog is expert guidance. A seasoned mentor, like a wise navigator, uses knowledge from previous journeys to spot patterns and suggest routes for new travelers. This is much like a recommender system in the world of computer science. We’ve built powerful recommender systems to serve us new movies or ads we might like to watch; the same algorithms can also be used for learning recommendations.

However, we’re still fumbling with a lantern when it comes to measuring human capabilities. If we want to make great mentorship commonplace, our first task is to build a better lantern: to develop improved assessments and clearer goal-setting methods, and to make sure a mentor can assess their learner as often as possible — perhaps, even constantly — to ensure that every suggestion they make will be more accurate than the last.

Using the Power of AI to Reimagine Personalized Learning

Assessments have been used throughout history, but today for the first time we have the opportunity to transform them on a conceptual level. Let me explain.

First, assessments were historically administered verbally or on paper, rather than digitally. Data wasn’t collected and tracked. We were not able to identify patterns from thousands of people taking tests to improve the assessment and make it adaptive. Today, we’re analyzing answers from all around the world to improve how we assess one person. Using those patterns, we are able to measure someone on a handful of skills and predict hundreds — without measuring them directly. In simple terms, if we know someone can solve a mathematical equation, like 2 x 2 = 4, we can predict that they can also solve 2 + 2 = 4 and 2 – 2 = 0. This means we can begin to ask them more difficult questions, because we understand the patterns of learning and cognition.

Second, assessments have often been used for potentially stressful, high-stakes situations — for example, in job selection processes — rather than for upskilling and mentorship. Today, we have the opportunity to reinvent and promote assessment for formative usage. The goal is not to pass or fail someone, but instead to identify their skill strengths and skills gaps, and then suggest ways to close these gaps. Of course, this means the entire interface of assessments needs to change: They should be empowering, flexible and cutting-edge, rather than serious, traditional and rigid.

Beyond assessments, we also have the opportunity to define goals like never before, through skills ontologies. Skills ontologies are a must-have in most large, contemporary companies, becoming the de facto vocabulary and communication system when discussing skills. They’re a categorization of skills that builds a common language of skills that define certain aspects of the job. They need to be structured and flexible at the same time: structured, because most of us (across companies, countries, etc.) use common, similar or shared skills.

For instance, it’s possible that 70% to 80% of the skills of software engineers in two separate companies are, in fact, the same — with room for flexibility in the remaining 20% to 30% to account for differences in factors like project, industry, language and culture.

Consider this example:

The process of rice cultivation across two countries, such as India and Italy, is likely to be similar — the skills of those people cultivating it is, therefore, likely to be similar. But the rice from India is Basmati, a longer variety, whilst a rice from Italy, such as Arborio, is shorter and rounder. The differences in cultivating and cooking these two rice types will be reflected in the cultivating and cooking skills of those who cultivate and cook them.

While they might find generic skills in a structured skills ontology, organizations need to add new skills to the ontology to cover 100% of what they need. Those skills would also need to be measurable and tied to learning content.

In the past, creating new skills, assessments and learning content would have been virtually impossible without the input of many psychometricians, assessment designers and learning content creators, since the volume and scope of material needed to establish a comprehensive and valuable learning experience would be enormous.

Today, generative AI is capable of taking the bulk of that work and infusing it with information from across the internet, assisting experts in creating skills ontologies, assessment and learning content that’s both psychometrically and educationally sound, and hyper-specific to the individual users based on a variety of factors including, but not limited to, their role, skills, proficiencies, industry and so on.

The Future of Skills Development Rests in AI’s Hands

AI’s productivity boost will alter job demand, while simultaneously creating and defining new roles, propelled by emerging regulations. As we move toward the largest career shift in human history, the scale of transition will necessitate AI-based mentorship.

With increasing data, AI systems harness global expert knowledge, tailoring it to individuals’ unique skills and needs. As digitalization permeates our work, AI’s understanding of career progression and skill acquisition grows, leading to refined skills assessment.