Tigran Sloyan is CEO and co-founder of CodeSignal. He answered HRTech Edition’s questions on skills evidence, measuring AI use, AI training and fairness in AI evaluation.
By Tigran Sloyan, CEO of CodeSignal
1. Why do organizations still struggle to identify employees’ actual skills, despite having more workforce data than ever?
More data doesn’t necessarily mean better evidence. A job title tells you what someone was hired to do. A completed course tells you what they studied. Neither tells you how well they can do a particular task.
That distinction matters when you’re hiring, promoting someone, or putting together a project team. Our research found that 63% of managers made workforce decisions based on assumptions rather than reliable data in the past 12 months, and 86% said candidates performed differently on the job than their resumes or interviews suggested.
AI makes this more urgent. The tasks people do are changing quickly, so even an accurate picture of someone’s skills a few years ago may not tell you what they can do today. Organizations need evidence from the work people actually perform.
2. How can HR leaders move beyond resumes, performance reviews, and self-assessments to make more reliable talent decisions?
I recommend HR leaders become masters of ‘skillspotting,’ which is the practice of surfacing skills employees already have within an organization, including capabilities that may fall outside their formal job descriptions. If you only look at titles or self-reported profiles, you can miss someone who is already capable of contributing to a new team or project.
Start by defining the skill you need to identify. What task does someone need to perform, and what would good performance look like? Then give them a chance to demonstrate it. For employees already doing the work, observe the work itself.
What does observing the work itself mean? It depends on the job, but that could mean looking at a customer interaction, a piece of code, or a completed project against a clear, job-relevant rubric. And don’t stop at asking whether someone has a skill or not. It’s most important that you understand their level of proficiency.
3. With AI investments rising but measurable returns remaining limited, how can organizations determine whether employees are actually becoming more effective with AI?
Counting AI licenses, token usage, or course completions can’t tell you whether they’re using AI well.
Can they provide the right context, delegate tasks appropriately, evaluate the output, and recognize when it’s wrong? Working with AI is a skill in itself, and employees need to develop the judgment to use these tools effectively.
Our research found that 84% of organizations expanded their AI software investments and 79% increased AI training over the past year, yet 79% said they couldn’t measure the ROI of that spending. That points to a major visibility problem.
As AI takes on tasks like writing SQL or visualization code, the human skills around understanding requirements, interpreting results, and exercising judgment become even more important.
4. Why does corporate AI training often fail to translate into everyday workplace performance, and what needs to change?
Our research found that, among employees who received employer-provided AI training, only 26% said they could apply most or all of it in their daily work.
That’s because you can’t build skill by watching a video. Skill is built through practice, feedback, and trying again. AI can make that practice much more available and interactive. For example, someone may need a conversation partner to practice giving difficult feedback or managing conflict. AI role-play conversations can give employees a low-stakes opportunity to practice these skills before using them in a real work setting.
But AI shouldn’t do the learning for you. If the goal is to develop judgment, employees still need to work through problems themselves and demonstrate what they’ve learned.
5. As AI reshapes job roles, how can organizations identify emerging skills gaps before they affect business outcomes?
Look at how the actual work, and the tasks associated with that work, are changing. A data scientist may still be expected to turn data into useful insights, but AI changes which tasks they do themselves and which skills they need to do the job well.
Organizations then need to compare those requirements with what employees can actually demonstrate. Skills intelligence is designed to evaluate skills in the context of real work and identify proficiency and gaps. Within CodeSignal’s platform, our AI coach Cosmo can then offer targeted practice to help employees develop the skills they need.
The aim with agentic skills intelligence is to find those gaps early enough to do something useful about them before it becomes a problem.
6. How can companies use skills intelligence to make better decisions about internal mobility, reskilling, and hiring?
When looking at internal mobility, skills intelligence helps companies identify employees who have the skills that are needed for another role, rather than having to source externally. In the same realm, it allows organizations to bring people together internally with the right skills for specific projects. For reskilling, organizations can identify specific skills gaps and give employees personalized opportunities to practice and develop those skills. When it comes to hiring, companies can focus on demonstrated capabilities rather than relying solely on resumes, job titles, or other proxies.
Better skills information gives organizations a clearer view of the talent they already have and gives people more ways to demonstrate what they could do next.
7. As organizations increasingly use AI to evaluate employee capabilities, how should they address concerns around privacy, transparency, and fairness?
The way an AI evaluation system is designed matters. The starting point is a clear, job-related rubric. As for CodeSignal’s Agentic Skills Intelligence, we score work against structured rubrics grounded in industrial-organizational psychology, the same scientific rigor behind the more than 21 million skill evaluations completed on CodeSignal.
There is also a question of what AI can and cannot capture. A great manager may pick up on things that a rubric doesn’t account for, which is why human oversight remains important. The goal is never to remove humans from important talent and workforce decisions. It’s to give them better information so they can focus on the decisions that require their judgment.
