AI skills assessment: stop guessing who can work with AI

You cannot hire for AI capability if you cannot measure it

Every hiring team is under pressure to find people who can actually work with AI. Most are failing at it in silence.

Not because they are not trying. Because the tools have not existed. A candidate writes "proficient with AI tools" on their CV. You ask a question in the interview. They answer fluently, confidently and in enough detail to sound credible. You move them forward.

Then they join. And you find out that sounding fluent about AI is not the same as being able to use it effectively under real conditions.

This is the dominant pattern in AI hiring right now. Self-reported confidence is the primary signal, and it is the weakest one available. The research on self-assessment is consistent on this point: a metasynthesis of 22 meta-analyses found that the average correlation between people's self-evaluations of their abilities and their measured performance is only .29. There is no standard for AI skills, no structured measurement, and no consistency across hiring decisions. Companies are making consequential choices about who can work with AI based on guesswork and gut feel.

That gap is now closed.

Introducing AI proficiency assessment in Maki

Maki can now test for AI proficiency. Our AI skills assessment puts candidates in realistic work scenarios and measures what they can actually do with AI, not what they claim.

The assessment is built on a validated four-dimensional framework, developed by Maki's science team using established psychometric processes. This is not a tool-name quiz or a knowledge checklist. It is a structured, evidence-based method for measuring the capabilities that actually matter when someone uses AI as part of their daily work.

It is available today for existing Shiro and Mochi customers, with no additional cost and no changes to your current workflows.

What is an AI skills assessment?

An AI skills assessment is a structured test of how well a candidate can use AI tools to do real work: choosing the right tool, prompting it, judging and refining its output, and knowing when not to use it. It is different from an AI literacy quiz, which checks what someone knows about AI, and from a tool checklist, which records which platforms they have tried. An AI proficiency assessment measures performance, not familiarity.

The four dimensions of AI proficiency at work

Most attempts to define "AI proficiency" land on a list of platforms. Have you used ChatGPT? Copilot? Gemini? That is tool familiarity, not capability, and it tells you almost nothing about performance.

Maki's framework identifies four dimensions of what it actually takes to work effectively with AI in a knowledge work context.

AI tool agility

How quickly and effectively someone can learn, adapt to and troubleshoot AI tools. The AI landscape changes fast. The people who perform well are not the ones who memorised a specific tool. They are the ones who can pick up a new one, work out what it can and cannot do, and get useful outputs from it quickly.

Human-tool interactivity

The quality of someone's collaboration with AI. Can they prompt well? Can they evaluate outputs critically and refine them? Can they integrate AI into a workflow in a way that actually improves the output, rather than just adding a step? This is where a significant portion of real-world AI performance lives, and it is almost entirely invisible in a traditional interview.

Ethical use of AI

Awareness of risk, fairness and responsible use, including the judgement to know when not to use AI at all. As AI-assisted decisions become more common across functions, this dimension is no longer optional. Candidates who lack it are a liability.

Data and algorithmic literacy

A baseline understanding of how AI systems work, what affects output quality, and how to recognise when something has gone wrong. This is not a technical qualification. It is the practical knowledge a knowledge worker needs to use AI outputs responsibly.

Together, these four dimensions give a picture of how someone operates with AI in practice, not just whether they have heard of it.

Three assessment methods, not one quiz

A single multiple-choice test cannot capture all four dimensions reliably. That is why Maki's AI skills assessment uses a multi-method approach.

Situational judgement tests present candidates with hypothetical work scenarios involving AI. They are asked to decide how they would use AI to solve a problem, troubleshoot an issue or improve an outcome. You see how they think and prioritise, not how they describe what they might do in the abstract.

Structured behavioural grids capture how candidates have actually applied AI in practice. These are behavioural questions with a structured response format, designed to surface real past experience rather than rehearsed answers. The signal is observable and comparable across candidates.

Conversational cognitive tasks, coming in a future Mochi release (Mochi 3.5), will go further. Open-ended, complex problems where you can see how someone approaches a real challenge using AI. The quality of their judgement, their approach to uncertainty, and their ability to integrate AI meaningfully into problem-solving will all be visible.

Each method maps deliberately to the four dimensions of the framework. The combination gives a more complete and reliable signal than any single format could provide. The same principle runs through Maki's wider AI assessment software: several validated methods beat one test.

Live now in Shiro and Mochi

Situational judgement tests and structured behavioural grids are available today inside Shiro and Mochi. If you are an existing customer, there is nothing to integrate, no new platform to onboard, and no change to your pricing.

Drop the assessments into existing role profiles and flows. The AI proficiency score sits in the candidate's profile alongside their other results, role requirements and full application record, visible at the moment a hiring decision is being made. Not in a separate platform. Not requiring an export. Where it actually matters.

Conversational cognitive tasks will be available in Mochi soon.

What this means for your hiring

If your team is currently assessing AI capability through CV bullets and interview impressions, you are operating on the weakest possible signal for one of the most consequential hiring criteria in the market right now.

Maki's AI proficiency assessment gives you something different: a structured, science-backed, observable measure of how candidates actually perform with AI at work. Built for graduates, knowledge workers and managers, the population that most hiring teams are actually trying to assess, and the population where a wrong call is most expensive in high-stakes specialist roles. Validated by a dedicated science team. And available today without disrupting a single thing in your existing process.

Hiring for AI capability without measuring it is not a gap in your process. It is a bet you are placing every time you make an offer.