What is the five-level proficiency scale?

Each capability is read on the same five levels: L1 Emerging, L2 Developing, L3 Proficient, L4 Advanced, L5 Strategic. The levels describe observable behaviour — for example, from accepting AI output uncritically (L1) to designing standards that guide responsible AI use across an organisation (L5).

Can your people actually work with AI? The 10 capabilities that decide it

← Knowledge Hub

Guide11 min read18 September 2026

Can your people actually work with AI? The 10 capabilities that decide it

Q: What are the AI-era capabilities that actually matter?

Working with AI is not one skill. GoMeasure's framework separates it into ten measurable capabilities: AI Work (productivity with AI), AI Oversight (supervising and owning AI work), AI Trust (calibrating when to trust or verify), AI Verification (catching AI errors), Judgment, Human Contribution (value beyond the model), Delegation, Problem Framing, Adaptive Reasoning and Decision Defence. Each is defined on a five-level scale from Emerging to Strategic.

Q: Why isn't 'can you use AI' a good enough measure?

Tool familiarity and course completion say nothing about whether a person supervises AI responsibly, catches its errors, or adds real value on top of it. Two people with identical AI access can produce very different quality and risk. The AI-era capabilities measure how well someone works with AI, not whether they have touched it.

Q: How do you measure these capabilities objectively?

Not with self-ratings or multiple-choice quizzes. GoMeasure measures them from demonstrated work — AI-enabled work simulations, adaptive AI interviews with identity assurance, and evidence from real tasks — and places each person on the five-level scale with the evidence attached. This is a framework and rubric for measurement, designed to be piloted and calibrated before high-stakes use.

AI has made output cheap and effort invisible. The people who win are the ones who can supervise it, verify it, judge with it and own the outcome. Here are the 10 AI-era capabilities that separate them — and how to measure each one.

Ankit KhandelwalFounder · GoMeasure AI

AI ReadinessSkills IntelligenceAssessmentFuture of WorkHR in the AI Era

In short: "Can you use AI?" is the wrong question. AI-era performance is a set of distinct, measurable capabilities — supervising AI, calibrating trust, catching its errors, framing the problem, making and defending the call, and adding value the model can't. This is GoMeasure's 10-capability framework, each read on the same five-level scale, so you can measure who can genuinely work with AI instead of guessing from résumés and tool familiarity.

Give ten people the same AI tools and you get ten very different results — different quality, different risk, different amounts of real human value added. The difference isn't access. It's capability. Yet most organisations still assess it with a binary: has this person used AI or not? That question can't tell the person who supervises a model and owns the outcome from the one who pastes its output unread. This is the operating problem behind our 16 HR struggles in the AI era — specifically measuring AI-era capability and performance that no longer reflects contribution.

Working with AI is not one skill

It decomposes into capabilities that can be defined, observed and scored independently. Some are about producing with AI; some are about controlling it; some are about the distinctly human judgement that decides whether the output is any good. GoMeasure organises them as ten AI-era capabilities — each expressed as a Quotient, our way of measuring and scoring that capability from demonstrated work.

The 10 AI-era capabilities

Capability What it measures Who relies on it

AI Work Quotient (AIWQ)Working productively and safely with AI — decomposition, prompting, orchestration, verification.Talent Acquisition, L&D, Digital Transformation

AI Oversight Quotient (AOQ)Supervising, verifying, challenging and remaining accountable for AI-assisted work.Governance, Risk, Compliance, Internal Audit

AI Trust Quotient (ATQ)Calibrating when to trust, verify, challenge or override AI — neither blind trust nor wasteful scepticism.Governance, Compliance, Quality Assurance, Risk

AI Verification Quotient (AVQ)Detecting errors, checking evidence and resisting hallucinations before they become business errors.QA, Internal Audit, Legal, Finance, Research

Judgment Quotient (JQ)Decision quality, trade-offs, ambiguity and ownership when evidence is incomplete or AI conflicts.Talent Management, Leadership, Risk, Strategy

Human Contribution Quotient (HCQ)The value a person adds beyond AI output — original framing, insight, correction, ownership.Talent Acquisition, Performance Management, L&D

Delegation Quotient (DQ)Deciding what AI should do versus what humans should retain, and where to draw the line.Digital Transformation, Process Excellence, Ops

Problem Framing Quotient (PFQ)Defining the right problem before solving it — the new bottleneck once solving is cheap.Product, Strategy, Consulting, Innovation

Adaptive Reasoning Quotient (ARQ)Revising strategy intelligently when assumptions or AI outputs fail.Change Management, Leadership, Project Management

Decision Defence Quotient (DDQ)Independently explaining and defending a final decision — without leaning on the AI.Leadership, Legal, Compliance, Finance, Audit

The six linked above are covered in depth in this series; Delegation, Problem Framing, Adaptive Reasoning and Decision Defence follow next.

One scale for all ten: Emerging to Strategic

Every capability is read on the same five levels, so a person's AI-era profile is comparable across dimensions and over time. The levels describe observable behaviour, not self-belief:

Level What it looks like (using AI Oversight as the example) What it means

L1 · EmergingAccepts AI output with weak supervision or ownership.Capability is present in name only; risk is high.

L2 · DevelopingPerforms basic checks, but applies the skill inconsistently.Reliable on easy cases, not on hard ones.

L3 · ProficientReliably verifies, challenges and owns routine AI-assisted work.Trustworthy in normal conditions — the working bar.

L4 · AdvancedHandles complex or high-risk AI work independently.Can be trusted with consequential, ambiguous work.

L5 · StrategicDesigns standards and guides responsible AI use across the organisation.Sets the bar for others; a multiplier.

Why measure capabilities, not tools

Access is now universal; capability is not. The differentiator has moved from whether people use AI to how well they control it.
Output hides the gap. AI lets a weak performer ship acceptable-looking work, so a normal review never sees that the judgement underneath is thin. Capability measures catch what output conceals.
Different roles need different capabilities. An auditor lives on Verification and Oversight; a strategist on Problem Framing and Judgment. The framework lets you require the right ones per role.
It makes development concrete. "Move from L2 to L4 on Verification" is a coachable, re-measurable goal in a way that "get better at AI" never is.

How GoMeasure measures them

Not with self-ratings or multiple-choice quizzes — those measure confidence and recall, not capability. GoMeasure reads each Quotient from demonstrated work: AI-enabled work simulations, adaptive AI interviews with identity and integrity assurance, and evidence from real tasks — then places the person on the five-level scale with the evidence attached in a Verified Passport. The result feeds hiring, deployment, development and governance decisions with proof rather than assumption. Explore the assessment portfolio to see how the capabilities map to roles.

An honest note on maturity: this is a capability framework and rubric — a structured way to define and measure AI-era work. The task families, level thresholds and decision uses are designed to be piloted, calibrated and reviewed for fairness before any high-stakes use, not treated as finished, norm-referenced instruments. We would rather be precise about that than over-claim.

Key takeaways

"Can you use AI?" can't distinguish real AI-era performance. Working with AI decomposes into ten measurable capabilities.
They span producing with AI (AI Work), controlling it (Oversight, Trust, Verification, Delegation) and the human judgement that decides quality (Judgment, Human Contribution, Problem Framing, Adaptive Reasoning, Decision Defence).
Each is read on one five-level scale — Emerging → Developing → Proficient → Advanced → Strategic — describing observable behaviour.
Measuring capabilities, not tool familiarity, catches what AI-inflated output hides and makes development concrete and re-measurable.
GoMeasure measures them from demonstrated work, with evidence attached — as a framework designed to be piloted and calibrated before high-stakes use.

Frequently asked questions

What are the AI-era capabilities that actually matter?

Ten measurable capabilities: AI Work, AI Oversight, AI Trust, AI Verification, Judgment, Human Contribution, Delegation, Problem Framing, Adaptive Reasoning and Decision Defence — each defined on a five-level scale from Emerging to Strategic.

Why isn't "can you use AI" a good enough measure?

Tool familiarity says nothing about whether a person supervises AI responsibly, catches its errors or adds real value on top of it. Two people with identical AI access can produce very different quality and risk.

How do you measure these capabilities objectively?

From demonstrated work — AI-enabled work simulations and adaptive AI interviews with identity assurance — placing each person on the five-level scale with the evidence attached, rather than relying on self-ratings.

Start with the six flagship capabilities below, or read the wider context in the 16 HR struggles in the AI era. To see how the capabilities map to roles, explore the assessment portfolio.

Get the frameworks