Frameworks & Templates4 min read

AI Skills for Employees: A Role-Based Skills Matrix

Build an AI skills matrix for employees that maps framing, creation, verification and decision skills to real role tasks and observable evidence.

Bokili EditorialPublished August 14, 2026· Verified August 14, 2026

A role-based matrix maps employee tasks across framing, creation, verification and decision skills.

An AI skills matrix for employees should answer a practical question: what must each person be able to do in their real work? A generic list of tools or prompt techniques cannot do that. The finance analyst checking a variance, the recruiter drafting a candidate message and the manager approving an AI-assisted decision need some shared foundations—but different proof of capability.

The most useful matrix links four core skills to representative tasks and observable evidence. It becomes a design tool for learning and development: teams can see what is essential for everyone, what changes by role and where practice must deepen before consequence or autonomy increases.

The rule

Map skills to work samples, not confidence. “Feels comfortable with AI” is not evidence that someone can protect data, verify a claim or recognise a decision boundary.

What AI skills for employees should the matrix include?

The four skills: FRAME, CREATE, VERIFY, DECIDE

FRAME the task

Choose an appropriate task, state the goal, provide relevant context and define the shape of a useful result.

CREATE with control

Use an approved tool and permitted input, guide the work in small steps and preserve the human contribution.

VERIFY the result

Check material facts, sources, calculations, completeness, bias signals and alignment with the original task.

DECIDE and escalate

Know what can be used, what needs review, what must stop and who has authority for the final decision.

Use three levels of observable capability

	Level 1 — Assisted	Level 2 — Independent	Level 3 — Workflow owner
Frame	Completes a task brief with guidance	Frames familiar tasks without help	Defines task boundaries and good examples for others
Create	Uses an approved template and sample	Adapts the method to routine work	Designs and tests the repeatable workflow
Verify	Follows a check list	Selects checks based on the output	Defines evidence, tests and quality thresholds
Decide	Recognises clear stop cases	Handles routine exceptions and escalates	Owns approvals, controls and workflow changes

These are capability levels, not seniority grades. A director may be Level 1 in a new tool. An experienced operations specialist may be Level 3 for one narrow workflow. Assign the level by task, consequence and ownership.

This fits the European Commission’s current AI-literacy guidance, which asks organisations to consider staff knowledge and experience, the systems and risks involved, and the context of use. The Commission also says differentiated learning approaches can be appropriate. NIST’s Generative AI Profile reinforces the need to connect capability with risk management, testing and human oversight.

Build the role-based AI skills matrix

Reading is a start. Practice makes it stick.

Start learning

From job title to work evidence

1
1. Choose one role family
Start with a group that already performs similar tasks. Avoid mapping the whole company in one workshop.
2
2. List five repeated tasks
Name outputs people recognise: a forecast note, supplier summary, hiring email, campaign brief or service response.
3
3. Mark AI suitability
Identify which tasks are approved for AI assistance, which need extra controls and which remain outside scope.
4
4. Apply the four skills
For each task, describe how the employee must frame, create, verify and decide.
5
5. Set the required level
Use Assisted, Independent or Workflow owner for each cell. Base the choice on work and consequence.
6
6. Write one work sample
Define a short task that makes the skill visible, including one uncertainty or failure to catch.
7
7. Review gaps quarterly
Update the matrix when roles, tools, policies or important workflows change.

Worked example: three roles, one shared foundation

	Representative task	Proof of capability
HR coordinator	Draft an interview invitation from approved details	Uses permitted data, checks names and dates, and keeps the hiring decision outside the tool
Sales representative	Prepare a first-pass account brief from approved sources	Separates source facts from hypotheses and cites decisive evidence
Finance analyst	Explain a budget variance from a checked table	Verifies calculations, labels assumptions and routes material exceptions to the owner
Shared foundation	Use an approved assistant for routine work	Protect inputs, frame the task, check material claims and know the escalation path

The matrix now tells L&D what to teach and managers what to observe. It also prevents a familiar mistake: giving everyone the same feature tour, then expecting role-specific performance to emerge on its own.

Turn the matrix into a learning path

Matrix design checklist

Every cell describes an action a person can demonstrate.
Shared foundation skills are separated from role-specific practice.
Approved tools and data boundaries are named.
Required levels reflect task consequence, not hierarchy.
Each role has at least one realistic work sample.
Verification and escalation are taught alongside creation.
Managers know what evidence shows readiness.
The matrix has an owner and a review date.

Map one role in ten minutes

Choose one role and one repeated task.
Write one observable behaviour for FRAME, CREATE, VERIFY and DECIDE.
Assign Assisted, Independent or Workflow owner to each skill.
Add one mistake or uncertainty the work sample must expose.
Name the manager or expert who will review the result.

Use the matrix alongside related Bokili Learn guides: design depth with https://bokili.com/en/learn/risk-based-ai-literacy-training, measure behaviour with https://bokili.com/en/learn/measure-ai-training-beyond-completion, and support local practice through https://bokili.com/en/learn/ai-champions-network-support-bottleneck.

A strong matrix is small enough to use and specific enough to change learning decisions. Start with one role, one task and four skills. Expand only after the work sample shows what the labels really mean.

Bokili turns matrix cells into short, role- and tool-specific missions with practice and feedback. Explore practical AI skills for real work at https://bokili.com/en.