How to Build an AI Competency Framework for Your Organisation

Most organisations know they need their people to develop AI skills. Far fewer know exactly which skills, to what level, for which roles, and in what order. The result is scattered training efforts — some teams get workshops, others get LinkedIn Learning licences, a few enthusiasts teach themselves, and the organisation has no clear picture of where it stands or where it needs to go.

An AI competency framework solves this problem. It defines what AI skills your organisation needs, maps them to specific roles, creates clear levels of proficiency, and gives you a structured way to assess where people are and close the gaps.

This is not an academic exercise. A well-built framework becomes the foundation for every training investment, every hiring decision related to AI skills, and every conversation about AI readiness at the leadership level. Without one, you are investing in AI training without knowing what you are buying or whether it is working.

This guide walks through how to build one, step by step.

What Is an AI Competency Framework?

An AI competency framework is a structured document that defines:

• The AI competencies your organisation needs (the specific skills and knowledge areas)
• The proficiency levels for each competency (from beginner to expert)
• The role mapping that connects competencies to specific jobs (what each role needs to know)
• The assessment criteria for measuring where people currently stand
• The development pathways for closing the gaps between current state and target state

Think of it as a skills map. It shows you the territory — what skills exist, who needs them, to what level — and gives you a way to navigate from where you are to where you need to be.

The framework is not a training programme. It is the strategic document that tells you what training programmes to build, who to enrol in them, and how to measure whether they are working.

The Six Core AI Competencies

After working with organisations across multiple industries, we have found that AI competency breaks down into six core areas. Most roles need some combination of these, but not all of them, and not to the same depth.

Competency 1: AI Literacy

Understanding what AI is, what it can do, what it cannot do, and how it is relevant to the organisation and the individual's role. This includes knowing the major categories of AI tools, understanding basic concepts like prompting and output evaluation, and being able to have informed conversations about AI's role in the business.

Who needs this: Everyone in the organisation.

Competency 2: Prompt Engineering

The ability to communicate effectively with AI tools to get useful, accurate, and well-formatted outputs. This includes writing clear prompts with context, task, and format specifications; iterating on outputs; using personas and examples; and chaining prompts for multi-step tasks.

Who needs this: Anyone who uses AI tools directly, which should eventually be most of the organisation.

Competency 3: AI-Assisted Content Creation

Using AI to draft, edit, refine, and produce written content, visual assets, presentations, and other deliverables. This goes beyond basic prompting into the editorial layer — knowing how to evaluate AI-generated content, maintain quality and voice, and use AI as a force multiplier rather than a replacement for human judgement.

Who needs this: Marketing, communications, HR, sales, consulting, and any role that produces written or visual deliverables regularly.

Competency 4: AI-Assisted Analysis

Using AI to process, analyse, and extract insights from data, documents, and research. This includes uploading data to AI tools for analysis, asking the right analytical questions, interpreting AI-generated insights critically, and using AI for research and competitive intelligence.

Who needs this: Finance, operations, strategy, research, sales, and any role that works with data or needs to synthesise large amounts of information.

Competency 5: Workflow Automation

The ability to identify repetitive processes that can be automated, design automation workflows using no-code or low-code tools, and maintain those automations over time. This includes familiarity with tools like Make.com, Zapier, or n8n, and the ability to think in terms of triggers, actions, and integrations.

Who needs this: Operations, project management, IT, and identified "automation champions" in each department.

Competency 6: AI Strategy and Governance

The ability to evaluate AI tools and opportunities, assess risks, make decisions about AI adoption, develop AI policies, and lead AI transformation within a team or function. This includes understanding data privacy implications, ethical considerations, vendor evaluation, and change management for AI adoption.

Who needs this: Leadership, IT, compliance, and department heads.

The Four Levels of AI Competency

Each competency area should be assessed and developed across four progressive levels. Here is what each level means in practice:

Level 1: Awareness

Definition: The person understands the concept and can explain its relevance to their role. They have seen it in action but have not practiced it independently.

What it looks like: "I know what prompt engineering is. I have seen someone use it. I understand why it matters for my work. I have not done it myself, or I have tried once or twice with mixed results."

Assessment criteria: Can explain the concept in their own words. Can identify at least two ways it applies to their role. Has seen a demonstration or worked example.

Level 2: Foundational

Definition: The person can perform basic tasks in this area with guidance or reference materials. They use it occasionally and can produce acceptable results with effort.

What it looks like: "I can write a decent prompt if I think about it carefully. I sometimes use AI for my work, usually for straightforward tasks. I still need to look things up or ask for help with anything complex."

Assessment criteria: Can perform basic tasks independently. Uses it at least weekly. Can produce acceptable output for standard tasks. Still requires support for complex or novel situations.

Level 3: Proficient

Definition: The person uses this competency regularly and effectively as part of their normal workflow. They can handle complex situations, adapt their approach, and produce high-quality results consistently.

What it looks like: "AI is part of how I work every day. I can handle complex tasks, iterate effectively, and get strong results quickly. I can troubleshoot when something does not work and adapt my approach."

Assessment criteria: Uses it daily as part of normal workflow. Can handle complex and novel situations. Produces consistently high-quality output. Can explain their approach and reasoning to others.

Level 4: Expert

Definition: The person can teach others, develop best practices, evaluate new approaches, and innovate in this area. They are a go-to resource for their team or organisation.

What it looks like: "I teach others how to do this. I develop our team's best practices. I evaluate new tools and techniques. I find novel applications that others have not considered."

Assessment criteria: Can teach the competency to others effectively. Develops best practices and templates for the team. Evaluates and recommends new tools and techniques. Identifies and implements innovative applications.

Mapping Competencies to Roles

This is where the framework becomes specific to your organisation. For each role or role family, define which competencies are relevant and what level is the target.

Here is a sample mapping for a mid-sized organisation:

All Employees (Every Role)

• AI Literacy: Level 2 (Foundational)
• Prompt Engineering: Level 2 (Foundational)

Marketing Team

• AI Literacy: Level 3
• Prompt Engineering: Level 3
• AI-Assisted Content Creation: Level 3
• AI-Assisted Analysis: Level 2
• Workflow Automation: Level 2

Finance Team

• AI Literacy: Level 2
• Prompt Engineering: Level 3
• AI-Assisted Analysis: Level 3
• AI-Assisted Content Creation: Level 2
• Workflow Automation: Level 2

HR Team

• AI Literacy: Level 3
• Prompt Engineering: Level 3
• AI-Assisted Content Creation: Level 3
• AI-Assisted Analysis: Level 2

Department Heads

• AI Literacy: Level 3
• Prompt Engineering: Level 2
• AI Strategy and Governance: Level 3

AI Champions (Selected from Each Team)

• AI Literacy: Level 4
• Prompt Engineering: Level 4
• Core functional competency: Level 4
• Workflow Automation: Level 3

Your mapping will look different based on your industry, team structure, and strategic priorities. The key is to be specific about what each role needs. Vague targets produce vague training and vague results.

Assessment: Measuring Where People Are Today

Once you have defined the target state, you need to assess the current state. The gap between the two is your training roadmap.

Assessment Method 1: Self-Assessment Survey

The simplest and fastest approach. Create a survey that asks employees to rate themselves on each relevant competency using the four-level descriptors above. Self-assessment is not perfectly accurate — some people overestimate, some underestimate — but it provides a useful baseline and can be deployed organisation-wide in days.

When to use: For initial baseline assessment across the whole organisation. Quick, low-cost, and scalable.

Limitation: Subjective. Some employees will rate themselves higher or lower than their actual skill level.

Assessment Method 2: Practical Assessment

Give employees a set of tasks to complete using AI tools and evaluate the quality of their output. For example: "Use an AI tool to draft a client proposal based on these notes" or "Analyse this data set and produce a summary of key findings." This tests actual ability, not perceived ability.

When to use: For targeted assessment of specific teams before designing customised training. More accurate than self-assessment but requires more time to administer and evaluate.

Limitation: Time-intensive. Best used for teams of thirty or fewer at a time.

Assessment Method 3: Manager Assessment

Managers assess their direct reports using the competency framework, based on observed performance and work output. This adds a layer of objectivity and can be incorporated into existing performance review processes.

When to use: As a complement to self-assessment, especially for identifying discrepancies between perceived and actual skill levels.

Limitation: Only as good as the manager's own understanding of AI competencies. Managers may need training on the framework before they can assess others accurately.

Combining Approaches

The most effective assessment combines all three: a self-assessment survey for breadth, practical assessments for targeted depth, and manager input for calibration. The combined data gives you a reliable picture of your organisation's AI readiness.

Building the Training Roadmap

With the gap analysis complete — you know the target state, you know the current state — you can build a training roadmap that is specific, prioritised, and measurable.

Step 1: Prioritise by Impact

Not all gaps are equally important. Prioritise based on two factors: the size of the gap (how far people are from the target) and the business impact of closing it (which gaps, when closed, will produce the most value).

A marketing team at Level 1 in Content Creation when the target is Level 3 has a large gap with high business impact — that is a top priority. A finance team at Level 1 in Automation when the target is Level 2 has a smaller gap with moderate impact — that can wait.

Step 2: Group into Cohorts

Create training cohorts based on current level and target level, not by department. People at Level 1 heading to Level 2 need different training from people at Level 2 heading to Level 3. Mixing them in the same session wastes everyone's time.

Step 3: Design Progressive Learning Paths

Each path should have clear milestones:

• Level 1 to Level 2: Half-day to full-day introduction workshop, followed by two weeks of supported practice, ending with a practical assessment.
• Level 2 to Level 3: Multi-session programme over three to four weeks, with application tasks between sessions, coaching support, and a project-based assessment.
• Level 3 to Level 4: Advanced programme over six to eight weeks, including mentorship, a capstone project, and train-the-trainer elements.

Step 4: Schedule in Waves

Do not try to train everyone at once. Roll out in waves, starting with the highest-priority cohorts. Use learnings from early waves to refine the programme for subsequent groups. Champions from early waves can support later ones.

Step 5: Build Ongoing Learning Infrastructure

The framework does not stop after the initial training. Build infrastructure for ongoing development: monthly refresher sessions, a community of practice, regular framework reassessment (quarterly), and pathways for people to progress to higher levels over time.

Common Pitfalls to Avoid

Pitfall 1: Making the Framework Too Complex

If your framework has twenty competencies with five levels each and a hundred behavioural indicators, nobody will use it. Keep it simple enough that a manager can understand and apply it in a single sitting. Six competencies with four levels is manageable. If you feel the need for more granularity, add it later after the core framework is established.

Pitfall 2: Building It in Isolation

A framework built by HR or L&D without input from department heads and frontline employees will miss critical context. The people doing the work know which AI skills would help them most. Include their perspective in the design process.

Pitfall 3: Treating It as a One-Time Exercise

AI evolves rapidly. A framework built today will need updating within six to twelve months as new tools emerge, new capabilities become available, and the skill requirements shift. Plan for regular reviews and updates from the start.

Pitfall 4: Assessing Without a Plan to Act

Assessment without follow-through is worse than no assessment at all. If you survey your organisation and discover significant gaps but then do nothing about it, you have created awareness of the problem without providing a solution. That breeds cynicism. Do not assess until you are ready to act on the results.

Pitfall 5: Ignoring the Human Side

Competency frameworks are inherently evaluative — they create categories of "better" and "worse" performers. Handle this sensitively. The framework should feel like a development tool, not a judgement tool. Emphasise that the goal is growth, not grading. Everyone starts somewhere, and the framework exists to help everyone progress.

Pitfall 6: Not Connecting to Business Outcomes

The framework should tie directly to business goals. If AI literacy for the sales team is a priority, be explicit about why: "So we can personalise outreach at scale and shorten the sales cycle." If AI analysis for the finance team matters, say so: "So we can produce monthly reports in two days instead of five." Abstract frameworks get filed away. Connected frameworks get funded.

Getting Started: Your First Steps

You do not need to build the entire framework before you start. Here is the minimum viable version:

1. Pick two or three priority departments where AI skills would have the highest impact.
2. Identify the three most relevant competencies for those departments from the six listed above.
3. Define the target level for each role in those departments — keep it simple.
4. Run a quick self-assessment survey to understand the current state.
5. Design training to close the largest, highest-impact gaps first.
6. Measure the results and use them to build the case for expanding the framework to the rest of the organisation.

Start small, prove value, then scale. This approach builds momentum and political support far more effectively than trying to launch a comprehensive framework for the entire organisation in one go.