Short answer
An AI training needs assessment works out who needs which AI skills, at what level, by comparing what each role must do with AI against what people can do today. This template uses six skill areas, a 0 to 3 proficiency scale, a role target matrix, short practical checks and a gap score that ranks training priorities so budget goes where gaps are largest.
- Assess six skills, from safe data use to governance, not just prompting.
- Set target levels by role before measuring anyone.
- Pair self-ratings with a five-minute practical check.
- Rank gaps by size, headcount and importance to safety or compliance.
Run this assessment before you book training. It takes each person about 20 minutes, and the manager or AI champion about two hours to compile. The output is a ranked list of gaps that tells you what to train, for whom, and in what order. Turn the result into a schedule with the AI learning plan for teams.
The training needs assessment template
Part A: Skill areas and proficiency scale
| # | Skill area | What level 2 (practitioner) looks like |
|---|---|---|
| S1 | Safe use and data rules | Knows which information can go into which approved tool, and acts on it without asking |
| S2 | Prompting and context | Gives AI the task, audience, format and source material, and gets usable output first or second time |
| S3 | Checking output | Checks facts, figures and references against sources before using output |
| S4 | AI in our core tools | Uses the AI features in the software we already have (Microsoft 365, Google Workspace, Xero, our CRM) |
| S5 | Workflow design | Breaks a task into steps and spots where AI can assist or automate |
| S6 | Governance and oversight | Approves tools, reads vendor data terms, owns risks and incidents |
Proficiency: 0 none; 1 aware (can explain it); 2 practitioner (does it unaided in daily work); 3 lead (can teach others and set rules).
Part B: Role target matrix
Set targets before measuring anyone. Edit the roles to match your organisation.
| Role | S1 | S2 | S3 | S4 | S5 | S6 |
|---|---|---|---|---|---|---|
| All staff | 2 | 2 | 2 | 1 | 0 | 0 |
| Team leads | 2 | 2 | 2 | 2 | 1 | 1 |
| AI champion | 3 | 3 | 3 | 3 | 2 | 1 |
| Directors and senior managers | 2 | 1 | 2 | 1 | 1 | 2 |
| Operations or IT administrator | 2 | 2 | 2 | 2 | 2 | 2 |
Part C: Self-rating and practical check
Each person rates themselves 0 to 3 on each statement, then completes the matching check. Use the lower of the two results.
| Skill | Self-rating statement | Five-minute practical check |
|---|---|---|
| S1 | “I know what information I must not put into each AI tool we use.” | Sort six real items (client email, public web page, payslip and so on) into allowed and not allowed |
| S2 | “I can get a usable first draft from AI for my regular tasks.” | Write a prompt for a real task; score the output against a short rubric |
| S3 | “I check AI output against sources before I use it.” | Find the two planted errors in an AI summary of a short document |
| S4 | “I use the AI features built into our main software.” | Complete one task using the AI feature in your main application |
| S5 | “I can describe a task as steps and say where AI could help.” | Map a familiar task into five steps and mark one for AI |
| S6 | “I could decide whether a new AI tool is safe to approve.” | Read a vendor’s data terms and answer three questions about training and storage |
Part D: Gap score and priority
For each role and skill:
Priority score = average gap (target minus current, never below zero) × number of people in the role × importance.
Importance: 1 nice to have; 2 needed for current work; 3 needed for safety, privacy or compliance. S1 and S3 are always importance 3. Rank all scores and train the top three first.
How to run the assessment
- Confirm the policy first. People cannot be assessed on rules that do not exist yet. Adopt an AI use policy before measuring S1.
- Set role targets. The manager and AI champion agree Part B.
- Send the self-ratings and checks. Allow 20 minutes per person. Make clear it is for planning training, not performance review.
- Score and compile. Take the lower of self-rating and check, then calculate priority scores.
- Rank and plan. Pick the top three gaps; choose a format for each, using our comparison of AI workshops and self-paced courses.
- Reassess in six months. Use the same checks so the results are comparable.
Worked example
Illustrative example: a 22-person engineering consultancy in Hamilton with 12 engineers, 5 project administrators, 3 directors, 1 AI champion and 1 IT administrator.
| Gap | Target | Average current | People | Importance | Priority |
|---|---|---|---|---|---|
| Engineers, S1 safe use | 2 | 0.8 | 12 | 3 | 43.2 |
| Engineers, S3 checking output | 2 | 1.0 | 12 | 3 | 36.0 |
| Directors, S6 governance | 2 | 0.3 | 3 | 3 | 15.3 |
| Administrators, S2 prompting | 2 | 0.6 | 5 | 2 | 14.0 |
| Administrators, S4 core tools | 2 | 0.8 | 5 | 2 | 12.0 |
The practical checks changed the picture. Engineers rated themselves 2 on safe use on average, but most put a client report in the “allowed” pile for a personal AI account. The plan became: a 90-minute safe use session for everyone, a hands-on checking session for engineers using their own documents, and a governance briefing for directors through an executive AI workshop.
Common mistakes
- Training only prompting. The costliest mistakes tend to come from gaps in S1 and S3, not weak prompts.
- Trusting self-ratings. Use the practical checks.
- No time to practise. WorkSafe’s psychosocial guidance lists a lack of training to use new equipment or technology among workplace hazards (WorkSafe); rolling out tools without training is a people risk as well as a quality one.
- Ignoring security training. The joint small-business guide from NCSC New Zealand and Australian partners includes giving staff AI security training in its checklist (NCSC).
- Treating it as a test of people. If staff think results affect their jobs, they will rate themselves high and learn less.
Next step
Schedule the top gaps in the AI learning plan for teams, name an AI champion using training an internal AI champion, and baseline usage with the AI adoption survey. If you want training designed around your own documents and tools, see AI training for business teams.
Sources
- Managing psychosocial risks at work, WorkSafe New Zealand, 10 April 2025
- Quickstart: Businesses using AI tools, Ministry of Business, Innovation and Employment, accessed 19 October 2026
- Artificial intelligence for small business, ASD’s ACSC with NCSC New Zealand and COSBOA, 14 January 2026
Written by the Levy Street team in Wellington. Last updated 19 October 2026.