Short answer
An AI learning plan for teams sets out what people will learn, in what order, how they practise on real work and how progress is checked. This template runs for 12 weeks in four stages (safe basics, daily tasks, team workflows, measure and embed), needs about two hours per person a week, and points to free Levy Street lessons for the key weeks.
- Two hours a week for 12 weeks beats a one-day workshop.
- Safe use and checking output come before productivity tips.
- Every week ends with practice on a real task and evidence.
- Measure adoption before week 1 and after week 12.
Use this plan after you have an AI use policy and approved tools in place. Start with the AI training needs assessment if you have not run one: it tells you which weeks to stretch and which to shorten. Each week follows the same rhythm: about 30 minutes on a lesson, an hour of practice on real work, and a 30-minute team show-and-tell.
The 12-week AI learning plan
| Week | Stage | Topic and lesson | Practice on real work | Evidence |
|---|---|---|---|---|
| 1 | Safe basics | Rules and approved tools: protecting customer information when using AI | Sort ten real items into allowed and not allowed; set up your business account | Sorting exercise passed; account active |
| 2 | Safe basics | Checking output: checking AI answers against sources | Check an AI summary of one of your own documents | Errors found and noted |
| 3 | Daily tasks | Prompts: writing useful business prompts | Rewrite prompts for three of your recurring tasks | Before and after outputs |
| 4 | Daily tasks | Context: giving AI the background, examples and sources it needs | Build a context pack for one recurring task | Context pack saved |
| 5 | Daily tasks | Email drafts with human review | Draft five real replies with AI; send only after full review | Time per reply noted |
| 6 | Daily tasks | Meeting notes into actions | Turn two real meetings’ notes into owners and dates | Action lists shared |
| 7 | Team workflows | Summarising documents with citations | Summarise a document your team uses, with page references | Summary checked by a colleague |
| 8 | Team workflows | Spreadsheets and data | Ask three questions of a real spreadsheet and verify the answers | Answers verified |
| 9 | Team workflows | Shared prompts: building a reusable prompt library | Contribute one tested prompt with a known-good output | Prompt in the library |
| 10 | Team workflows | Mistakes and escalation: what to do when AI gets it wrong | Walk through two incident scenarios | Team knows who to tell |
| 11 | Measure and embed | Pick one team workflow to improve | Time the workflow today; propose where AI helps | Baseline timings |
| 12 | Measure and embed | Review progress | Repeat the AI adoption survey; agree next quarter’s focus | Survey results and a plan |
Individual plan card
Each person fills in one card in week 1 and reviews it with their manager in weeks 6 and 12.
| Field | Entry |
|---|---|
| Name and role | |
| Target skill levels | From the training needs assessment |
| Two tasks I will use AI for | 1. 2. |
| Weekly time block | Day and time, in the calendar |
| Learning buddy | |
| Check-in dates | Week 6 and week 12 |
| Evidence collected | Links to outputs, timings, prompts added |
How to run the plan
- Baseline first. Run the adoption survey before week 1 so week 12 has something to compare against.
- Book the time. Put a recurring two-hour block in every calendar, in work time.
- Name a champion. They run the show-and-tell and keep the prompt library; see training an internal AI champion.
- Use real work, not exercises. Every practice task comes from the person’s own job.
- Collect evidence weekly. Five minutes at the end of each session; it makes week 12 easy.
- Adjust after week 6. Stretch weeks where people are struggling; skip what they already do well.
Worked example
Illustrative example: a 10-person reservations and office team at a Queenstown tourism operator follows the plan from August to October, with the operations coordinator as champion. Weeks 5 to 8 were adapted to their work on reservation enquiry handling:
| Week | Their practice task |
|---|---|
| 5 | Draft replies to five real booking enquiries, checked against the live availability system before sending |
| 6 | Turn the weekly operations meeting into an action list in the shared channel |
| 7 | Summarise supplier terms for a new activity partner, with clause references |
| 8 | Ask questions of the season’s booking spreadsheet, such as cancellations by source, and verify each answer by filter |
By week 12 the team had eight tested prompts in its library, and the survey showed most of the team using AI weekly for enquiries. The workflow chosen in week 11, first replies to group booking requests, became the next quarter’s project.
Common mistakes
- A one-off workshop. Without practice and follow-up, use fades within weeks.
- Generic exercises. People learn fastest on their own tasks.
- Skipping the rules. Weeks 1 and 2 prevent the mistakes that end AI use.
- No time allocated. Learning squeezed around a full workload does not happen. MBIE’s quickstart also notes staff training may need to change as AI is introduced (MBIE).
- Measuring attendance. Count changed work and evidence, not sessions attended.
WorkSafe lists a lack of training to use new equipment or technology among psychosocial hazards at work (WorkSafe). Giving people time to learn is part of introducing the tools well.
Next step
Use rolling out AI to a small team to plan the change around the learning. If you want the first weeks delivered on your own documents and tools, see our 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
Written by the Levy Street team in Wellington. Last updated 19 October 2026.