Short answer
AI training for business teams teaches staff to use AI tools safely and productively on their real work. Levy Street runs practical, hands-on sessions for New Zealand teams built on your own documents, templates and software, whether that is ChatGPT, Claude, Microsoft Copilot or Gemini, covering prompting, checking answers, privacy rules and turning repeat tasks into reusable workflows, followed by free lessons on our Learn site.
- Exercises use your own documents and tools, not generic examples.
- Every session covers checking AI output and what must never go into a prompt.
- Only 24% of New Zealanders have done AI training, against 39% globally (KPMG, 2025).
- Free Learn lessons follow each session so practice continues.
What our AI training covers
AI training for business teams means people leave able to do their own work better with AI, safely, the next morning. Our sessions are practical and hands-on: participants work in their own accounts, on their own documents, templates and spreadsheets, with the tools your organisation has approved. Every session covers two things generic courses skip: how to check what the AI produced, and what must never be typed into it.
The gap is real. Only 24% of New Zealanders have done any AI-related training, against 39% globally, and only 36% believe they have the skills to use AI tools appropriately (KPMG and University of Melbourne, 2025). Among small businesses, mistrust of output quality (37%) and privacy concerns (40%) are the top reasons for not using AI more (Xero, May 2026). Both are teachable.
Who it is for
- Whole teams getting a tool such as Microsoft Copilot, ChatGPT Business or Claude Team, who need to use it well from the first week.
- Role groups with specific work: finance, HR, customer support, sales, operations.
- Internal champions who will support colleagues after we leave.
- Developers adopting AI coding agents.
- Leaders, who usually need a decision workshop more than tool training. See executive AI workshops.
The curriculum
We choose modules to fit the team. Each one has an exercise on your own material and a free lesson on our Learn site for practice afterwards.
| Module | What people learn | Exercise on your material | Free lesson |
|---|---|---|---|
| Foundations | What current AI tools can and cannot do | Try three real tasks and compare results | AI foundations |
| Prompting with context | Instructions, examples and source material | Rewrite a weak prompt for a real task | Writing useful business prompts |
| Checking answers | Spotting errors, verifying against sources | Fact-check an AI summary of your own document | Checking AI answers against sources |
| Protecting information | What never goes in, which plans are safe | Sort sample data into allowed and not allowed | Protecting customer information |
| Documents and data | Summaries with citations, spreadsheet analysis | Summarise a real report; analyse a real export | Analysing a spreadsheet with AI |
| Drafting with review | Emails, proposals and replies a person signs off | Draft three real replies and edit them | Drafting emails with human review |
| Reusable workflows | Turning a repeat task into a saved prompt or automation | Build a team prompt for a weekly task | Building a reusable prompt library |
| Safety | Prompt injection and when to escalate | Spot planted instructions in a sample document | Recognising prompt injection |
What participants can do afterwards
- Choose the right approved tool and plan for a task, and explain why
- Write a prompt with clear instructions, context and an example
- Check an AI answer against its source before using it
- Say what information must never be entered, and why
- Use AI on at least two real tasks from their own job
- Save and share a working prompt with the team
- Report an AI mistake through the agreed route
Formats you can choose
Formats are options, combined to suit the team: on-site hands-on workshops for whole teams; smaller clinics for role groups working on their own tasks; a champion programme that prepares a few people to support everyone else; remote follow-up sessions where people bring problems from the weeks since; and developer sessions on coding agents inside your own repositories. Our comparison of workshops and self-paced courses helps decide the mix.
How the work runs
- Discover: find the real tasks. A short training needs assessment and a few conversations identify the tasks worth teaching, the tools and plans in use, and any policy gaps. We collect real documents and templates for the exercises.
- Prove: pilot with one group. The first session runs with one team. Exercises that fall flat are rewritten before the wider rollout.
- Scale: roll out and measure. Sessions run across teams, champions take over day-to-day support, and adoption is measured against the baseline agreed at the start.
- Session plans
- Built around your tasks, tools and rules.
- Exercise packs
- Real material, with personal details removed.
- Starter prompt library
- Working prompts for the team’s most common tasks.
- One-page rules
- What is allowed, what is not, and who to ask.
- Adoption measures
- A baseline and a follow-up check.
- Lesson path
- The free Learn lessons each participant should do next.
New Zealand rules every session covers
The Privacy Commissioner advises against entering personal or confidential information into a generative AI tool unless the provider has explicitly confirmed it is not retained or disclosed (OPC). In practice the plan matters: OpenAI’s pricing page says consumer ChatGPT plans use content to train models unless people opt out, while business plans do not (OpenAI). Staff using personal accounts for work is a common gap: 52% of larger NZ organisations in Datacom’s 2025 survey called unapproved “shadow AI” a problem (Datacom). Training is the moment to fix it.
What drives the cost, and what training cannot do
Cost depends on the number of people and sessions, how much preparation goes into exercises on your material, whether sessions are on site outside Wellington, and whether a champion programme or follow-up clinics are included. Training cannot fix the wrong tool, a missing policy or a workflow that should not use AI at all. One session also rarely changes habits on its own, which is why follow-up and measurement are built in. In Datacom’s 2026 survey, 66% of larger NZ organisations had provided AI skills training in the past year, yet skills remained the most-cited barrier to scaling AI (Datacom).
Work this draws on
- MOGGED Digital: Levy Street’s AI-native Wellington digital studio, which says it has served 100+ businesses, mostly in New Zealand. AI runs through its research and drafting while “every decision is still made by a person” (MOGGED), which is the working habit we teach.
- World of ClaudeCraft: our open-source browser MMO, whose first playable version was built in about 48 hours with the Claude Fable 5 coding model. What we learned building and maintaining it with coding agents shapes how we train developers to use them.
Training for a specific tool
If your team has already chosen its assistant, these pages cover tool-specific sessions, starting with the admin and data settings: Microsoft Copilot training, ChatGPT training and Claude training.
Sources
- Trust, attitudes and use of artificial intelligence: a global study 2025, KPMG New Zealand and University of Melbourne, 2025-05
- 2026 State of AI Index: AI uptake at 91%, leadership lags, Datacom, 2026-08
- Techweek: barriers preventing Kiwi SME AI adoption, Xero, 2026-05-18
- Generative artificial intelligence: expectations, Office of the Privacy Commissioner, 2023-06-15
- ChatGPT pricing, OpenAI, accessed 2026-10-19
- 2025 State of AI Index, Datacom, 2025-08
- Master of Artificial Intelligence, University of Canterbury, 2026-08-10
- About MOGGED, MOGGED, accessed 2026-10-19
Written by the Levy Street team in Wellington. Last updated 19 October 2026.