ChatGPT Training NZ, home

4.8% of Canterbury employees work in government

ChatGPT training for Canterbury public-service staff

Develop bounded AI tasks that staff can review before they affect a public service. Define the question, limit the sources and keep a person accountable for the accepted output.

EnquireWhere ChatGPT helps
Illustration generated with AI.
Employees in government
16,087Canterbury, February 2025
Share of Canterbury employees
4.8%5 of 16 regions
Share across New Zealand
6.4%all regions, February 2025
Location quotient
0.75×less concentrated than nationally

Government in Canterbury

Environment Canterbury describes report review, draft editing and consultation themes among its AI productivity uses.1 It also publishes explicit exclusions. A fictional agency workshop can examine a public-programme briefing without treating local AI activity as permission to automate a council decision. Start with the administrative work around the decision and the evidence it needs.

A ChatGPT Project can organise related instructions, files and chats.5 Define the agency's source scope and required output there. We teach staff to ask for an evidence inventory and missing inputs before requesting a recommendation. If the answer depends on an unavailable document, the useful result is an unresolved question for its owner.

Paid employees in government went from 13,791 in February 2020 to 16,087 in February 2025 (+16.6%).

Share of paid employees in government, by region

Share of employees in government by region, Canterbury highlighted
RegionChartShareNZ×
Wellington 17.8% 6.4% 2.8×
Manawatū-Whanganui 9.5% 6.4% 1.5×
Northland 6.2% 6.4% 1.0×
Waikato 5.0% 6.4% 0.8×
Canterbury (this page) 4.8% 6.4% 0.8×
Gisborne 4.8% 6.4% 0.7×
Marlborough 4.8% 6.4% 0.8×
Auckland 4.7% 6.4% 0.7×
Hawke's Bay 4.6% 6.4% 0.7×
Bay of Plenty 4.5% 6.4% 0.7×
Otago 4.4% 6.4% 0.7×
Taranaki 4.4% 6.4% 0.7×
West Coast 4.4% 6.4% 0.7×
Nelson 4.2% 6.4% 0.7×
Southland 3.6% 6.4% 0.6×
Tasman 2.1% 6.4% 0.3×

Paid employees at business locations in the region, February 2025. The tick marks the national share. × is the regional share divided by the national share.

Source: Stats NZ, Business Demography Statistics, February 2020 and 2025.

Local context

  • Selwyn's economic strategy includes engagement with business and work on investment attraction.2 This gives a local context for a generic service-planning exercise using public strategy documents. It does not imply the council has bought ChatGPT or that TheColab works for it.2

Organisations are named for local context only. No relationship with them is implied.

Towns in Canterbury

Each town page has its own figures and local brief.

Or see ChatGPT training across Canterbury.

Where ChatGPT helps government teams

  • Organise a source-led briefing Project

    Use an approved public strategy, fictional project notes and a glossary of service terms. Ask for a briefing outline that marks unsupported statements. Keep the source owner involved when the documents disagree. Staff should be able to trace a proposed option to evidence rather than to the phrasing of a previous chat.

    • Policy and service advisers
    • Projects

  • Analyse invented programme records

    ChatGPT analysis can retain calculations and caveats for reviewer inspection.8 Practise grouping fabricated programme categories and detecting missing labels. Require a table of exclusions alongside any chart. An analyst checks the raw rows independently before discussing the findings, with no inference about actual residents or public-service outcomes.

    • Performance and policy analysts
    • Data analysis

  • Trial a drafting helper with a stop rule

    Where supported, a custom GPT can guide an internal briefing routine under controlled sharing. OpenAI also documents moving GPTs to plugins.6 Test an incomplete request and a question about individual eligibility. The helper should request evidence or direct staff to an authorised officer, rather than manufacture a decision.

    • Capability and knowledge leads
    • Custom GPTs and controlled sharing

  • Collect a public-document evidence pack

    Connected apps and ChatGPT Work support multi-step tasks with authorised context.7 Specify an approved folder and request a list of retrieved documents before synthesis. Check currency and completeness. Keep the rehearsal limited to a draft pack; changes to systems, external communication or public publication need the agency's normal approvals.

    • Digital and project staff
    • Connected apps and agent workflows

Test a public-service draft for missing evidence

  1. Create a fictional agency question with a deliberately incomplete source pack.
  2. Set Project instructions to expose gaps and distinguish evidence from proposed options.
  3. Request an outline and inspect whether the missing source is correctly identified.
  4. Supply the checked information, review the revised draft and record the accepting officer.

Courses

  1. 1Introductory

    ChatGPT essentials workshop

    A hands-on first session: clear prompts, files, data analysis, Projects and search, practised on tasks your team already does.

  2. 2Organisation

    ChatGPT Business and Enterprise rollout

    Set up a ChatGPT workspace properly: plan choice, admin settings, an AI use policy, shared skills and Projects, and support that keeps people using it.

  3. 3Advanced

    Skills, plugins, workspace agents and connected apps

    For power users and technical leads: skills, plugins, Projects, ChatGPT Work, connected apps, data analysis and workspace agents.

Government in other regions

Other industries in Canterbury

See ChatGPT for government and public services across New Zealand, or all training in Canterbury.

Government in Canterbury: questions

Can the training make a procurement decision?

A workshop can help define tasks, data boundaries and questions for procurement colleagues. The agency keeps responsibility for tool selection and its approval process. Use learning results as evidence for that discussion, without treating them as a completed assurance review.

Do the national AI principles create a legal mandate?

The Public Service AI Framework is explicitly non-binding. It offers a useful reference for agencies and may have broader application. Check your own organisation's obligations and policies rather than describe alignment as a law that every council must follow.3

How do we stop an agent taking unintended actions?

Start with a tightly scoped source-gathering task and review enabled tools. Keep sending, editing and publishing outside the initial exercise. Confirm the workspace and source permissions with your administrator, and test what the task does when access is missing.7

Where can Canterbury agency teams attend?

We deliver workplace sessions across the region, with material the agency has cleared for learning. Public workshop days suit staff starting individually. For a team pilot, include the officer who owns the task and someone who can decide information-access questions.

Train your government team in Canterbury

  1. Send us a few linesHow many people, where you are, and the jobs that eat your week.
  2. We reply with a planA suggested session and format, plus what to bring: real quotes, emails and questions.
  3. Your team does the workAt your place or over video. They finish with ChatGPT doing real tasks, not demos.
  • In person
  • Remote
  • Half-day workshop
  • Full-day workshop
  • Team rollout

Independent training by TheColab. Not affiliated with OpenAI.

Enquire

Tell us about your team in Canterbury. We reply by email with a suggested course and format.

We use these details only to reply to you, and store them securely with TheColab. Privacy

Sources

Written 6 October 2026. Employment figures come from Stats NZ Business Demography Statistics.