Claude training for NZ teams.

What Claude training covers

Claude training is hands-on practice in your own Claude Team organisation, on the documents your people work with: contracts, tenders, policies, board papers and reports. Claude is strongest at long, careful document work. Projects hold the reference material and instructions for a repeat task; artifacts are, in Anthropic’s words, “anything Claude makes for you that you’d put in front of someone”, from a document to a small interactive tool.

Both can be shared, so training starts with the owner’s settings. Teams using several assistants can combine this with our broader AI training for business teams.

Who it is for

• Document-heavy teams on Claude Team: legal, finance, policy, tenders, compliance and consulting.
• Pilot groups of two or more seats (Team’s minimum), trying Claude on real documents before a wider rollout.
• Owners and Primary Owners, who control sharing, connectors and feedback.
• Teams moving off personal Pro accounts onto Team, where training on content is off by default.

If the tool decision is still open, our Claude guide covers plans and data handling, and the comparisons of ChatGPT and Claude and Claude and Gemini test them side by side.

Session options, starting with owner settings

Choose the options that match the team’s work. The owner settings come first, because projects and artifacts can be shared.

Settings record

Each owner setting, its value and the reason, signed off by an owner.

Shared projects

Two or three tested projects for repeat tasks, with instructions and reference files.

Exercise pack

Role tasks built from your documents, with personal details removed where the task allows.

Checking rules

How each output type is verified before use, and who signs off.

Adoption baseline

Timed target tasks and current usage, for comparison later.

The organisation settings we check first

From Anthropic’s help and privacy pages, checked on 19 and 20 October 2026:

• Training and feedback: Team content is not used for training by default. The exception that catches people is the thumbs up or down button: rating a chat keeps it for up to five years. Owners can turn the rating buttons off for everyone.
• Retention: Team chats are kept until someone deletes them, and are removed from back-end storage within 30 days after that. Conversations with Claude Fable carry a default 30-day retention for safety monitoring.
• Project sharing: the Share projects and Public projects settings are both on by default. Members can be given Can view or Can edit.
• Artifacts: they need “Cloud code execution and file creation” turned on. They stay inside the organisation by default, but email invitations to outside people are on by default on Team, up to 50 per artifact.
• Artifact connectors: one organisation-wide switch. Owners cannot limit which apps artifacts reach, and each user connects their own.
• Custom connectors: on Team only Owners and Primary Owners add them, and they are called from Anthropic’s cloud, not your network.

A sample exercise: comparing a marked-up contract with your standard terms

For anyone who negotiates supplier or client terms. It is a comparison aid, not legal advice: material changes still go to your approver or a lawyer.

1. Choose a live contract. Your standard terms and the other party’s marked-up version. Remove personal details the comparison does not need.
2. Set up a project. Call it “Contract review” and add your standard terms and a one-page negotiation position: what you accept, what you never accept, who approves exceptions.
3. Run the prompt below in a chat inside the project, with the marked-up version attached.
4. Trace every row. Check each quoted clause against both documents by clause number. Then read five clauses Claude did not flag, to test what it missed.
5. Escalate, do not decide. Rows marked “no” or “unclear” go to the approver with the original documents.
6. Score and save. Record missed clauses, wrong flags and minutes saved against a manual comparison. If it passes, keep the project for the team at Can view.

Compare the attached marked-up contract with our standard terms in this project.

Create a document with one row per changed clause: clause number, our wording, their wording, what the change does in plain New Zealand English, and whether it sits inside our negotiation position (yes, no or unclear).

Quote both wordings exactly; never paraphrase inside a quote column. Include clauses that appear in only one document. If you are unsure whether two clauses correspond, say so rather than guessing.

End with the clauses marked "no" or "unclear" for [approver's role] to decide. Do not recommend accepting or rejecting any change.

Our lesson on summarising documents with citations and the contract clause review workflow go deeper on the same checking habit.

Measuring what changes after Claude training

Afterwards, repeat document tasks run from shared projects, every quoted clause or figure is traced before use, and nothing leaves the organisation without someone deciding it should. Anthropic’s plan comparison lists usage analytics on Team, which shows who is active. For value, compare a timed manual run of each target task with the Claude version, before training and at 30, 60 and 90 days, using the method in our lesson on measuring adoption after training. For document work, count missed clauses as well as minutes: a faster review that misses a change is not a saving.

Where Claude falls short

• One shared usage pool. Limits reset on a rolling five-hour window with weekly caps, and chat, desktop and Claude Code draw on the same pool. Heavy users need Premium seats; developers belong in a separate Claude Code session.
• No audit trail on Team. Audit logs, the Compliance API and custom retention are Enterprise only.
• Fitting is not reading. A long tender pack fits in one conversation, yet a clause can still be missed. Spot-check unflagged sections.
• Rollout gaps. When Cowork merged into the main app on 16 September 2026, the rollout began with Pro and Max, and Team was to follow. We check what your plan shows before teaching it.
• Artifacts that look finished. A polished Doc or dashboard reads like checked work. Label it a draft until someone has traced the figures.

How this differs from Anthropic’s free training

Anthropic’s Claude Academy offers free courses, tutorials and use cases. AI Fluency: Framework and foundations runs to 14 lessons and about four hours, built around delegation, description, discernment and diligence, and AI capabilities and limitations builds a realistic picture of what models can do. We recommend both as pre-reading, especially for champions.

A public course cannot open your organisation settings, decide whether outside invitations stay on, or test a prompt on the contract on your desk. Our sessions do that, and measure it. Workshops versus self-paced courses sets out the trade-offs.

New Zealand notes: privacy, data location and GST

• Privacy Act 2020. On Team with feedback off, Anthropic handles data on your behalf, so under section 11 you remain responsible for it. Feedback is different: a rated chat kept for up to five years and possibly used for training is, in the Privacy Commissioner’s terms, a provider using information for its own purposes, which can be an offshore disclosure under IPP 12. IPP 9 says not to keep personal information longer than needed, and Team has no custom retention, so we teach a deletion habit. Sharing an artifact containing personal information outside the organisation can be a disclosure under IPP 11.
• Data location. On Anthropic’s own platform, inference runs in “global” or “us” locations only, and workspace storage is US-only. Amazon Bedrock’s Australian profile keeps Claude inference within Australia and New Zealand, but that is a developer route, not the Claude app.
• Prices and GST. As shown to NZ visitors on 19 October 2026, a Team Standard seat is $40 a month billed annually or $50 monthly. The page says prices include 15% GST but never labels the currency. If it is NZD, $40 including GST is $34.78 excluding GST. Confirm on your first invoice.

Written by the Levy Street team in Wellington. Last updated 19 October 2026.