Skip to content

How we handle your systems and data.

AI agents work inside the systems your business depends on. These are the practices we follow on every engagement, so your risk, legal and IT teams know what they are approving.

Your systems of recordLeast-privilege accessHuman approvalAgent09:02 read invoices (read-only)09:03 draft payment run09:03 wait approval: CFO09:41 done approved, loggedAudit trail

How every agent is built

The same controls apply whether the agent drafts a report or prepares a payment run.

  • Least privilege Each agent gets only the access its job needs, through its own account, so its permissions can be reviewed and revoked on their own.
  • Read-only by default Agents read from systems of record wherever writing isn't needed. Where an agent writes, it writes only to the fields and records in its scope.
  • A person approves what matters Payments, messages to customers, changes to financial records and anything above a set limit wait for a named person to approve.
  • Every action logged What the agent read, what it decided, what it did and who approved it, kept where your own team can review it.
  • Your environment where possible We build inside your Microsoft 365, cloud and systems wherever we can, so your data stays under your existing controls and contracts.
  • No training on your data For the agents we build, we specify enterprise AI services whose commercial terms exclude training on customer data, and we confirm the service and its terms with you in the design.

Your data

Where it is hostedAustralian or New Zealand hosting regions wherever the service offers them. Where a service is only available offshore, we tell you before it is used and record it in the design.
CredentialsKept in your password vault or secrets manager, never in documents, chat or email. Access is removed at handover or whenever you ask.
Copies of your dataWe avoid taking copies. Where test data is needed, we agree what, where and for how long, and delete it at the end of the engagement.
Scope in writingEach agent has a written scope: what it may read, what it may change, what needs approval and who is accountable for it.

Aligned with Australian and New Zealand guidance

In October 2025 the Australian Government replaced the Voluntary AI Safety Standard with the Guidance for AI Adoption and its six essential practices. Here is how our way of working maps to each one.

Australia's six essential practices, and what we do for each
Essential practiceHow we meet it
Decide who is accountable Each agent has a named business owner, agreed in the assessment and recorded in its scope.
Understand impacts and plan accordingly The assessment sets out who is affected by each automation, what changes for staff and customers, and what happens when it is wrong.
Measure and manage risks Risks are rated per use case, and the controls (access, approval limits, monitoring) are designed to match them.
Share essential information Staff know when they are working with an agent, and customer-facing output is reviewed by a person before it goes out.
Test and monitor Agents are tested against real cases before go-live and monitored afterwards: what they did, how often a person corrected them, what they cost.
Maintain human control Approval steps, a way to pause any agent and a person who can always override it.

New Zealand

For New Zealand organisations we work to MBIE's Responsible AI Guidance for Businesses and the Privacy Act 2020, including its information privacy principles.

Privacy law

Where an agent handles personal information, the design follows the Australian Privacy Principles under the Privacy Act 1988 (Cth), or New Zealand's Privacy Act 2020, and your own privacy policy.

This website

  • HTTPS only Every request is served over HTTPS with TLS 1.2 or later.
  • No tracking cookies Visitor numbers come from Cloudflare Web Analytics, which doesn't use cookies or follow you across sites.
  • Locked-down pages A strict content security policy allows scripts only from this site and from Cloudflare: its visitor analytics, and on the Contact page its check that a person is sending the form. Only the Contact page can send a form, and only to this site.
  • A way to report issues Our security contact is published at /.well-known/security.txt.

Found a security issue?

Email tom@spurbuck.com with the details. We will acknowledge it promptly and keep you informed until it is fixed. Our full contact details are in security.txt, and our privacy policy explains how we handle personal information.

Questions people ask us

How does Spurbuck Advisory keep AI agents secure?

Every agent runs with its own least-privilege account, reads rather than writes wherever it can, waits for a named person to approve anything consequential and logs every action. Each agent has a written scope that says what it may read, change and escalate.

Is our data used to train AI models?

For the agents it builds, Spurbuck Advisory specifies enterprise AI services whose commercial terms exclude training on customer data, confirms the service and its terms with the client in the design, and builds inside the client's own environment wherever possible.

Where is our data hosted?

In Australian or New Zealand hosting regions wherever the service offers them. If a service is only available offshore, the client is told before it is used and the decision is recorded in the design.

Start with a free discovery session.

Thirty minutes with Tom, with your IT or risk lead welcome, to go through how an agent would work inside your systems. You leave with a clear view of what is worth automating, what it would take, and what it would return. No obligation.

We use the details you give us only to arrange and hold your session. See our privacy policy.