The step your best people do from memory, done the same way every time. We find the one manual process your team queues behind, build AI for it inside your own systems, and prove it beside your people before it touches a customer.
The lead on every engagement spent 18 years on operations and data systems for BHP, Rio Tinto and Senex Energy, gained while working with previous consulting employers. These companies are not Solve8 clients.
We qualify work by the problem, never by headcount. These are the patterns we look for.
Firms where quotes, job sheets, supplier bills and client documents still pass through someone's inbox before they reach a system. The work queues behind the two or three people who know how each one should be handled.
Correspondence, ministerial and OIA workloads, and the suppliers who have to show an agency how an algorithm reached its answer. Here the governance documentation matters as much as the automation.
Production and maintenance data spread across a historian, an ERP and spreadsheets, plus compliance paperwork for every job. Reporting that someone fixes by hand every month is usually the first thing to go.
Dairy, horticulture, forestry and the processors and freight operators around them. Seasonal peaks expose every manual step, so the case for automation is easy to measure: count the overtime in the busiest month.
New Zealand has no standalone AI act. The Government's first AI strategy, published by MBIE in July 2025, relies on the laws already in place and adds voluntary Responsible AI Guidance for Businesses. In practice the Privacy Act 2020 does most of the work, and a privacy officer will ask the same questions about an AI model that they ask about any other system holding personal information.
The question that decides most designs is where the data goes. IPP 12 restricts disclosing personal information overseas, while a provider that only stores or processes it for you is generally holding it on your behalf. So we build inside your tenant, keep prompts and outputs there, and switch off any vendor option that trains on your data. Each data flow is written down in Discovery, which also gives agency suppliers what they need to support an Algorithm Charter review.
The same thinking runs through our Australian work. Our guide to AI agent governance, data access and human override applies almost line for line to a New Zealand deployment.
Microsoft's New Zealand North Azure region went live in December 2024, so AI workloads can now run in country. If a model you need is only offered in Australia East, the design document says so before anything is built.
Private AI inside your tenantSupplier bills coded and matched without rekeying, with GST at 15 per cent, IRD payday filing and trans-Tasman groups reporting across Xero organisations in both currencies.
Consolidating several Xero organisationsMost New Zealand organisations already pay for Microsoft 365. Before Copilot is switched on, we check who can see what, because Copilot will surface any file a person has access to.
Copilot readiness assessmentAn ERP, a job system, a CRM and a spreadsheet each holding part of the answer. We connect them through whatever they expose, whether that is an API, a database or a nightly export.
System integrationWe measure the bottleneck as it runs today, agree the accuracy threshold with you, and design the solution inside your tenant. Fixed price, and you keep everything if you stop here.
The AI runs beside your team on real work. Each week we compare its output with what your people decided. Nothing reaches a customer.
The AI drafts and a person approves every output. Code, documentation and support move to your IT team or your managed service provider.
More on the method in what an AI strategy engagement should deliver, and on choosing the right tool in Power Automate or an AI agent.
Yes. Discovery workshops, shadow mode reviews and handover all run on Microsoft Teams, so an organisation in Auckland, Wellington, Christchurch or the regions gets the same engagement as one in Brisbane. Brisbane sits two to three hours behind New Zealand, so an NZ morning meeting is an early start for us rather than a late night for you. Where being in the room genuinely helps, for example walking a process with the people who run it, we travel.
In most cases, yes. We build inside your own Microsoft, AWS or Google tenant, so the data stays where your tenant already keeps it. Microsoft opened its New Zealand North Azure region in December 2024, which lets AI workloads run in country. Where a model or service you need is only offered in Australia East, we say so in the design document before anything is built, and you decide whether that is acceptable under your privacy obligations.
The thirteen information privacy principles apply to personal information whether a person or a model handles it. The ones that come up most are IPP 1 and 3 (only collect what you need and tell people why), IPP 8 (check accuracy before you use information), and IPP 12, which restricts disclosing personal information overseas unless comparable safeguards apply. A cloud provider that only stores or processes data on your behalf is generally treated under section 11 as holding it for you, which is why where your tenant runs matters more than which AI vendor you pick. We map each data flow against these principles during Discovery.
Not a standalone one. The Government's AI strategy, published by MBIE in July 2025, chose to rely on existing law, mainly the Privacy Act, the Fair Trading Act and the Human Rights Act, with voluntary Responsible AI Guidance for Businesses alongside it. Public sector agencies that signed the Algorithm Charter have extra transparency commitments, and suppliers to those agencies are often asked to support them. Our governance documentation is written so it can be handed to a privacy officer or an agency reviewer without rework.
Yes. Xero started in Wellington and a lot of New Zealand finance teams run it across several organisations. We automate supplier bill coding, reconciliation and multi-entity reporting in Xero and MYOB, with the NZ details handled properly: GST at 15 per cent, IRD payday filing for payroll, and trans-Tasman groups that report in both currencies.
With one process. Phase 1 is three to four weeks at a fixed price: we measure the bottleneck as it runs today, agree the accuracy threshold with you, and design the solution inside your tenant. You can stop there and keep everything delivered. If you proceed, the AI runs in shadow mode beside your team before a person approves any output it produces.
Thirty minutes on Teams. Bring the step that eats the most hours, and we will tell you whether AI is the right fix or whether a simpler change would do.