AI & Machine Learning Company in Auckland

A scoped AI proof of concept at Auckland rates runs NZD $55,000 to $130,000 over six to ten weeks, covering data audit, a baseline model, and a hosted demo. A custom production ML model (fraud scoring, churn prediction, document extraction, demand forecasting) typically lands at NZD $160,000 to $400,000 including MLOps, monitoring, and a Privacy Act 2020 aligned model card. Full production AI systems with RAG, multiple models, fine-tuning, and enterprise integrations range from NZD $350,000 to $1.6M. USD pricing for Auckland SaaS exporters serving offshore customers starts roughly 40 percent lower at our offshore hub rates. Auckland sits above Wellington and Christchurch on cost because of the Xero, Spark, and Trade Me talent premium. We give fixed-fee proposals rather than open T and M estimates.

The Privacy Act 2020, enforced by the Office of the Privacy Commissioner, applies thirteen Information Privacy Principles and adds mandatory breach notification, cross-border transfer restrictions under IPP 12, and Compliance Notices the Commissioner can issue with binding effect. Our discovery phase runs a Privacy Impact Assessment, classifies whether the system makes or supports decisions about identifiable individuals, and locks down cross-border data flows. AI systems that profile, score, or rank New Zealanders trigger heightened transparency expectations under the OPC's published AI guidance and the Government Algorithm Charter. We deliver a model card, an IPP-by-IPP compliance matrix, and a breach response runbook that names the OPC notification path and the 72-hour timeline.

Yes. RealMe is the NZ Government's verified identity service operated by the Department of Internal Affairs, and the Digital Identity Services Trust Framework Act 2023 (DISTF) created an accredited identity provider regime that came into force in 2024. For citizen-facing AI assistants, decision support, and case management tooling, we integrate RealMe assertions, scope the system against DISTF accreditation expectations where relevant, and keep authentication and identity data inside NZ-resident infrastructure. We have patterned deployments after DIA technical guidance and the Government Service Standard, and we coordinate with the agency's privacy officer before any model touches identity data. Outputs feed audit logs that survive OIA requests and Ombudsman reviews.

Maori data sovereignty (rangatiratanga over Maori data) is a live obligation for Crown-funded and iwi-partnered AI work. We follow Te Mana Raraunga Maori Data Sovereignty Network principles, scope iwi consultation into discovery when Maori data, imagery, or whakapapa is in training sets, and propose data governance arrangements that preserve iwi authority over derivative models. For Crown clients we align with the Government Data Investment Plan, the Mana Orite Relationship Agreement frame where it applies, and Te Kahui Raraunga guidance. We do not pretend a global SaaS vendor's defaults satisfy these obligations, and we will tell you when an off-the-shelf US model is the wrong tool for a kaupapa Maori use case.

AWS opened the Auckland Local Zone (ap-southeast-2-akl-1) in December 2023, giving NZ the first hyperscaler-resident compute in the country. We use Auckland Local Zone for inference where latency or Privacy Act 2020 residency strongly favours staying in NZ, AWS ap-southeast-2 (Sydney) for primary training and storage where trans-Tasman is acceptable under a Privacy Impact Assessment, Azure Australia East and GCP australia-southeast1 (Sydney) where client standards specify them. For LLMs that must stay in NZ or AU we self-host Llama 3.3 or Mistral on GPU instances. Cross-border to US or EU regions is acceptable only when a PIA explicitly signs off, typically for non-sensitive internal tooling. Residency lives in the model card and the DPA.

A typical Auckland fintech model (fraud scoring, AML transaction monitoring, credit decisioning, KYC document extraction) takes fourteen to twenty-two weeks from kickoff to production, assuming clean historical data and RBNZ BS11 outsourcing documentation as part of scope. Week 1 to 4 is data audit, PIA, and baseline. Week 5 to 12 is modelling, iteration, and challenger testing. Week 13 to 18 is MLOps, monitoring, shadow mode deployment, and internal audit review. Week 19 onward is gradual rollout under FMA conduct expectations where applicable. If you are pre-data or need labelling, add six to eight weeks. We have shipped to this cadence inside RBNZ-supervised and FMA-licensed stacks.

Yes. The Algorithm Charter for Aotearoa New Zealand, signed by more than two dozen central government agencies, commits signatories to transparency, partnership with Maori, focus on people, data integrity, privacy and ethics, and human oversight in operational algorithms. Our public sector engagements ship with a Charter-aligned algorithm statement, a plain-language explanation suitable for a public-facing page on the agency website, and a human review pathway documented in the operating procedure. We coordinate with Stats NZ guidance, the Government Chief Data Steward, and the agency Privacy Officer. Where the Algorithm Impact Assessment Toolkit is in use, we deliver the AIA artefacts alongside the model card.

Yes, and we design for it on day one. Xero, Pushpay, Eroad, and Vend-style Auckland SaaS exporters serve US, UK, EU, and AU customers, which means GDPR, CCPA, the EU AI Act, and UK ICO obligations stack on top of the Privacy Act 2020. We architect models with regional data isolation, route EU customer inference through AWS eu-west-1 or eu-central-1 with a GDPR-compliant data processing agreement, route US customer inference through us-east-1 or us-west-2, and keep NZ and AU customers inside Sydney or Auckland Local Zone. EU AI Act risk classification ships in the model card from the start so a future Conformity Assessment is not a rewrite.