The AI consulting market in New Zealand is young and fragmented. Some operators have genuine engineering capability. Others are resellers wrapping API calls in a slide deck. Telling the difference before signing a contract is the challenge.
This guide covers what to evaluate, what to avoid, and how to structure an engagement that protects your business while moving fast enough to capture the advantages AI offers.
What should you look for in an AI consultant?
Start with the work. Ask to see production systems, not proofs of concept. A consultant who has shipped AI that handles real traffic, real edge cases, and real users is fundamentally different from one who has built demos. Demos prove familiarity with the tools. Production systems prove the ability to solve problems under constraints.
Look for domain knowledge. AI integration is not just about the model. It is about understanding the business process well enough to know where automation will actually reduce cost or increase revenue. A consultant who asks detailed questions about your workflow before mentioning any technology is a stronger signal than one who leads with the latest model release.
Check for a clear methodology. Good consultants can describe their process: how they scope projects, how they handle uncertainty, how they measure success. Vague promises about "transformation" are a red flag. Specific language about discovery phases, pilot deployments, and success metrics is what you want.
What should you avoid?
Be wary of anyone who leads with the technology rather than the problem. If the first conversation is about GPT-4 or Claude rather than your workflow, that is a signal that the consultant is more interested in the tools than the outcome.
Avoid firms that cannot explain their pricing. AI projects have genuine uncertainty — a responsible consultant acknowledges this. But they should be able to scope a discovery phase and give you a range before committing to a full build. "It depends" is acceptable as an initial answer. "It depends" as the only answer after a detailed briefing is not.
Watch for over-promising. AI is powerful but has real limitations. Any consultant who guarantees specific outcomes without understanding your data, systems, and processes is selling you confidence, not competence. The best consultants are honest about what AI can and cannot do for your specific situation.
How should you structure the engagement?
The most effective structure for a first AI engagement is a phased approach:
Phase 1 — Discovery (1-2 weeks). The consultant examines your workflows, data, and systems. They identify where AI can add value and estimate the return. This phase should be priced as a fixed-fee deliverable with a clear output: a recommendation document.
Phase 2 — Pilot (4-8 weeks). Build and deploy a minimum viable solution for the highest-value use case identified in discovery. Measure results against the baseline. This phase validates the approach before committing to a full rollout.
Phase 3 — Scale. Expand the solution based on pilot results. This is where ongoing development, integration with other systems, and training happen.
This structure protects you in two ways. First, the discovery phase is low-cost and gives you a concrete deliverable regardless of whether you proceed. Second, the pilot phase produces measurable results before you commit to the full investment.
How much should AI consulting cost in New Zealand?
Rates vary significantly. Solo consultants and small firms typically charge $150-$300 per hour, or $5,000-$15,000 for a discovery phase. Larger firms and Big Four consultancies charge $300-$600 per hour, with discovery phases running $20,000-$50,000.
The relevant question is not the hourly rate but the return. A $10,000 discovery phase that identifies $200,000 in annual savings is a strong investment regardless of the daily rate. Focus on the output, not the input.
Be cautious of consultants who only offer large, multi-month engagements. A competent consultant should be willing to start small and earn the larger engagement through demonstrated results.
What questions should you ask before signing?
Five questions that separate strong consultants from weak ones:
- "Can you show me a production AI system you have built?" — Not a demo. A system handling real users.
- "What happened when something went wrong?" — How they handle failure reveals more than how they handle success.
- "How will we measure whether this worked?" — Forces specificity about outcomes.
- "What would you recommend we do not automate?" — Tests whether they prioritise your interests over their billable hours.
- "Who will actually do the work?" — In larger firms, the person who sells the project is rarely the person who builds it.
FAQ
Should I choose a local NZ consultant or an international firm?
Local consultants offer advantages in understanding NZ business context, regulations, and time zones. International firms may have deeper technical specialisation. For most NZ SMEs, a local consultant with proven production experience is the stronger choice. The communication overhead of offshore engagement often erodes any cost savings.
Do I need AI consulting, or can my team learn to do this internally?
Both approaches have merit. If you have developers on staff and the use case is straightforward (e.g., adding a chatbot to your website), internal capability building may be more cost-effective. For complex integrations involving multiple systems, data pipelines, or custom model training, external expertise typically delivers faster and more reliable results.
How long does a typical AI integration project take?
Discovery takes 1-2 weeks. A pilot deployment takes 4-8 weeks. Full production rollout varies from 2-6 months depending on complexity, integrations required, and organisational readiness. Be sceptical of anyone promising production-ready AI in under a month for anything beyond simple use cases.