Every enterprise in the country is trying to hire AI talent. Most of them are looking for the wrong people, in the wrong places, for the wrong reasons.
What You Need to Know
- The AI talent market is a seller's market, and it will be for years. Competing on salary alone is a losing strategy for NZ enterprises.
- You don't need a team of data scientists. You need domain experts who understand AI, not AI experts who need to learn your domain.
- The most effective AI teams are cross-functional: business analysts + engineers + domain experts. Pure-AI teams build impressive models that nobody uses.
- Internal capability building is faster and more sustainable than external hiring. Your existing team already knows your data, your processes, and your customers.
- The real bottleneck isn't AI talent. It's AI literacy across leadership and operations.
57%
of organisations cite AI skill shortages as a primary adoption challenge
Source: McKinsey & Company, The State of AI in 2022, December 2022
The Hiring Fallacy
The default enterprise response to AI is to hire a data science team. Post job ads for ML engineers, offer above-market salaries, compete with Google and Microsoft for a tiny pool of specialists.
This approach has three problems:
The talent doesn't exist at the scale you need. The global demand for AI talent far outstrips supply, and NZ is competing with markets that can pay 2-3× more. Even if you could hire a world-class AI team, it would take 12-18 months to assemble, and by then the market has shifted again.
Pure-AI teams lack domain context. A brilliant ML engineer who doesn't understand your claims process will build a brilliant model that doesn't fit your claims process. The domain knowledge gap is where most enterprise AI initiatives fail, not the model development.
Hiring is a dependency, not a capability. When key hires leave (and they will, since AI talent is the most mobile in tech), your capability leaves with them. Sustainable AI capability needs to be embedded in your organisation, not concentrated in a few individuals.
What to Build Instead
1. AI Literacy Across the Organisation
Before you hire specialists, invest in AI literacy for your existing leadership and operations teams. Not deep technical training, but practical understanding of what AI can do, what it can't do, and how to identify opportunities.
This means:
- Executive workshops on AI capability and limitation (not vendor demos, but actual learning)
- Operations team exposure to AI tools in their specific domain
- Clear frameworks for identifying and evaluating AI use cases
The goal isn't to turn everyone into a data scientist. It's to give your existing experts enough understanding to recognise where AI helps and to collaborate effectively with AI specialists.
2. Cross-Functional AI Teams
The most effective AI teams we've seen aren't AI teams. They're cross-functional delivery teams that include AI capability. The composition:
- Domain expert (2-3 people): the claims specialists, the operations managers, the advisors who do the work. They define the problem, validate the output, and know the edge cases.
- AI/ML engineer (1-2 people): builds the models, manages the pipeline, handles the technical complexity. This can be in-house or a partner.
- Integration engineer (1-2 people): connects the AI to existing systems. Often the most underestimated role.
- Product/project lead (1 person): coordinates delivery, manages stakeholders, tracks outcomes.
Notice the ratio: more domain experts than AI specialists. That's deliberate. The domain knowledge is what makes enterprise AI enterprise AI.
3. Strategic AI Partnerships
Instead of trying to hire everything, partner for the specialist capability and invest in capability transfer. The right AI partner:
- Builds the complex stuff (model development, knowledge architecture, integration frameworks)
- Transfers knowledge to your team throughout the engagement
- Designs systems your team can maintain and extend
- Reduces their involvement over time as your capability grows
The Capability Transfer Test
Ask any potential AI partner: "What does our team look like after your engagement ends?" If the answer is "you'll need us for ongoing support," you're building a dependency, not a capability.
4. AI Champions, Not AI Departments
Identify 3-5 people across your organisation who are curious, capable, and well-connected within their teams. Give them time, training, and mandate to explore AI applications in their domain. These "AI champions" become the bridge between AI capability and business application.
This is more effective than a centralised AI department because:
- Champions understand their domain deeply
- Champions have existing relationships and credibility with their teams
- Champions can identify opportunities that a centralised team would miss
- Multiple champions create distributed AI literacy
The Realistic Timeline
Building organisational AI capability isn't a quarter-long project. Here's what a realistic timeline looks like:
Months 1-3: AI literacy programme for leadership and key operations staff. Identify AI champions. Establish AI governance basics.
Months 3-6: Run first AI initiative with a cross-functional team, partnering for specialist capability. Champions embedded in the delivery.
Months 6-12: Internal team takes increasing ownership. Champions active across departments. Governance framework maturing. Second AI capability in progress.
Months 12-18: Internal team leads most AI work. Partner involved for complex or novel challenges. AI literacy widespread. Champions leading their domain's AI roadmap.
- What should we budget for AI capability building?
- Allocate roughly 30% of your AI initiative budget to capability building: training, workshops, knowledge transfer. It's the most underinvested area and the one that determines long-term sustainability.
- Should we hire a Chief AI Officer?
- Not yet, for most NZ enterprises. A designated AI lead (often the CTO or a senior operations leader) with a clear mandate and a small cross-functional team is more effective than a C-suite hire. The CAIO role makes sense once you have 3+ AI capabilities in production and need strategic coordination across them.