# GTC 2025: Next Steps for New Zealand Businesses - Turning AI Fatigue into Tangible Outcomes > Move beyond AI hype and 'shiny object syndrome' with practical steps to implement AI solutions that deliver real value for New Zealand businesses following NVIDIA's GTC 2025 conference. **Canonical:** https://infainite.ai/blog/gtc-new-zealand **HTML:** https://infainite.ai/blog/gtc-new-zealand **Source:** src/data/blog/gtc-new-zealand.md --- > **After GTC 2025** > > New Zealand businesses do not need another AI hype cycle. They need practical steps that improve operations, skills, and trust. > > *These seven actions turn conference noise into a plan you can run this year.* GTC is a telescope, not a strategy. This article is a **playbook for NZ organisations**, what to do, what to measure, and what to avoid. ## At a glance | # | Action | NZ angle | |---|--------|----------| | 1 | Realistic AI audit | Start with export, logistics, agri, public, or professional services pain you already measure | | 2 | Data quality | Fragmented ERP + spreadsheets is common, fix access before models | | 3 | Skills | Polytechs + industry partnerships beat hiring-only strategies | | 4 | Collaboration | Small market, pool pilots with peers or research bodies | | 5 | Ethical AI | Trust is export brand; document limitations | | 6 | Government support | Callaghan Innovation, NZTE, R&D incentives | | 7 | Pilot and learn | Publish internal lessons; scale what clears ROI | ## In this article Seven actions. Each includes **why it matters here**, **what to do this quarter**, and **pitfalls** we see in NZ programmes. --- ## 1. Conduct a realistic AI audit **Why it matters in NZ:** Many firms are mid-market, big enough for complexity, small enough that failed “enterprise AI programmes” hurt. **Do this quarter** - List top 10 processes by **cost, delay, or error rate**. - Mark where AI is **assistive** (human decides) vs **automated** (rules + agents). - Pick **one** pilot with a metric you already report (e.g. turnaround hours). **Watch out** - Boiling the ocean with a 24-month roadmap before one proven win. --- ## 2. Prioritise data quality and access **Why it matters in NZ:** Data often lives across Xero, legacy ERP, regional spreadsheets, and partner systems. **Do this quarter** - Name **owners** for critical datasets (customer, product, financial, operational). - Fix **one** silo: e.g. customer view or inventory truth. - Document what may **not** go to a model (privacy, contractual limits). **Watch out** - Buying models before anyone can answer “where is the source of truth?” --- ## 3. Invest in skills development **Why it matters in NZ:** Talent competition is global; local upskilling retains institutional knowledge. **Do this quarter** - Run literacy sessions: prompts, risks, when **not** to use AI. - Sponsor **one** technical upskilling path (data, ML ops, or AI-assisted dev). - Pair with a **polytechnic or university** programme where possible. **Watch out** - Tool licences without training, adoption stalls in month two. --- ## 4. Explore collaborative innovation **Why it matters in NZ:** Shared pilots de-risk cost and surface lessons faster in a compact market. **Do this quarter** - Join or form an **industry working group** on a shared use case. - Engage a **research partner** for evaluation rigour. - Share **non-competitive** components (benchmarks, governance templates). **Watch out** - NDAs so tight that nobody learns from the pilot. --- ## 5. Focus on sustainable and ethical AI **Why it matters in NZ:** Customers and export partners increasingly ask **how** AI is used, not only that it is. **Do this quarter** - Publish internal principles: fairness, transparency, human oversight. - Require **explainability** for decisions that affect people materially. - Log model version, data source, and reviewer for audit trails. **Watch out** - Marketing claims that outrun governance reality. --- ## 6. Leverage government support **Why it matters in NZ:** National programmes exist to offset R&D and internationalisation cost. **Do this quarter** - Map eligibility for **Callaghan Innovation**, **NZTE**, and R&D tax settings. - Align pilot timing with **funding windows** and reporting needs. - Brief leadership on **non-dilutive** support before capital raises. **Watch out** - Grant-led strategy with no product owner inside the business. --- ## 7. Pilot, iterate, and learn **Why it matters in NZ:** Fast feedback loops beat perfect plans, especially when testing across regions or business units. **Do this quarter** - Run a **6 to 8 week** pilot with weekly metrics. - Kill or scale at week 8, no infinite “pilot mode.” - Write a **one-page retrospective**: data, cost, people, next step. **Watch out** - Success theatre (demos without production metrics). --- ## The bottom line | Conference takeaway | NZ execution | |--------------------|--------------| | “AI everywhere” | **One measured workflow** | | “Bigger models” | **Better data + governance** | | “Future platforms” | **Skills + partners + ethics** | GTC informs direction. Your board and customers want **evidence**. Planning AI work this year? [Compare notes with us](/contact) on where agentic and multi-agent patterns fit your sector.