Field Notes 21 June 2026 8 min read
AI filmmaking for NZ business: where the tools actually earn their place
After a year of hybrid AI video production for NZ clients, here's what works, what doesn't, and where the economics change.
After a year of hybrid AI video production for NZ clients, here's what works, what doesn't, and where the economics change.
The thing most NZ businesses want from video content is simple: it should look like it was made with intent. Not cinematic, not high-concept, just considered. For a long time, producing that level of quality meant a production budget that excluded most SMEs from the conversation.
AI video generation is changing that calculus, but not uniformly, and not in the ways most coverage suggests. After running hybrid AI production projects for NZ clients across real estate, hospitality, professional services, and B2B over the past year, the clearest thing we can say is this: AI filmmaking tools are genuinely useful for a specific set of production problems, and nearly useless for a different set that often overlaps with what clients initially want to use them for.
Understanding where the line falls is more useful than any tool comparison.
The tools in current production use, Runway, Luma Dream Machine, Sora, Pika, and Kling in particular, can generate short video clips from text or image prompts at significantly improved quality compared to twelve months ago. A ten-second atmospheric clip of waves hitting a rocky coastline. Golden hour light filtering through native bush. A time-lapse of clouds moving across mountain ranges. A close-up of materials arranged on a workbench. For content that is abstract, atmospheric, or conceptual, the quality is now high enough to be usable in professional production.
What these tools don’t do well is anything requiring human subjects, specific recognisable locations, consistent character appearances across more than a few seconds, or verifiable detail. Ask the tool for a wide shot of Wellington Harbour from Petone and you’ll get something that looks vaguely plausible and is immediately wrong to anyone who knows what Wellington Harbour actually looks like. Ask it for a general coastal landscape with native flora and you’ll get something usable.
This distinction is the first thing to understand. AI filmmaking in its current form is a tool for generating generic, atmospheric, or conceptual footage. It is not a tool for producing recognisable representations of specific NZ places or realistic portrayals of real people doing things.
NZ businesses often want video that signals local specificity. A Queenstown adventure tourism operator, a Bay of Plenty kiwifruit exporter, a Wellington architecture firm. Part of what makes their visual identity theirs is the place itself: the light, the landscape, the specificity of geography that reads as New Zealand before anyone has said so.
AI-generated footage struggles here. The tools are trained on global datasets that underrepresent NZ’s specific visual signature. Ask for “rolling green hills with sheep” and you might get something broadly plausible. Ask for “Hawke’s Bay wine country in summer” and you’ll receive a Mediterranean vineyard with different light and the wrong vegetation. The gap is small enough to miss in a quick look, and exactly large enough to feel wrong to anyone who knows.
For brands where local specificity is part of the value proposition, AI-generated footage is not a substitute for sourced or original NZ footage. It fills the production gaps that sourced footage can’t efficiently cover.
A tourism marketing video might open with original drone footage of a specific location, use AI-generated atmospheric B-roll for landscape transitions between sections, and close with real footage again. The AI doesn’t replace the location shoot. It extends it, filling in the atmospheric content that would otherwise require a second shoot day or obvious compromises in the edit.
The strongest case we’ve encountered in practice is for businesses that need regular short-form video content across social channels but don’t have a location-specific visual requirement for every piece.
A professional services firm posting weekly to LinkedIn doesn’t need to show its actual office or a recognisable Auckland streetscape. It needs professional, relevant, non-generic visuals that frame the content. Abstract footage of documents, collaborative spaces, city light at golden hour. At NZD $40 to $80 per month for a Runway subscription and around an hour of generation time per week, a business can maintain a library of B-roll assets for social video at a fraction of what sourcing stock footage or commissioning a shoot would cost.
For businesses already running a structured social media marketing programme, AI-generated B-roll removes one of the most persistent production bottlenecks: sourcing visual content that is relevant and not recognisably generic stock. A consistently posting LinkedIn or Instagram account needs new visual material every week. A library of generated atmospheric footage, refreshed monthly with a generation session, solves that supply problem without a proportionate increase in cost.
Real estate is a more nuanced case. AI-generated footage cannot substitute for actual property walkthroughs, which are what buyers need to make decisions. But it can generate atmospheric lifestyle footage for property marketing materials and background content for agent profile videos. Paired with AI avatar production, a real estate agent can produce a polished branded property update video in under two hours without a crew: the avatar handles the presenter role, AI-generated lifestyle footage provides the property context, and sourced photography covers the actual listing.
The hospitality sector has a specific and serviceable need. Restaurants, cafes, and accommodation providers all need atmospheric content: food in good light, room textures, landscape settings. Much of that content is about feeling rather than specificity. AI-generated footage can cover a meaningful proportion of the ambient visuals a hospitality brand needs without the logistics of a food shoot or location session.
What’s become clear from client work is that the most effective use of AI filmmaking is in hybrid production: combining AI-generated footage with real NZ footage, AI avatar presenters, and edited graphic elements. No single element has to carry the entire weight of realism. The AI elements handle atmospheric, abstract, or transitional sections. Human footage handles the specific and the personal.
This changes the economics of a production without compromising the parts that actually carry credibility. A client testimonial needs to be real. A brand story needs to feature actual people and places. But the transitions, the title cards, the B-roll cutaways, the background footage in a branded video, all of that can come from a generation workflow rather than a shoot.
For the AI filmmaking projects we produce with clients, the planning step is now a clear question: which elements of this production require reality, and which require quality? The first category demands camera crews, sourced NZ footage, and real environments. The second is where generated content earns its place in the budget.
The combination of AI-generated B-roll, avatar-driven presentation, and purpose-built AI automation to move content from script to platform is how the production cost of a weekly branded video drops from a day of crew work to a focused two-hour workflow.
People remain the primary limitation. AI video tools generate humans that look approximately correct at low resolution and obviously wrong when anyone pays close attention. Hands in particular are still a consistent tell. NZ clients who want video featuring recognisable human subjects in natural interaction, staff profiles, customer service explainers, team culture content, need human subjects and real production. Generated human footage at this stage creates an uncanny quality that works against the brand rather than for it.
Narrative continuity is the second limit. A ten-second atmospheric clip is manageable. A two-minute video with consistent characters, environments, and a story that holds together across it requires either a highly skilled prompting process with significant iteration time, or real production. The effort required to generate a coherent two-minute narrative from AI footage alone often exceeds the effort of organising a focused real-world shoot.
The third limit is any content where the audience has motivation to verify what they’re seeing. A claims-bearing testimonial, a product demonstration where the viewer wants to check real-world performance, a news context. Generated footage introduces doubt in those situations that undermines the content’s purpose regardless of visual quality.
For most NZ businesses approaching this for the first time, the right question isn’t “can we use AI filmmaking?” It’s “which specific production problems does AI filmmaking solve for us at better cost than the alternatives?”
For businesses with a regular social content cadence and no strong location specificity requirement, AI-generated B-roll is ready to use now. A Runway or Luma subscription at NZD $50 to $80 per month, a defined generation workflow, and integration into a broader content pipeline can reduce the time cost of visual content for social to a manageable weekly session.
For businesses where NZ specificity matters, the better entry point is identifying which elements must be shot or sourced, and which can be generated. That hybrid approach reduces production cost while protecting the elements that carry the credibility signal.
The tool set is improving at a rate that makes small, regular experiments more valuable than a one-time assessment. What isn’t usable at professional quality today may well clear that bar in three or four months. Businesses that have already established even a simple AI production workflow will incorporate those improvements faster than those starting from scratch, because they already know which production problems they’re trying to solve and have a workflow ready to plug new capabilities into.
From automated guest comms to multilingual booking content, here's where the tools are earning their place.
The tools getting traction in real estate and property management look nothing like the early hype.
NZ accountants, lawyers, and advisers carry heavy admin loads. Here's where automation is already paying off.