New Zealand insurers don't have a documented, industry-wide claims backlog, but the pressure is real. Volumes are up, disputes are climbing and delay is the common thread. Custom AI can help, but the evidence suggests integration, data and governance decide the outcome, not model capability.
Is there actually a claims backlog in NZ?
Not in any way the public record proves. We found no recent statement from the Insurance Council of New Zealand (ICNZ) or the Natural Hazards Commission describing a sector-wide backlog. We also found thin evidence of a broad insurer move to custom claims AI.
What the evidence does show is a system under strain. That's a more useful story than a headline backlog number.
Where is the pressure showing up?
Three places, and each points at the same problem: slow, hard-to-follow claims.
- Volume. Gallagher's September 2026 NZ Insurance Market Update reports an 8% year-on-year rise in its claim volumes. NZI reported an 11% fall in gross written premium for FY2026 alongside a $53 million increase in natural peril claims. That's more work against tighter margins.
- Disputes. The Insurance and Financial Services Ombudsman (IFSO) accepted a record 600 disputes in 2024-25, more than double its 2022 total. Delays and poor communication were the consistent driver, not coverage disagreements. Insurer disputes lodged with Financial Services Complaints Limited rose from 46 in 2023/24 to 64 in 2025/26, according to Insurance Business NZ.
- Surges. IAG's brands received 33,174 claims from 46 storms between March 2025 and February 2026, up from 9,324 claims over 29 storms the year before (NZ Herald). That's older background, but it shows how quickly capacity gets tested.
Notice what the dispute data says. People aren't mostly angry about being declined. They're angry about waiting and not knowing what's happening.
Why is claims the under-addressed AI opportunity?
Insurers are spending on AI. A KPMG survey reported by Insurance Business NZ found nearly half of AI budgets go to operational and back-office efficiency. Yet only 11% of insurers say they have the data foundation and governance to scale AI. Just 3% are functionally redesigning claims management.
One caveat: New Zealand isn't broken out in that data. Treat it as a directional signal, not a local measurement.
The gap matters. Bolting a tool onto an unchanged claims process tends to speed up one step while the queue moves somewhere else.
What actually limits the return?
Integration. A Finity survey of Australian and New Zealand general insurers (reported by IT Brief NZ) found most have technology strategies but struggle to deliver them. Only 29% reported major gains last year, and integration speed limited returns. It's a trans-Tasman finding, not NZ-only, but it will sound familiar to most claims leaders.
That favours AI built to work alongside existing claims systems rather than replace them. It also explains why generic tools often disappoint. They don't match how a particular insurer triages, requests documents or hands off to assessors.
It's also why vendor ROI claims need care. Several NZ vendors publish large time-saving figures. We haven't seen them independently verified, so ask for baselines, sample sizes and live-claims evidence.
How does NZ governance shape the design?
New Zealand has no AI Act. In practice, that means the Privacy Act applies, alongside the fair conduct duty insurers have owed since March 2025. Public attitudes add a second constraint: Policywise reports only 35% of New Zealanders are comfortable with AI making automated financial decisions.
Adoption looks cautious too. Policywise says only one of the five health insurers it works with discloses using AI to help assess claims. In a GlobalData poll from May 2026, close to one in four respondents said AI isn't ready for widespread insurance use.
The Contracts of Insurance Act 2024 takes effect on 15 November 2027 and will affect underwriting and claims data and workflows. Any system built now should expect that change.
The design implication is clear:
- Keep a person accountable for every claims outcome.
- Use AI to prepare, sort and summarise, not to decline.
- Log what the system saw and suggested, so decisions can be explained to a customer or the ombudsman.
What should an insurer check before building?
A practical order of work:
- Find the delay. Map where claims actually wait: first notification, document chasing, assessor allocation or customer updates. Don't assume it's the decision step.
- Check the data. Can you reliably pull claim notes, emails, photos and policy wording from existing systems?
- Decide the human role. Define what the AI may suggest and who signs off.
- Plan the integration first. Work out how outputs land in the current claims system before choosing a model.
- Set measures upfront. Agree cycle time, rework and complaint baselines, so gains are provable rather than claimed.
- Prototype small. Test on one claim type, such as a storm-damage surge, before scaling.
The same logic applies to anyone in the sector, including training and workforce teams who need to upskill claims staff to work with these tools.
Key takeaways
- The NZ "backlog" is better described as rising volume, surge exposure and delay-driven disputes. No sector-wide backlog statement was found.
- IFSO's record 600 disputes in 2024-25 were driven by delay and poor communication, not coverage disagreements.
- KPMG's survey (not NZ-specific) found only 11% of insurers have the data and governance to scale AI, and 3% are redesigning claims.
- Integration speed limits returns, so AI that works with existing claims systems has the edge.
- With no AI Act, the Privacy Act and fair conduct duty favour human-in-the-loop designs with audit trails.
Our take
The model is rarely the hard part. Most insurers can get a decent document-reading prototype working in weeks, and then stall on system access, messy data and sign-off. We think the sensible NZ approach is narrow and accountable: pick one painful claim type, fit it into the current workflow and keep people in charge of outcomes. The insurers who treat AI as claims redesign, rather than a bolt-on, will have the better complaints data in two years.
If you're looking at your own claims queue, the useful question is where claims wait, not which tool to buy.
FAQ
Do NZ insurers have a claims processing backlog?
There's no verified, sector-wide backlog statement from ICNZ or the Natural Hazards Commission. There is clear pressure: Gallagher reports an 8% rise in claim volumes, and IFSO accepted a record 600 disputes in 2024-25, mostly linked to delay and poor communication.
Are NZ insurers adopting custom AI for claims?
Evidence of a broad shift is thin. Policywise says only one of five health insurers it works with discloses using AI to help assess claims, and NZ insurers have been slower than banks. KPMG's survey shows AI spending is growing, but only 3% are functionally redesigning claims management, and NZ isn't broken out.
Is there an AI law governing insurance claims in NZ?
No. New Zealand has no AI Act, so the Privacy Act applies. Insurers have also owed a fair conduct duty since March 2025. The Contracts of Insurance Act 2024 takes effect on 15 November 2027 and will affect claims data and workflows.
Why does integration matter so much for claims AI?
A Finity survey of Australian and New Zealand general insurers found only 29% reported major technology gains last year, with integration speed limiting returns. AI that can't connect to existing claims systems tends to move delay around rather than remove it.
Can I trust vendor claims about AI cutting claims time?
Treat them as unverified until proven. Ask for baselines, sample sizes and results from live claims, and agree your own measures before a pilot starts.