The state of NZ mortgage broking, mid-2026

Market8 May 20263 min readBy The Twio team

A snapshot of where the broker market sits halfway through 2026 — volumes, rates, regulation, and the operational reality on the ground.

Halfway through 2026, the New Zealand mortgage market sits in a recognisable but uncomfortable shape. Rates are off their peak. Volumes have lifted from the 2023 trough. Regulatory expectations have crystallised. And brokers are doing more work per deal than they've ever done.

This post is a short snapshot of where things sit, and what we're hearing from the brokers we work with.

Volumes have recovered, but not evenly

Settled lending volumes through the broker channel have lifted meaningfully from the 2023–2024 floor. The headline numbers look healthy; the texture underneath is more complicated.

First-home buyer activity is the strongest segment, helped by the OCR easing cycle and the lingering effects of the LVR adjustments. Refix volume is high — and remains the steady cash-flow base for most broker businesses. Investor lending is patchy, with appetite varying considerably by lender.

What this means in practice: the brokers we talk to aren't short of conversations. They're short of time to get to all of them.

Compliance has moved from concern to constraint

The 2023 transitional licensing regime feels distant now. The FMA has published two rounds of monitoring observations, and the patterns are clear: the regulator expects each piece of advice to be reasoned, documented, and traceable to a customer-specific assessment. "I always do this" is not a defensible answer.

For most broker businesses, this hasn't required a redesign of the work — but it has required a redesign of the record. File notes that used to be a paragraph are now half a page. Suitability rationale that used to live in the broker's head now has to land on paper, attached to the loan file, before the application goes in.

The administrative weight of this shift is, in our observation, the single biggest input to the four-to-six hours per deal that has become routine.

AI is no longer hypothetical

A year ago, AI in broking was a conference talk. Today it's in the workflow of every broker we know — variably, unevenly, sometimes well, often not. The FMA's 2024 guidance landed at the right time: it set the expectation that adviser tools using AI must be auditable, that the adviser remains the decision-maker, and that customers must be informed where appropriate.

The brokers using AI well tend to share three habits:

  • They use it for structuring (extracting data, drafting, filling) rather than for deciding.
  • They have a clear story for customers about what AI is doing in the workflow.
  • They keep records of AI output the same way they keep records of every other artefact — under their licence, in their file.

The brokers using AI badly are, mostly, copy-pasting from a chatbot into a customer-facing email and hoping for the best. The gap between those two postures is widening fast.

Lenders are tightening operationally, not in policy

Headline lending policy hasn't changed much. But operationally — turnaround times, document tolerances, repricing windows — most lenders have become measurably stricter. A clean application gets a quick yes. A messy one gets a long list of questions and then, often, a no.

This puts a premium on application quality at submission. Brokers who can submit clean files first time are winning lender relationships and turnaround speed. Brokers who treat the application as something to fix during the back-and-forth are losing both.

What this means for broker businesses

The aggregate picture for 2026 H2 looks like this:

  • Demand is there. Conversations are not the bottleneck.
  • Compliance load per deal has structurally risen and will not fall.
  • AI is a viable lever for the administrative half of the work, if used inside a framework the broker controls.
  • Lender operational tolerance is dropping. Quality at first submission matters more than ever.

The brokers who pull ahead from here will be the ones who use the time saved by AI on the work that AI cannot do: judgement, customer relationships, and the sharp end of strategy.

We're biased — we built Twio around precisely this thesis. But the trend would be there with or without us.