The self-service ceiling

The trust gap hiding in services

Digital financial tools have never been more capable. In a recent study we conducted on fintech and open banking sentiments, 97% of our participants said they felt comfortable managing their own finances digitally, and 83% actively preferred it over talking to a person. About what we’d expect.

But there was a catch. The moment a task shifts from paying to planning, between 50-60% went looking for a human. That gap, between what people can do on their own and what they're willing to commit to without reassurance, is where services are quietly losing user confidence.

“I have a tendency to think, especially for bigger value things like our mortgage, I liked being able to discuss that with a person.”

For a $20 coffee card purchase, or even a $1000 dollar investment, sure, people are happy to trust an app to do exactly what they intend. But for an $800,000 mortgage? A business strategy? An investment plan? That's where the cognitive weight of a decision exceeds their confidence in the technology, and they go looking for a human to reassure them and validate their decisions. The problem is, getting to one isn't easy. This is what we've called the trust gap, a space we've inadvertently created by optimising self-service so thoroughly that human support became inaccessible.

‍

‍Why Octave started asking these questions

Over the past 10 years, Octave has worked alongside clients like Payments NZ, Co-op bank, TSB, SBS, and Bizzy (Invoice Hub), helping design and build their websites and tooling. Alongside this we’ve hosted meet-ups and supported start-ups. All this is to say, we’ve been invested in all the recent changes impacting this industry, and have ensured we stay on top of industry change, technology shifts and customer sentiment. Octaves prioritises continuous research and development, and in this moment of historical technological shift we wanted to ask how Kiwis are feeling about their finances and the influx of AI and open banking enabled tools on the market.

With open banking now live in New Zealand and AI embedding itself into every layer of financial tooling, we wanted to understand something specific: how is all of this actually landing with the people using it? Are people becoming more financially capable and confident, or are they feeling increasingly out of their depth?

We surveyed over 100 people and conducted five in-depth interviews, filtering for people who actively manage their household finances using digital tools to reach informed, engaged users that fintech products are built for.

‍

What we found: people avoid help, even when they want and need it

The barriers to accessing human support are high enough that most people just don't bother trying. From our research:

• 38% would only reach out for support after multiple failed attempts
• 9% said they would never reach out to a person at all

This isn't because people dislike talking to humans. People really value a human voice and their expertise for reassurance and advice. They avoid reaching out because the experience of getting through to someone has become genuinely painful.

“To me, it's time. So the majority of us work Monday to Friday, nine to five... I’m sure as heck am not gonna pick up the phone and try and call a bank call-centre during my lunch break with like, 300 other people.”

The result is that humans have been repositioned as a ‘fail state,’ a reactive event, only triggered when self-service falls apart completely. This has consequences, users get frustrated, they distrust the product, and some leave entirely.

“I would like to be able to not use people who don't have people answering the phones. It's just becoming increasingly hard.”

‍‍

The role of AI: optimism with a firm ceiling

The current solution that seems to appear obvious when trying to fill the void humans leave is AI. AI on its own is seen positively. Our research found that 61% of participants were optimistic about AI's role in their financial tools, particularly when it came to saving time and reducing manual admin.

But optimism has clear boundaries. A consistent and vocal minority, roughly 11%, were staunchly opposed, and said they would abandon a product that implemented AI. Their concerns ranged from data security to ethical discomfort.

"In terms of talking to a person, if I have a problem with a bank or insurance and I call them up, I want a person to answer that call. I get any kind of chatbot or AI thing, I lose my mind and go, 'where is the person?'"

What we learnt is: AI is welcome as a tool for efficiency and accessibility, right up until it starts to feel like a barrier to a human. The risk appetite for AI drops sharply when the decision at hand is high-stakes or emotionally weighted. People want AI to help them assemble and interpret their finances, not to stand between them and someone who can guide them through a difficult choice.

‍

A better model: collaborative service

The fix isn't to keep optimising self-service until users feel comfortable managing their own uncertainty. The more valuable opportunity is to make it easier to work with a person. A collaborative service model, rather than a pure self-service one.

There are clear parallels in other industries worth learning from:

• At a self-service checkout, a human is summoned for two specific moments: an error that needs resolving, or a judgement call (like age verification). 
• Self-driving cars initiate a 'Request to Intervene' when encountering edge cases: confusing roadworks, extreme weather, sensor failure. The car handles expected outcomes; the human handles the unexpected. 
• The hospitality industry is adopting ‘invisible concierges:’ Integrated AI, automation, and emerging technology that is streamlining backend decisions so that staff can provide proactive and tailored human services.

These examples point to three transferable principles:

1. Human-tech synchronisation: When technology hands over to a human, it must bring that person up to speed on what happened. No one should have to repeat themselves.
2. Managed oversight: One support person can clear friction across many self-driven experiences.
3. Expert service tiers: Tech handles the administrative. Humans appear as a value-add, not just a fixer, which elevates the overall service experience.
   ‍

‍

How to apply this in your own service

Step 1: Identify your high-impact moments
Map the moments in your user journey that carry either high cognitive load or emotional significance. In finance and health especially, a task can be technically simple but emotionally significant (like reviewing a retirement balance for the first time). These are your intervention points.

When our clients struggle to identify these moments we might look into user behaviour in analytics and testing: Where do users linger? Where do drop-offs spike? Sentiment analysis and session data can surface the friction that user feedback might not.

Step 2: Deploy humans and machines asymmetrically
Machines should handle the high-volume, low-complexity 90%. Humans should ease friction with the low-volume, high-complexity 10%.
The question of how the human appears depends on the nature of the challenge:

• Emotionally weighted tasks: slow things down, build high-trust patterns, make users feel safe. Meaningful friction is reassuring here.
• Technically difficult tasks: apply humans as friction removers. This can actually build trust through what's known as the service recovery paradox: trust increases after a problem is resolved, even when the service caused the problem in the first place.

Step 3: Ensure the human arrives informed
Nothing erodes trust faster than having to repeat yourself to a support person. When a handover happens, technology should prepare a context 'dossier' of what the user was doing, what they tried, and what went wrong. Ideally, this moves toward a co-browsing experience, the user and support person working from the same screen, in real time.

‍

What this means for fintech specifically

At Octave we've seen this play out. We work organisations that represent the highest end of customer service in their fields. We’re currently using the insights discussed in this article to inform our designs in the income and life insurance space with MAS, where people are navigating what they might do in the most stressful moments of their lives. Here we see an example of an emotionally weighted task benefiting from access to a human; when confronting mortality and its impacts, taking a moment to weigh decisions with an expert feels a lot better than getting hurled through to an optimised purchase flow.

In 2026, a frictionless experience isn't enough. People expect a supportive experience, one they can mostly manage on their own, but that won't leave them stranded when the stakes get high. People are optimistic about AI, for assembly of information, not when it gets in the way of a reassuring human voice.

The goal of the modern designer is no longer to build a system so perfect that a human is never needed. It is to build a system so well-integrated that a human can step in exactly when their empathy and judgment adds the most value.