By Benjamin Baume, Co-Founding father of Craggle
Introduction
In a time when belief within the dwelling mortgage trade is eroding and prospects are navigating ever extra advanced lending landscapes, there’s a rising name for change. In accordance with a latest Deloitte examine, 54% of Australian shoppers say they don’t absolutely belief their financial institution or dealer to behave of their greatest pursuits. But, greater than three-quarters of consumers nonetheless stick with the identical dwelling mortgage supplier even when higher offers can be found elsewhere.
At Craggle, we imagine there’s a greater approach. We’ve developed an AI-driven voice agent that helps Australians uncover tailor-made, unbiased dwelling mortgage choices—with out the stress, complexity, or lengthy wait occasions. However constructing this agent wasn’t only a technical problem; it was a philosophical one too. The query on the coronary heart of our journey was: how can we make voice-based AI interactions really feel pure and humanistic with out ever deceptive the consumer into considering they’re talking with an actual human?
The reply, because it seems, got here from a dialog I had over a number of glasses of whisky with the Shackleton Rescue Committee.
A Dialog Over Scotch: Ethics and Expectations
The Shackleton Rescue Committee is a small however mighty group of execs from various industries, united by a shared love of scotch, the historical past of whisky, innovation, and an ever-changing debate on the highest 10 films of all time. Our backgrounds span a broad spectrum: data expertise (together with software program growth and networking), monetary companies, danger administration, industrial building, rigging and constructing, safety and disaster administration, navy, portray, instructing—to call just some.
We meet recurrently below that tongue-in-cheek title to debate every thing from private development to trade disruption to survival methods (literal and metaphorical). It’s a spot the place no concept is simply too daring to be explored. Conversations typically drift into philosophical territories: what made sure ventures profitable? What did we be taught from failure? What does moral management appear like in a tech-saturated world? It’s additionally a sanctuary—an area the place we help each other throughout completely different phases of life and profession, difficult assumptions and lending our respective skillsets.
Throughout one such session, the subject of Craggle’s AI Agent got here up. I shared that we have been constructing a voice-based generative AI that may converse to dwelling mortgage prospects on the cellphone, serving to them perceive their choices, assess eligibility, and examine offers. We mentioned what this AI ought to be capable to do, the way it ought to reply, and—most significantly—what strains ought to by no means be crossed.
Our group’s resident danger administration skilled chimed in with a passionate objection: “No matter you do, don’t let the AI fake to be human.” And he was proper. It’s a line of deception that’s not simply ethically doubtful—it dangers breaking consumer belief earlier than it even kinds.
The Dilemma: Human Expertise vs Moral Transparency
Right here’s the paradox: for a voice-based AI to achieve success, it should sound pure. Meaning mimicking human conversational patterns—pauses, intonation, empathy cues, adaptive phrasing. If it doesn’t, folks wrestle to interact with it. However the second it sounds too human, we danger creating the phantasm that there is a human on the opposite finish.
The answer isn’t to make the AI sound robotic; it’s to make it clear what it’s, why it exists, and how it helps—with out trickery. The voice agent can nonetheless use humanistic speech patterns and pure language understanding, however with full transparency. We begin every interplay with a greeting that clearly introduces the AI as a digital assistant powered by Craggle, and we reiterate this periodically to maintain expectations clear.
It’s the distinction between respectful design and manipulative design. And it issues.
The Design Method: Conversational AI That Listens Like a Human
One of many largest benefits of generative AI is its means to interpret nuance. Let’s say a consumer says, “I need a mortgage the place I could make additional repayments if I’ve a very good month.” A static type or rule-based bot would miss this totally until the consumer chosen the precise checkbox labelled “further repayments.”
However our voice AI listens like a human. It understands the intention behind the assertion and might match it to the suitable characteristic. This doesn’t simply cut back friction; it empowers folks to precise themselves in their very own phrases, identical to they’d in a pure dialog.
Importantly, we recognised early on that not all prospects converse in monetary phrases. The truth is, many don’t know the technical title of the precise characteristic they need. So it’s not nearly detecting nuance—it’s about enabling folks to explain their wants with out jargon. When prospects say, “I would like to have the ability to redraw a few of my repayments if I would like money later,” CLARA understands they’re speaking a couple of redraw facility. This lifts the extent of transparency by permitting customers to precise themselves naturally and helps cut back the discomfort many really feel after they don’t know the ‘proper’ monetary time period to make use of.
Why is that this essential?
“We don’t ask people to discover ways to speak to bots. We construct bots that know tips on how to hearken to people.” – Benjamin Baume
This precept is on the coronary heart of Craggle’s consumer expertise philosophy. We’re not making an attempt to trick folks. We’re making an attempt to make it simpler for them to make higher monetary choices.
CLARA: Bringing the Imaginative and prescient to Life
A lot of what we’ve realized has culminated within the growth of CLARA—our Craggle Lending and Charge Assistant. Initially constructed as a text-based AI chat expertise, CLARA demonstrated how generative AI might information dwelling mortgage conversations in a pure, personalised, and unbiased approach. Constructing on this basis, we’ve been working to carry CLARA to life by voice—a conversational AI Agent able to real-time help.
CLARA remains to be in growth and present process testing. Whereas not but out there to the general public, the design and testing phases have already revealed the ability and suppleness of generative AI when used to ship clear, human-like interactions. By means of this journey, we’ve been capable of prototype and take a look at how voice-based AI can lengthen the attain of our digital companies, providing purchasers a hands-free, accessible expertise that meets them on their phrases.
Bringing CLARA right into a voice format has meant reimagining not simply how folks speak to AI—however how AI could make advanced processes really feel easy. Actual-time help removes the necessity for limitless form-filling and repetitive inputs. With pure dialog, we cut back not simply paper, however digital friction—changing clicks with easy, direct engagement. Greater than a device, CLARA turns into a information.
And maybe most crucially, CLARA delivers this service with none of the unconscious biases a human might carry. There’s no gross sales stress. There’s no judgment. It’s a constant, unbiased service expertise each time. By permitting customers to maneuver at their very own tempo, and by serving to them construct confidence by readability and transparency, CLARA offers Australians the instruments to make the following step towards their monetary targets—with out worry or friction.
Privateness Boundaries and Buyer Consolation
One other perception from our Shackleton Rescue Committee dialogue was round privateness. I hypothesised that whereas folks could also be hesitant to share private particulars with an AI agent, they’d be extra snug discussing monetary situations. This turned out to be true.
Most members agreed: they wouldn’t need to give their full title, tackle, or e mail to a machine. However they’d be comfortable to say, “I earn about $120K a 12 months, and I’m trying to refinance my dwelling to avoid wasting on repayments.”
This aligns with broader market analysis. A 2024 survey by PwC discovered that 62% of Australians are extra open to digital monetary instruments in the event that they imagine their information is anonymised and used responsibly. When constructing Craggle’s AI Chat Agent, information safety and privateness have been paramount. We developed an automatic PII obfuscation resolution, which permits delicate information to stream by the system with out leaving a hint. This resolution remains to be forward of the curve right now and has been seamlessly built-in into CLARA. The one exception to this obfuscation rule is the consumer’s first title, retained to help personalised interactions and pure human-like engagement.
We’ve additionally launched a number of pure speech options into CLARA based mostly on real-world testing and consumer suggestions. As an example, customers can interrupt CLARA mid-sentence to ask for clarification or shift route—one thing that naturally occurs in human conversations. This helps construct belief and reduces frustration.
CLARA additionally helps dynamic interplay stream. If a consumer needs to start out by speaking a couple of specific concern, like switching from interest-only to principal and curiosity repayments, CLARA will comply with their lead—then circle again to gather the remaining required particulars in a structured approach.
Maybe most spectacular is CLARA’s energetic listening means. Identical to a talented human advisor, CLARA can detect when a bit of data shared in a single a part of the dialog is related elsewhere. If a buyer solutions one query however features a element helpful for one more afterward, CLARA recognises this, shops it, and applies it on the proper time. It’s not simply key phrase or intent matching—it’s contextual recall that evolves with the dialog.
Why This Issues for Dwelling Mortgage Accessibility
For too lengthy, the house mortgage course of has been inaccessible, overly advanced, and riddled with jargon. Many Australians really feel unsure about the place to start or anxious that asking a query might reveal what they don’t know.
Our work with CLARA goals to flip that script. Pure language processing permits prospects to talk in their very own phrases, not in monetary terminology. Actual-time voice interplay replaces friction with stream. CLARA doesn’t push customers towards a specific product or pathway—it guides them with data they will perceive and at a tempo they’re snug with.
That is important not just for transparency, however for inclusion. When the AI listens, interprets, and responds like a human—with out pretending to be one—it ranges the enjoying subject for debtors who’ve felt excluded by conventional techniques. And in doing so, it makes the trail ahead clearer for everybody.
Business Tendencies Supporting AI-Led Conversations
The route of shopper expectations is obvious:
- 80% of shoppers say the expertise an organization offers is as essential as its services or products (Salesforce, 2024).
- 47% of Australians now say they like digital-first interactions for monetary companies, up from 32% in 2021 (Roy Morgan).
- Voice AI adoption in monetary companies is projected to develop by 24% YoY by 2027 (Gartner), particularly in use circumstances the place resolution help is required.
Craggle’s AI isn’t simply maintaining with the pattern—it’s setting a brand new benchmark for moral, humanistic interplay.
Conclusion: A Higher Approach Ahead
Know-how ought to at all times serve folks—not the opposite approach round. At Craggle, we’ve constructed a voice-based AI Agent that does simply that. It listens. It understands. It by no means pretends.
Our objective is to simplify the house mortgage course of and take away the stress and bias that usually comes with it. We imagine that by designing AI that communicates clearly, transparently, and humanistically, we create a greater expertise for all Australians—not simply these fluent in bank-speak.
“It’s not about changing people. It’s about restoring humanity to the method.” – Benjamin Baume
CLARA is extra than simply an AI. She’s the embodiment of a promise: that expertise can be utilized to extend entry, cut back complexity, and empower higher monetary choices—anytime, wherever.
As we proceed to refine and scale Craggle’s voice AI, we maintain tight to the insights from our Scotch-fuelled brainstorming classes: transparency issues, empathy issues, and how we construct is simply as essential as what we construct.
If we will do this proper, then maybe the way forward for lending doesn’t simply look extra environment friendly—however feels extra clear too.