Choice Hub.

The Challenge

ChoiceHub, a leading Australian financial services comparison platform, faced a critical conversion crisis that threatened their growth trajectory. Despite their success in phone-based sales, their digital platform was hemorrhaging potential customers at an alarming rate.

The numbers painted a stark picture: 75% of users who started the online journey never completed it or spoke to an advisor. This represented millions in lost revenue and thousands of frustrated customers who couldn't navigate the complex world of financial products alone.

The root causes ran deep:

• Static forms created friction - lengthy questionnaires that felt interrogative rather than consultative

• Trust barriers - customers hesitated to share phone details without understanding the value they'd receive

• Experience gap - the quality chasm between their award-winning phone advisors and the sterile online experience

• Legacy system constraints - existing technology couldn't support the dynamic, consultative approach that made their human agents so successful

"Of the people that start the form, we talk to 25%. So we're losing 75% of people we can never talk to…"
ChoiceHub CEO

The company's growth was essentially capped by their inability to digitally replicate the confidence, trust, and personalisation that their best advisors provided over the phone.

The Solution

Rather than simply automating existing processes, we envisioned something revolutionary: Australia's first AI financial advisor that could conduct natural voice conversations while leveraging the institutional knowledge of ChoiceHub's top-performing sales agents.

The solution centreed on three core innovations:

1. Sales Agent DNA Capture

We conducted extensive research with ChoiceHub's highest-performing advisors to understand their techniques:

• Trust-building patterns - how top agents establish credibility in the first 30 seconds

• Dynamic questioning flows - adaptive conversation paths based on customer responses and life stages

• Objection handling playbooks - particularly around complex topics like waiting periods and coverage gaps

• Emotional intelligence cues - adapting tone and pace for logical versus emotional decision-makers

2. Multi-AI Orchestration Architecture

The solution employed multiple specialised AI agents working in concert:

• Conversation Agent - natural voice interaction using ElevenLabs' advanced voice technology

• Policy Analysis Agent - extracting and comparing complex policy documents using Google Gemini

• Recommendation Engine - dynamic matching algorithms that consider individual circumstances

• Transcription Agent - real-time conversation analysis for sentiment and intent detection

• Context Orchestrator - managing seamless handoffs between AI and human agents

3. Omnichannel Continuity Design

Built for flexibility across interaction modes:

• Voice-first approach - natural conversation as the primary interface

• Live transcription - visual conversation flow for accessibility and reference

• Seamless handover - complete context transfer to human agents when needed

• Multi-device support - consistent experience across web, mobile, and future avatar implementations

In additional, the platform was designed as a completely standalone system to avoid legacy constraints, with the strategic vision of launching as a new brand to capture market attention with "Australia's First AI Financial Advisor."

Key Innovations

1. Intent-Driven Dynamic Questioning

Unlike static forms, our AI adapts its questioning strategy based on customer responses and detected intent patterns. For example, if a user mentions "starting a family," the system automatically shifts to pregnancy-focused coverage options and adjusts its recommendation weighting accordingly.

2. Real-Time Policy Document Analysis

The AI can process uploaded policy documents in real-time, extracting key coverage details and identifying gaps or opportunities. This transforms document comparison from a weeks-long research project into a 5-minute conversation.

3. Emotional Intelligence Integration

Using linguistic analysis and conversation pacing, the system detects whether customers are analytically-driven (preferring detailed comparisons) or emotionally-driven (seeking reassurance and simplicity) and adapts its communication style accordingly.

4. Contextual Handover Protocol

When transferring to human agents, the AI provides complete conversation context, sentiment analysis, and preliminary recommendations, enabling agents to continue seamlessly without making customers repeat information.

5. Regulatory Compliance by Design

Built specifically for Australian financial services regulations, with explainable AI reasoning for ACCC transparency requirements and audit trails for regulatory review.

The Results

Coming soon!

We're still testing it out.

Key Technical Decisions

AI-First Product Architecture

Rather than simply automating existing processes, we designed an entirely new product category that leveraged artificial intelligence across every customer touchpoint and business function. This represented a fundamental rethinking of how financial advisory services could operate in the digital age.

Multi-AI Ecosystem Integration

The solution employed seven distinct AI systems working in orchestrated harmony:

Customer Experience AI:

• Dynamic Conversation Engine - Natural voice interaction adapting to customer communication styles and emotional states

• GenAI Interface Cards - Real-time generation of visual comparison elements based on conversation context

• Intent Classification - Sophisticated understanding of customer needs from natural language expressions

Decision Intelligence AI:

• Profile Classification Engine - Automated customer segmentation across seven distinct profiles (family_switcher, baby_planner, senior_saver, etc.)

• Multi-Criteria Recommendation Engine - Advanced scoring algorithms processing 15+ weighted criteria per customer profile

• Conversion Propensity Modelling - Predictive analytics determining likelihood to complete purchase

Operations AI:

• Document Processing Agent - Real-time policy extraction and comparison using Google Gemini

• Provider Network Optimisation - Geospatial clustering for gap-free provider recommendations

Explainable AI Framework

Every recommendation generated a transparent rationale: "Your policy is #1 because it cuts obstetrics wait from 12 → 2 months and costs $4 less per week." This approach satisfied both regulatory requirements and customer trust expectations.

Adaptive Learning Architecture

The platform was designed to evolve from rule-based logic to machine learning without changing customer-facing interfaces. The initial rule matrix could be seamlessly replaced with ML-generated weights while maintaining full explainability for regulatory compliance.

Predictive Behavioural Insights

Beyond basic recommendations, the system anticipated customer actions:

• Disintermediation risk scoring - Identifying customers likely to bypass the platform after receiving quotes

• Next-best-question algorithms - Optimising conversation flow to maximise recommendation confidence

• Voucher allocation intelligence - Dynamic incentive offers based on conversion probability

Why This Worked