AI Onboarding & Document Intelligence Case Study
Client | Confidential
Engagement Overview
AJQ Technology designed and built an end-to-end digital application platform for a leading insurance and wealth management provider, spanning governed client-facing onboarding forms, automated document intelligence, and an AI-driven orchestration layer proving out unified, cross-system case handling.
The engagement covers the full lifecycle of a client application, covering multi-step digital forms with mandatory-field validation, automated extraction and structuring of supporting documents via AI-powered OCR, and a governed review pipeline giving the client’s operations team full visibility and control over every submission, hold-for-review, approve, reject, or return, all fully audited.
AJQ’s role spanned solution architecture, platform build, and delivery of a live proof-of-concept demonstrating how an AI assistant can operate across multiple backend systems using the Model Context Protocol (MCP), presented directly to the client and validated in a live demonstration ahead of the next phase of build.
Scope & Complexity
The platform operates in a highly regulated financial services environment, with data residency and auditability as first-order requirements throughout
- Governed digital forms with structured submission, validation, and a full approve/reject/hold-for-review workflow
- AI-powered document OCR extracting structured data from client-submitted documents, replacing manual keying
- Strict AU-only AI inference, every model call runs through a private, region-locked endpoint with no public internet path, giving the client a genuine data-residency guarantee rather than a policy promise
- A working proof-of-concept for unified AI case orchestration, using the Model Context Protocol to demonstrate an assistant that can reason and act across multiple backend systems from a single conversational interface
- JWT-based authentication throughout, with Microsoft Entra ID federated single sign-on for staff, enterprise users sign in with their existing corporate credentials, inheriting the client’s own MFA and conditional-access policies rather than a separate password to manage
- A zero-downtime identity migration pattern, where a dual-pool authentication bridge validates tokens from either the current or an incoming identity provider during a cutover, so staff access is never interrupted while an identity platform changes underneath them
- Separate, least-privilege access boundaries for staff and external clients, with role enforced from signed token claims at both the gateway and application layer, and every submission and decision captured in an immutable audit trail
AJQ Approach
- AWS
- Microsoft Entra ID
- CloudFront
- Model Context Protocol (MCP)
- React-based client and admin experiences
- Delivery Model & Ways of Working
AJQ’s platform team worked in tight, demo-driven iterations, building, validating, and showing working software back to the client at each step rather than working from a fixed up-front spec. Solution architecture was engaged from the outset to ensure the platform’s data-residency and audit posture would satisfy the client’s own regulatory obligations before a single line of application code was written.
The proof-of-concept phase was deliberately scoped to prove the hardest part first, which was whether an AI assistant could safely and usefully orchestrate real actions across a client’s real systems. Validating that early, in a live demonstration, gave the client concrete evidence to build the next phase’s business case on.
Outcomes Achieved
- A live, working platform handling real client submissions end-to-end, from digital form to reviewed, actioned outcome
- Automated document data extraction removing manual keying effort from the operations team
- A verifiable, private AI inference architecture meeting the client’s data-residency requirements
- A validated proof-of-concept for AI-driven, cross-system case orchestration, directly informing the client’s decision to proceed to full build