The Law Society of Singapore
AI-Native Legal Member Services
A professional body serving Singapore's legal community needed a governed AI-native service layer that could answer questions across 1,000+ approved sources, 143 knowledge folders, confidential legal materials, web, and mobile.
legal professionals served
approved data sources
avg response, down from days
legal answer accuracy
Legal knowledge support at institutional scale.
AI-native member service layer
<6s
routine answer time
1K+
approved sources
3
access platforms
>90%
validated accuracy
The Law Society of Singapore supports thousands of lawyers with professional resources, ethics guidance, member services, and operational support. Member questions were not low-stakes support tickets. They often needed accuracy, context, and source discipline.
Before JusAsk, answering routine inquiries required internal routing across departments and knowledge sources. The average response cycle could stretch to around two days, even when the member needed a straightforward answer from approved materials.
REPCONN built JusAsk as an AI-native service layer: one governed answer engine made available through web, iOS, and Android, with secure retrieval, source grounding, escalation paths, operating visibility, and controls for confidential data.
Five constraints shaped the system.
Member support served a national legal community
Questions came from a large professional audience that includes lawyers, legal practitioners, notaries, judges, and Singapore government stakeholders. Routine inquiries could still take around two days before a usable answer reached the member.
Knowledge lived across 143 folders
More than 1,000 approved sources had to be structured across legislation, practice materials, internal documents, guidelines, and operational references without flattening access rules.
Legal answers needed source discipline
Members needed responses grounded in approved materials, with clear source context and escalation when the question required human judgment or a restricted reference.
Confidential AML/CFT data required strict controls
The system had to process sensitive legal and regulatory materials, including confidential anti-money laundering and counter-terrorism financing references, under disciplined access, audit, and review workflows.
Access had to work beyond the website
Members needed a consistent experience across web, iOS, and Android, while the organization needed one governed answer layer behind every interface.
Our Solution
One answer layer across every member touchpoint.
REPCONN built JusAsk as a governed legal knowledge system. The operating layer connects approved sources, answer routing, mobile access, security controls, analytics, and human review into one managed service.
System 01
Secure Knowledge Ingestion
REPCONN structured LawSoc's approved knowledge sources into a governed retrieval layer across 143 folders. Public materials, member-restricted resources, internal guidance, legislation, and sensitive AML/CFT references were separated before answer behavior was configured.
What changed
1,000+ sources mapped by access level and use case
Approved documents prepared for retrieval and citation
Sensitive references separated from general member answers
Foundation created for a governed legal answer layer
Member service dashboard
7K+
served
<6s
routine answer
3
platforms
>90%
accuracy
Legislation lookup
Source cited
Answered
CPD guidance
<6s response
Answered
Member-specific issue
Human review
Escalated
AML/CFT reference
Policy guardrail
Restricted
System 02
Sourced Answer Engine
JusAsk was built to answer routine member inquiries in less than six seconds while preserving legal trust. The system retrieves relevant materials, drafts a response, cites its source context, and routes unclear or sensitive questions to review.
What changed
Routine questions moved from a 2-day average to under 6 seconds
Responses grounded in approved LawSoc materials
Clarification flow for ambiguous member questions
Escalation path for sensitive or high-risk queries
Answer routing
Sample cases
System 03
Web, iOS, and Android Access
The answer layer was deployed across the existing web environment and native mobile applications. Members could use the same governed service from the channels most natural to their workflow.
What changed
Web experience embedded into the existing website
Native iOS application delivered for mobile access
Native Android application delivered in parallel
Consistent answer behavior across all platforms
Embedded in the member website
Members can ask questions directly from the LawSoc web experience without leaving the existing WordPress environment.
System 04
Governance and Human Review
REPCONN designed the operating model so AI handles routine work and humans remain in control of the exceptions. Sensitive topics, low-confidence answers, restricted references, and confidential AML/CFT materials can be reviewed before the member experience is affected.
What changed
Flagged queries routed for human review
Confidence and source behavior monitored
Administrative access protected with role controls
PDPA and ISO 27001-aligned controls support operational review
Secure knowledge loop
Classify
Public, member-restricted, internal, and confidential AML/CFT sources are separated before retrieval behavior is configured.
System 05
Analytics and Continuous Improvement
The system gives the organization visibility into usage, query types, response quality, escalation patterns, and knowledge gaps. This turns member questions into a continuous improvement loop.
What changed
Usage patterns visible across channels
Response quality reviewed against member needs
Knowledge gaps surfaced from repeated questions
Operating layer improved after launch
Member service dashboard
7K+
served
<6s
routine answer
3
platforms
>90%
accuracy
Legislation lookup
Source cited
Answered
CPD guidance
<6s response
Answered
Member-specific issue
Human review
Escalated
AML/CFT reference
Policy guardrail
Restricted
From executive approval to ongoing production governance.
The project began in October 2024 and is still active. What started as a governed member-query answer layer became a broader legal knowledge operation with secure data ingestion, mobile access, CRM and SSO integration, legal analysis, contract review, audits, and ongoing model governance.
Oct 2024
Approval
Contract approved by executive management
The engagement was approved by Law Society executive management, including President Ms. Lisa Sam Hui Min, establishing JusAsk as a strategic AI-native member-service initiative.
Nov 2024
Development
Initial build using leading Google and OpenAI models
Development started with the strongest available Gemini and GPT models at the time, focused on governed retrieval, sourced answers, web access, mobile delivery, and secure handling of approved LawSoc knowledge.
Jan 2025
UAT
User acceptance testing and fine tuning
REPCONN coordinated testing with Director and Data Protection Officer Joseph Tay and the LawSoc IT Department, tuning answer behavior, source grounding, permissions, and escalation logic.
Feb 2025
Pre-production
Pre-production validation and sign-offs
The system moved through pre-production checks, operational review, security validation, and final sign-offs before production readiness.
Mar 2025
Phase 1
Production-ready status
Phase 1 closed with JusAsk production-ready: a governed member-query answer layer operating across web, iOS, and Android with source grounding and human review paths.
Apr-Aug 2025
Phase 2
Expanded data, UX, SSO, CRM, and legal analysis
Phase 2 expanded the product beyond member questions. REPCONN added more data ingestion, UX/UI improvements, CRM integrations with SSO, legal analysis workflows, and contract review capabilities.
Sep 2025-Jan 2026
Phase 3
Maintenance, audits, and model governance
The program moved into ongoing maintenance, ISO and PDPA audit support, operational tuning, and model governance. After evaluating model options, including Chinese model providers such as DeepSeek, REPCONN recommended US-based Anthropic models for security, governance, and enterprise-grade reliability.
Jan-Jun 2026
Current
Anthropic migration and ongoing optimization
The production model stack moved from Google and OpenAI models to Anthropic models in January 2026. In June 2026, REPCONN evaluated Fable 5 access, then switched to Opus 4.8 after external access restrictions changed availability.
Early usage showed real member demand.
Avg response, down from days
Routine legal knowledge questions moved from an estimated 2-day average response cycle to less than 6 seconds for approved-source answers.
Legal answer accuracy
The governed answer layer now operates above 90% legal accuracy for validated, approved-source member-service questions, with ongoing review and tuning.
Data sources governed
The system searches and retrieves from more than 1,000 approved sources, including legislation, internal documents, guidelines, and practice materials.
Knowledge folders structured
REPCONN organized the source base into 143 governed folders so the service can retrieve knowledge without losing permissions or context.
Secure processing built into the answer layer.
Legal member services require more than fast answers. JusAsk was structured so knowledge access, retrieval behavior, administrative permissions, confidential AML/CFT materials, and review workflows operate under clear controls aligned with PDPA expectations and ISO 27001-style practices.
Secure knowledge loop
Classify
Public, member-restricted, internal, and confidential AML/CFT sources are separated before retrieval behavior is configured.
Tiered knowledge access
Public, member-restricted, internal, and confidential AML/CFT references are treated differently before retrieval behavior is configured.
Data masking and minimization
Sensitive identifiers, member-specific context, and restricted references can be masked, excluded, or routed away from model-facing workflows.
Role-based administration
Administrative access is limited to authorized users, with higher-risk functions kept behind stricter controls and review paths.
PDPA and ISO alignment
Data handling was designed around PDPA expectations, auditability, access logging, and ISO 27001-aligned security practices.
JusAsk is a governed member-service operation that runs across every access channel, from public-facing questions to sensitive knowledge workflows.
Faster member service without losing institutional control.
6s avg response
Routine answers moved from an estimated 2-day response cycle to less than six seconds.
JusAsk gave LawSoc a practical way to improve member support without asking every question to pass through a manual departmental queue. Routine answers moved from a multi-day cycle to less than six seconds, while sensitive questions remained governed by review and access rules.
The value was not just speed. The system created a more visible member-service layer across more than 1,000 sources: what members ask, where knowledge gaps appear, what requires escalation, and which materials need improvement.
REPCONN designed the layer so LawSoc could adopt AI in a controlled, member-focused way, with approved knowledge, source discipline, secure access, and cross-platform availability.
Professional bodies hold valuable institutional knowledge, but members experience that knowledge through service channels. When those channels are slow, the knowledge is less useful than it should be.
An AI-native service layer makes approved knowledge more accessible without removing governance. That is the difference between a simple answer interface and an operating system for member support.
Confidentiality Notice
Specific internal workflows, data sources, confidential AML/CFT materials, system configurations, and access-control details are not published in this case study. Metrics represent rounded public-facing outcomes from operation through 2026.
Get Started
Want an operating layer like this in your organization?
We assess your current operations, identify where AI-native systems will have the highest impact, and build and manage the entire layer for you.