VIVE&BEST
EN中
Contact Us ↗

FINTECH PRODUCTS

Artificial Intelligence
Products

Connect knowledge, data and enterprise systems with AI built around your business.

From intelligent conversations
to business execution.

Bring AI into everyday banking work, from customer service and data analysis to application modernization. Our portfolio combines large language models, knowledge services and automation with banking domain expertise.

AI-POWERED APPLICATION MODERNIZATION

Modernize applications with AI.

Intelligent Code Conversion Platform

Use AI to convert enterprise COBOL and RPG applications to Java, supported by automated testing and generated documentation.

  • Support the complete export, import, conversion, verification and testing workflow.
  • Generate function specifications and design documentation alongside application code.
  • Combine code conversion with end-to-end testing to validate migrated applications.
Explore the migration solution

COBOL / RPG → JAVA

COBOL applicationsRPG applications
AI-assisted conversion

Application code + functional & design documentation

Java applicationsAutomated testing
  1. 01Export
  2. 02Import
  3. 03Convert
  4. 04Verify
  5. 05Test

AI FOR AML ALERT INVESTIGATION

AI-Powered AML Alert Triage

Combine business rules, machine learning and large language models to assess AML alerts, prioritize higher-risk cases and streamline lower-risk reviews. Connect transaction features, relationship networks and supporting evidence so specialists can focus their investigations.

More focused investigations. Consistent alert assessment.

Connect alerts & evidence

  • AML alerts
  • Customer & transaction data
  • External reference data

AI-assisted risk assessment

  • Business rules
  • Machine learning
  • LLM reasoning

Risk scores · Evidence · Priority

Higher-risk alerts

Prioritize the investigation queue and route cases to AML specialists.

Lower-risk alerts

Apply simplified review or rule-based exclusion under bank policy.

Expert review → Feedback → RefinementReviewed outcomes inform rules, features and models.

Give specialists more room to investigate.

High alert volumes

Each alert requires data gathering and assessment, placing pressure on review teams.

Repetitive reviews

Routine, lower-risk activity can repeatedly enter the same manual review process.

Inconsistent judgments

Similar alerts may receive different assessments across reviewers.

Expert capacity constraints

Lower-risk reviews take attention away from cases that need deeper investigation.

Four connected capabilities behind the assessment.

Bring fragmented data and specialist experience into a shared decision process.

Unified data access

  • Connect banking systems and data marts, with external company, legal and tax information where available.
  • Clean, reconcile and standardize data into trusted datasets for analysis.

Risk feature engineering

  • Extract structured transaction patterns, textual signals and transaction-network relationships.
  • Assess feature confidence and filter low-quality signals before downstream analysis.

Natural-language rule generation

  • Translate specialists’ descriptions into structured risk rules and calculate thresholds.
  • Generate executable rule code for publication, turning business experience into reusable logic.

Rules-and-model decisioning

  • Combine multi-source risk scores and supporting evidence with direct rule matches and model-based assessment.
  • Feed expert review outcomes back into training samples, rules and feature refinement.

AI FOR NAME & SANCTIONS SCREENING

AI-Powered Name Screening

Enhance existing name and sanctions screening with large language models, coordinated verification agents and reusable review knowledge. Assess name variants, field meaning and transaction context, then bring supporting evidence into the bank’s established review process.

Go beyond name similarity. Bring evidence into review.

Existing screening workflow

Business systemsTransactions & messages
Screening systemName matching & alerts

Screening and disposition continue through the bank’s established process.

Alerts & messages
Review outcomes

AI screening enhancement

Layered assessment
Rules & validationLLM semanticsVerification agents
Evidence & recommendations
Review workspaceConfirm · Escalate · Record
Parallel integration. Existing workflow continuity.Fallback to the established process if enhancement is unavailable.
  • Parallel integration
  • Multilingual analysis
  • Traceable evidence

Four capabilities. One connected assessment.

Combine configurable checks, semantic understanding, source verification and reviewed knowledge.

01

Configurable rules

  • Check common terms, field roles and entity types, with configurable thresholds and a traceable rule basis.
02

LLM semantic analysis

  • Assess aliases, transliterations and name variants alongside identity details and transaction context.
03

Verification agents

  • Coordinate controlled source checks and relationship analysis; capture source references and evidence for review.
04

Reviewed knowledge reuse

  • Turn reviewed cases into reusable knowledge, rule suggestions and model samples, validated before release.

Deeper checks for unresolved candidates.

Each stage contributes evidence to risk-based routing. Only candidates needing further assessment move to deeper checks.

Alerts + original messagesNormalize · Map fields · Preserve original data
  1. 01

    Rules & field validation

    • Common terms & core names
    • Field roles & entity types
    • Configurable thresholds
  2. 02

    LLM semantic assessment

    • Aliases & transliterations
    • Identity & address context
    • Knowledge-grounded reasoning
  3. 03

    Agent-assisted verification

    • Controlled source checks
    • Related information & evidence
    • Reviewable assessment summary
Combine available evidenceRoute by risk & bank policy
Low risk

Policy-based clearance

Approved false-positive rules support automated clearance with evidence and audit records.

Medium risk

Reviewer confirmation

Present AI recommendations and supporting evidence for authorized staff to confirm.

High risk

Priority expert review

Prioritize specialist and dual review, with escalation under the bank’s procedures.

Routing follows approved bank policy. Authorized staff and existing systems retain disposition control.

Review outcomes → Cases & samples → Validation & approval → Updated rules & models

Understand names in their context.

Support multilingual names, regional addresses and different entity types, including Hong Kong, Macao and ASEAN scenarios.

Name variants

Chan / Chen / Tan

Expand possible transliterations, then cross-check identity details. Spelling variants alone do not establish a match.

Address context

“Jia Yuan 8 Hao”

Identify an address fragment in its original field, rather than treating overlapping text as a personal-name match.

Company names

Trading / Technology / Limited

Reduce the weight of generic terms and compare the core name, registration details and location.

Designed for controlled banking operations.

Private deployment

Process models and transaction data inside the bank’s controlled environment.

Controlled verification

Use authorized external access with only the minimum necessary information, recording sources and times.

Validated rollout

Start with historical replay and shadow comparison, then enable in stages with fallback to existing workflows.

AI FOR AML INVESTIGATIONS & REPORTING

AI-Powered AML Report Generation

Use large language models, coordinated agents and risk-control data to prepare anti-money laundering investigation and suspicious transaction report drafts, with evidence checks and expert review built into the workflow.

AI drafts.
Experts review.

Connect risk data and institutional knowledge with AI-assisted analysis and report preparation.

LLM reasoningSpecialist agentsKnowledge retrieval
Customer &
transaction data
Cases &
regulatory knowledge
Large language modelsCoordinated specialist agents
AML report draftDraft

Customer profile · Transaction analysis
Risk findings · Supporting evidence

Expert review & approval
Illustrative report workflow.

Address the reporting bottlenecks.

Repetitive manual work

Customer data, transaction records and risk indicators take time to gather and organize.

Inconsistent report quality

Different levels of experience can lead to uneven analysis and supporting evidence.

Knowledge gaps

Past investigations and specialist expertise can be difficult to retrieve and reuse.

Demanding review requirements

Report conclusions need clear reasoning and sufficient supporting information.

From source data to an expert-reviewed report.

A connected workflow that keeps specialist judgment at the final review stage.

  1. Integrate & tag data

    • Collect, clean and reconcile source data; apply transaction-level scenario and risk labels.

  2. Analyze transaction risk

    • Calculate risk indicators and use transaction networks to examine relationships, fund flows and patterns.

  3. Retrieve relevant knowledge

    • Combine keyword, vector and knowledge-graph retrieval across past reports, investigation records and regulatory documents.

  4. Generate the report draft

    • Coordinate specialist agents to analyze similar cases and draft the report in a standard structure.

  5. Validate throughout

    • Check missing data before drafting, verify details during generation, and review consistency, facts and references afterward.

  6. Expert review & approval

    • Experts review and approve before submission; approved reports feed back into the case knowledge base.

AI-NATIVE QUALITY ENGINEERING

Intelligent Testing Platform

An AI-native quality engineering platform that connects requirements, test design, execution and reusable knowledge. Generative AI, RAG and knowledge graphs help teams reduce repetitive work and focus on quality decisions.

  • 13+ specialist AI skills
  • Self-healing execution
  • Full lifecycle coverage
  • Traceable knowledge graphs

One connected quality lifecycle.

Generative AI + RAG + knowledge graphs

  1. 01Understand requirements
  2. 02Generate test cases
  3. 03Review coverage
  4. 04Execute tests
  5. 05Analyze & self-heal
  6. 06Build knowledge

Reviewed project knowledge informs the next round of generation and review.

Generate test cases

Turn requirements into structured, business-aware test cases.

  • Import and merge requirements
  • Generate cases for each requirement
  • Ground generation in past cases
Explore capabilities
Multi-format import
Parse PDF, Word, Markdown and Excel through a unified requirements intake.
Large-document analysis
Split long documents into sections, process them in parallel and consolidate a structured requirements list.
Multi-document consolidation
Combine main requirements, change requests and interface documents while retaining key information.
Requirement-by-requirement generation
Generate titles, priorities, preconditions, steps and expected results for individual requirements.
RAG-enhanced generation
Retrieve historical cases and business context to make generated cases relevant to the application.
Outline expansion & export
Expand test outlines using rules, refine cases online and export to Excel.

Review test coverage

Check cases against requirements, with people in the approval loop.

  • Link requirements and cases by ID
  • Find gaps and missing scenarios
  • Route findings for human approval
Explore capabilities
Two-way traceability
Link requirements and test cases using their identifiers rather than keyword matching alone.
RAG gap analysis
Retrieve related cases for each requirement and assess the extent of test coverage.
Multi-step reasoning
Evaluate scenarios, compare coverage and identify gaps through a coordinated review process.
Coverage grading
Classify coverage as high, medium or low and trace each conclusion to its original requirement.
Missing scenario identification
Suggest uncovered exception and boundary scenarios, including locking, concurrency and timeouts.
Human approval workflow
Support review submission, individual approval, batch submission and resubmission.

Execute & self-heal

Generate and run API and UI tests, then investigate failures.

  • Build API and browser test scripts
  • Generate boundary and test data
  • Analyze failures and repair scripts
Explore capabilities
API contract import
Import OpenAPI definitions or cURL requests and use AI to populate interface schemas.
API & UI script generation
Generate API scripts from contracts and cases. Explore real UI pages to identify locators for browser tests.
Script updates
Compare scripts against contract changes, apply natural-language changes and confirm diffs before activation.
AI test data
Generate equivalence-class and boundary-value data, with dynamic unique values for test runs.
Failure analysis & self-healing
Use requests, responses and execution snapshots to identify causes and make local script repairs.
Coordinated execution
Use two execution engines with shared sessions and parameters; consolidate reports, export PDFs and share report links.

Build trusted knowledge

Connect requirements, cases and systems in a reusable GraphRAG foundation.

  • Combine vector and graph retrieval
  • Trace answers to source evidence
  • Reuse reviewed project knowledge
Explore capabilities
Multi-source knowledge ingestion
Ingest requirements, design documents and test cases for both vector and graph-based retrieval.
Knowledge graph construction
Use LLMs to extract entities and relationships into Neo4j, connecting requirements, cases and systems.
Source grounding
Keep original source references with knowledge items so supporting evidence can be inspected.
GraphRAG Q&A
Combine vector retrieval with graph expansion, return citations and let users inspect retrieved passages.
Enterprise knowledge layers
Organize knowledge by customer, system and project, and retrieve both system baselines and customer-specific cases.
Reviewed relationship discovery
Suggest relationships using similarity analysis; store them as organizational knowledge after human confirmation.

A shared AI foundation.

Specialist skills, model orchestration and connected knowledge.

  • LLMs
  • RAG + GraphRAG
  • AI skills & agents
Explore the AI foundation

Unified AI gateway

Separate business workflows from model orchestration through a common service gateway.

13+ specialist skills

Use focused AI skills with standardized JSON input and output contracts.

Vector + graph retrieval

Combine vector search and Neo4j for RAG and GraphRAG on one knowledge foundation.

Flexible model selection

Switch between GLM, DeepSeek and locally deployed models.

Real browser exploration

Use Playwright MCP to inspect actual pages and derive locators from the UI.

Built-in security

Store credentials separately, redact execution snapshots and isolate tenant data.

Faster test design

Reduce repetitive case authoring and review work.

Visible coverage

Trace requirement coverage and identify missing scenarios.

Less script maintenance

Use failure analysis and local repairs to support test upkeep.

Reusable experience

Turn reviewed project knowledge into an organizational asset.

LET’S DISCUSS YOUR NEXT STEP

Bring AI into your banking workflows.

Discuss the business scenarios, knowledge and systems you want to connect.

Talk to Our Experts

VIVE & BEST

Discuss Your Requirements ↗