Roadmap
This roadmap is a record of how NoahAI Labs has operated, verified, and expandedits AI asset decision infrastructure in practice.
These are not vague future promises—we separate what is already proven in production, what is actively in progress, and what could expand next.
Achieved (2024–2025)
Completed or currently operating and verified.
Personal financial AI live operation started
We started operating an AI financial decision structure for individual users on real accounts under an always-on monitoring framework. We validated that the pipeline connecting judgment, logging, and risk management works in production and laid the basis for expanding the structure validated in crypto to ETF, equities, and more.
Real-account continuous judgment, logging, and risk management
Guardrails, halt conditions, conservative decisions, full logs, reports, and reproducibility are in live operation.
- • Auditable logs for decision process and outcomes
- • Judgment rationale and replay per XAI policy
- • Anonymized pattern learning and risk signal detection (feedback) in operation
Multi-model comparison and verification
We operate evaluation, leaderboard, and replay systems (e.g. Alpha Arena) that compare and verify models/strategies under the same conditions. We validated multi-engine support (DeepSeek, Qwen, etc.) and version-level performance comparison and rollback.
Multi-exchange concurrent operation
We validated API integration, exception handling, and policy-change response through concurrent operation across six exchanges. We confirmed that per-exchange execution partner structure and judgment–execution separation work stably in production.
Enterprise integration design and operation
We designed the process: connection fit verification (PoC) → API integration → in-platform service expansion. We operate a responsibility structure where NoahAI handles judgment, explanation, risk, and logging; external financial institutions handle execution and product provision. The structure connects to exchanges, banks, brokers, and asset managers at the API and platform level.
In progress (2026)
Aligned with current work described on Enterprise and Technology. As of May 6 2026 we are stabilizing and improving under an always-on monitoring framework (v3.8.9.19). (Menu regression: 51 passed)
Life Finance dashboard tab advancement
Asset allocation diagnosis (allocation bar chart, HHI meter, correlation heatmap), Risk briefing (loss simulation -5% to -50%), Scenario validation (3-strategy database backtests), Security alerts (fraud_detection_service auto-detect), Tax calculation (year-end settlement, capital gains tax, ISA/IRP tax optimization), Korea stock market trends (KOSPI/KOSDAQ + Naver Finance API) completed.
Tax optimization features:
- • Automatic year-end settlement and capital gains tax calculation
- • Tax strategy comparison by account type (ISA/IRP)
- • Annual tax law updates and personalized analysis
Loan product comparison & recommendation engine
We are developing a decision-support system that analyzes user credit rating, loan objectives, and repayment capacity to recommend personalized loan products. We collect and normalize loan product data from banks, capitals, and savings banks to propose safe borrowing structures.
- • Credit rating–based loan availability and interest rate analysis
- • Monthly repayment simulation and total interest comparison
- • Risk assessment–based repayment plan proposals
Insurance product analysis & portfolio optimization
We diagnose insurance portfolio appropriateness and recommend personalized products by analyzing user life stage, family composition, and financial goals. We support rational insurance decisions through duplicate coverage detection, coverage gap analysis, and cost–benefit comparison.
- • Evaluate existing insurance portfolio appropriateness
- • Auto-detect duplicate coverage and coverage gaps
- • Insurance recommendation model by age, income, family composition
- • Optimize coverage level relative to insurance cost
AI Assistant: voice-based decision support
A conversational decision-support system where users speak about their financial situation and judgment, and AI analyzes the current state to help sound decision-making. Rather than simple command execution, it learns user preferences, risk orientation, and decisions to provide personalized advice.
Key features:
- • Voice input: "Should I buy this fund now?" → AI analyzes current assets, market conditions, and user goals
- • Decision explanation: provides reasoning ("why now?" "what's the risk?") alongside conclusions
- • Scenario comparison: visualizes and compares expected outcomes for each choice
- • Position adjustment: "When should I take losses?" → AI proposes optimal timing by combining technical and fundamental signals
- • User preference learning: understands user decision patterns and risk orientation through repeated consultations
ETF / Equities UI and broker API integration
Kiwoom OpenAPI+, Shinhan REST, Mirae Asset REST adapter integration regression tests complete. 14 guardrails passed. Real-order flag remains blocked (false) per safety policy.
Overseas equities and futures judgment structure testing
We are testing application of the same judgment–logging–verification pipeline to overseas equities and futures.
Common judgment pipeline for multi-asset expansion
We are refining multi-asset decision support infrastructure based on the operating engine validated in crypto. We continue to improve exchange order/execution stability and TP/SL and risk guardrail logic.
Open-source LLM–based financial infrastructure
We are advancing operating, verification, and explainability infrastructure to safely apply GPT-4-class or better LLMs to financial decisions; planned for 2026 Q2–Q4.
2026 Vertical AI government program
The technical structure is designed to support finance vertical AI R&D and the 2026 Vertical AI government program; we are pursuing participation.
Future expansion (sequential from 2026)
We reuse and extend the existing judgment pipeline and engine; we are considering sequential expansion around API integration and application. Stated at the level of plans and possibilities—no exaggeration, commitment, or return implications.
Real estate judgment structure expansion
We are considering expansion of AI judgment infrastructure for real estate market analysis and investment support. Included in R&D expansion plans.
Additional asset classes and industry integration
We have asset-neutral architecture to extend to securities, real estate, everyday finance, and other asset classes and industries.
Open-source LLM–based infrastructure refinement
We are considering continued refinement of operating, verification, and explainability infrastructure beyond the 2026 plan.
AI asset decision infrastructure standardization
We are considering standardization and institutionalization of the structure connecting logging, feedback, and verification in the financial domain.
See the structure in detail
Operating stack, industry integration, and the financial AI narrative are on the pages below.