Gen AI Reshaping the Asset Management Industry: Singapore Focuses on Investment-Driven Approach, Vertical Tools Become Key

Overview

This report is based on Bernstein's Future Tech Summit in Singapore, focusing on the application of generative AI in asset management. The core view is: although Singapore firms lag in AI maturity, their 'business-driven' model led by investment teams is rapidly advancing high-value use cases. Meanwhile, general chatbots cannot meet the deep needs of the financial industry; vertical platforms with data integration, security governance, and domain expertise (like LinqAlpha) will become core competitive strengths. The future of asset management will move toward a new paradigm of 'human-machine collaboration,' deeply integrating human judgment with machine automation.

Core views

Singapore's asset management institutions exhibit a unique 'investment-led' characteristic in adopting generative AI, with 57% of projects driven by investment teams, contrasting sharply with London (61%) and Hong Kong (52%), where technology teams dominate. While there is a maturity gap—only 24% of use cases have reached production stage, far below London/Hong Kong's 51%/52%—Singapore's pilot adoption rate (41%) leads, indicating a critical transition from exploration to scaled implementation. In terms of technical approach, Singapore firms heavily favor purchasing third-party solutions (68%), far exceeding self-build or hybrid models, with users typically using only 1-2 tools (e.g., Copilot, ChatGPT, Claude). This reflects strong market demand for plug-and-play, low-barrier solutions. However, this adoption has not delivered deep value, as general large models face three bottlenecks in handling financial data: first, data sources are fragmented and formats inconsistent, with different vendors (e.g., S&P, FactSet) having varying standards and citation rules, which general models cannot effectively reconcile; second, costs are uncontrollable, with direct API calls incurring high token fees; third, there is a lack of security control and permission management for sensitive internal data. Thus, vertical financial AI platforms (like LinqAlpha) have emerged. Their core value lies not in the models themselves but in building a 'smart pipeline': they first unify, clean, and normalize vast external and internal data (including emails, reports, transaction records) and establish a unified knowledge graph containing metadata like asset master data and financial calendars. Only then does the platform enable users to conduct cross-asset, cross-market deep queries. For example, when a major market event occurs, the system can automatically identify all affected companies, even if not explicitly mentioned. This 'connecting the gaps' capability is the future source of alpha. Ultimately, the future asset management architecture will not be isolated applications but a few deeply embedded 'intelligent agents' within organizational contexts. They are built by central teams but grant investment managers great flexibility to customize workflows. This means the role of fund managers will shift from information processors to 'system architects,' whose core responsibilities are defining signals, managing uncertainty, and intervening when machines fail. In this new paradigm, the core competency is no longer programming but 'judgment'—discernment in vast information, timing consensus errors, and decision-making under uncertainty.

Analysis framework

The report adopts a typical 'field observation + expert interviews + case extrapolation' analytical framework. First, through summits in Singapore, London, and Hong Kong, extensive first-hand voting data was collected, quantifying regional differences in AI adoption levels, leadership departments, usage frequency, and tool choices. Second, two senior professionals from J.P. Morgan and LinqAlpha were invited as panelists to delve into strategic planning, specific use cases (e.g., research assistants, intelligent monitoring), technical architecture (e.g., model-'engine' separation), and business models (e.g., output-based pricing). Finally, by dissecting these real cases, the report reveals the fundamental differences between general models and vertical platforms and prospectively outlines the 'human-machine collaboration' future. The analysis is logically clear, progressing from observed phenomena to underlying causes and future trends.

Methodology notes

• Industry Analysis FrameworkSupply-demand framework

  The financial industry's demand for AI is not simply efficiency improvement but a deeper need for 'data connectivity' and 'decision loops,' forming the fundamental driver for vertical platforms.

  This framework posits that any industry's core contradiction stems from supply-demand mismatches. In finance, suppliers (e.g., data providers) operate in silos, creating 'data islands,' while demanders (e.g., fund managers) need a unified view that connects all data sources for coherent insights. This 'connectivity' imbalance is the underlying logic driving vertical platforms.

• Competition & Strategy FrameworkMoat / competitive advantage

  Vertical platforms' competitive advantage lies not in model performance but in their 'data infrastructure' and 'organizational context' accumulation.

  The moat theory holds that long-term competitive advantage comes from hard-to-replicate barriers. For LinqAlpha, its moat is not algorithms but its years of accumulated, analyst-verified cross-source data standardization system and deep understanding of clients' internal investment philosophies, portfolios, etc. These are assets general models cannot easily replicate.

• Cycle & Sentiment FrameworkInflection Point Analysis

  Generative AI in finance is currently at an inflection point from 'proof of concept' to 'scaled production.'

  Inflection point analysis focuses on key turning points in industry lifecycles. The report notes that while hundreds of use cases were initially conceived, the focus has now narrowed to 50 high-return strategic priorities, marking the industry's shift from 'broad exploration' to 'selective implementation,' a typical upward inflection.

• Valuation methodsSOTP Sum-of-the-Parts Valuation

  Assessing vertical platforms' value requires splitting into: base service fees + output-based pricing for premium services.

  SOTP (sum-of-the-parts) valuation suits firms with multiple independent business lines. The report mentions LinqAlpha's revenue structure evolving from 'fixed platform fees + usage overages' to 'fixed platform fees + output-based pricing for advanced agents,' requiring investors to value it as a composite of basic functions and high-end intelligent services.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• LinqAlpha

  As a vertical financial AI platform, its core value lies in addressing general models' fundamental flaws in data integration, security, and context fusion, making it key to achieving high-order intelligence.

  Strengths

  Strong data standardization and cross-source integration capabilities; built-in security and permission management; deep integration with client internal contexts; strong product narrative and user reputation.

  Weaknesses

  Reliance on continuous high-quality data inputs; high client stickiness but potentially high new client acquisition costs.

  Comparison

  Compared to general models (e.g., Claude), LinqAlpha has overwhelming advantages in data accuracy, security, and domain depth; compared to other vertical platforms, it excels in product ecosystem and user growth.

  Risks

  Failure to maintain data quality and partnerships could lead to client attrition; intensifying competition may trigger price wars.

• J.P. Morgan Asset Management

  As an industry pioneer, its 'LLM Suite' platform enables large-scale internal empowerment, making it the best case validating 'democratizing AI.'

  Strengths

  Over 250,000 global employees using the platform; mature agile process from ideation to deployment; successful high-value use cases like 'research assistant' and 'proxy voting.'

  Weaknesses

  Large size slows decision-making and change speed; high internal IP protection requirements limit external collaboration flexibility.

  Comparison

  Compared to emerging vertical platforms, its advantage lies in resources and scale; but it may trail in flexibility and innovation speed.

  Risks

  Failure to consistently demonstrate high ROI may face internal budget pressure.

Key data

• Proportion of AI project leaders in Singapore firms57%Higher than London (61%) and Hong Kong (52%), showing investment team-led characteristics.
• Proportion of use cases in production stage in Singapore firms24%Significantly lower than London (51%) and Hong Kong (52%), reflecting relative maturity lag.
• Proportion of Singapore firms purchasing third-party solutions68%Far exceeds self-build or hybrid models, making it the mainstream choice.
• Average number of AI tools used by Singapore firms1-2Far fewer than London and Hong Kong's 3-5, showing highly concentrated tool usage.
• Top 3 horizontal general AI toolsCopilot (33%), ChatGPT (26%), Claude (23%)Adoption rates are largely consistent across Singapore, London, and Hong Kong.

Impact & implications

For the asset management industry, the meeting reveals two core impacts: First, the divergence in technical paths is now certain. General large models may continue to dominate in basic productivity tools, but 'high-order intelligence' that determines investment performance will be provided by vertical platforms. Second, talent roles will fundamentally shift. Future top fund managers will no longer need to be information gatherers but must be 'system architects' capable of designing, managing, and overseeing 'intelligent agents.' This will force the industry to redefine training systems and hiring standards. For investors, this means investment targets should not be selected based solely on traditional metrics but also on whether they possess robust internal data infrastructure and deep integration with vertical platforms.

Risks

• General large models in finance may face data confusion and compliance risks, leading to misjudgments.
• Vertical platforms' heavy reliance on internal data means output quality directly suffers if data sources are disrupted or degrade.
• Rapid technological change may render existing investments obsolete quickly.

What to watch

• How vertical platforms solve the 'output-based pricing' business model challenge and whether their pricing models gain market acceptance.
• Whether large asset managers can maintain efficiency while avoiding bureaucracy to truly achieve 'democratizing AI.'
• Whether regulators will issue specialized guidelines for generative AI use in finance.