11-20 更新 实时
首页/ 技术栈/ 正文

Databricks 获 A-Team 资本市场 AI 奖

技术栈 来源:Databricks Blog 2026-11-19 11:54 0 阅读 机器翻译
Databricks 获 A-Team 资本市场 AI 奖

领先机构已经在该平台之上构建应用。RBC Capital Markets 借助其 Aiden QuickTakes AI 智能体,将研究报告草稿的生成速度提升最高 95%;LSEG 则展示了 AI 智能体用自然语言筛选 35 万+ 项投资。

因 AI 与治理能力同时获得认可。

资本市场从不缺乏 AI 相关的宣称。每家厂商的推介材料都承诺智能、自动化和超额收益(alpha)。但 A-Team Group——其奖项由资本市场行业的从业者投票产生——选择认可另一种东西:在同一个平台上同时实现 AI 与治理。

这一区别很重要。

“祝贺 Databricks 在 2026 年第二届资本市场 AI 大奖(AI in Capital Markets Awards 2026)中赢得数据智能与治理最佳 AI 解决方案奖。该奖项由我们的行业社群评选,彰显了在资本市场推动 AI 能力方面的真正卓越与创新。Databricks 理应为这一成就深感自豪。”——A-Team Group 首席执行官 Angela Wilbraham

这是 Databricks 首次获得 A-Team 资本市场 AI 奖项。它到来的时机,正值行业在实验性 AI 与能够经受监管审查的生产级 AI 之间划出明确界线。

为什么治理才是资本市场 AI 真正的解锁点

资本市场机构面临的压力十分严峻。日内波动率仍处高位。北美的 T+1 结算正在压缩运营窗口期。监管机构正不断收紧对 AI 治理、模型透明度和数据血缘的期望——从 ESMA 关于算法交易的监管简报,到 FSB 2026 年关于 AI 稳健实践的工作计划。

与此同时,一项 2026 年针对管理约 2 万亿美元资产的管理人的调查显示,94% 的投资组合经理和分析师已在替代数据研究流程中使用 AI 或机器学习。但 71% 的人表示,整合来自不同来源的数据是他们最头疼的难题。

瓶颈不在于能否获取 AI,而在于其底层的数据基础。

大语言模型功能强大,但它们无法看到你公司专有的订单流、内部研究或风险头寸。要在云服务商上实现这一点,就意味着要把存储、计算、编排和治理等彼此独立的服务拼凑在一起。终端扩展插件无法跨你的内部系统进行推理。

资本市场 AI 的竞争优势不在于模型,而在于专有数据层。这正是 Databricks Data + AI Platform 的立足之地。

这在生产环境中意味着什么

A-Team 奖项表彰的是一个平台,但真正的证明在于资本市场公司基于它构建了什么。在 2026 年 Data + AI Summit 上,多家领先机构分享了他们如何利用 Databricks 改造研究与风险领域的各项工作流程。

加速研究与投资组合分析的 AI 智能体

构建这一专有数据基础,正是 RBC Capital Markets 着手推进的目标。该公司基于 Databricks 构建了 Aiden QuickTakes——一个专有 AI 智能体,在加快研究工作流程的同时,满足受监管资本市场所要求的治理与安全标准。通过将数据摄取、AI 驱动的分析和内容生成统一在一个受治理的环境中,Aiden QuickTakes 生成研究报告草稿的速度最高可提升 95%,让分析师得以专注于客户最看重的差异化洞察。

“资本市场上的每家公司都能用到大语言模型。区别在于,你能否在具备适当管控措施的前提下,让它们对接你自己的专有数据。Databricks 让我们一次性构建起这一基础,并在不牺牲安全或治理的前提下,把 AI 扩展到各类研究工作流中。”——Michael Tran,RBC Capital Markets 董事总经理、AI 与数字产品负责人

同样的模式也延伸到买方:各机构正将市场数据、基本面数据、ESG 数据和另类数据汇聚到同一个受治理的湖仓一体(lakehouse)中,再在其上叠加 AI 智能体,以加速信号研究、因子回测和组合构建。

研究与筛选:从数小时缩短到数分钟

LSEG 与 Databricks 合作,将可直接用于 AI 的金融数据原生接入分析、AI 应用和智能体。在 2026 Data + AI Summit 上,LSEG 演示了 AI 智能体如何用自然语言查询 35 万+ 项投资标的,把原本需要数小时的人工基金筛选缩短到数分钟。这一合作让机构级金融数据可直接在 Databricks Data + AI Platform 内获取,消除了拖慢大多数机构效率的集成摩擦。

风险、安全与合规

资本市场机构也在借助 Databricks 实现交易后运营的现代化。一些机构在 Databricks 上构建了统一的网络安全分析平台,将结构化、半结构化和非结构化安全数据集中起来,以支持更快的威胁检测、取证和合规报告。另一些机构则将核心银行系统迁移到以 OpenSharing 和 Unity Catalog 构建的受治理湖仓一体架构上,以满足监管要求,用受治理的自动化流水线取代脆弱的电子表格和人工对账。

贯穿全生命周期:统一的受治理前端

交易前研究、交易中筛查和交易后报告都汇聚到同一套受治理的定义上,但需要这些答案的人很少在数据平台内部工作。Databricks Genie One 是面向业务用户的数据智能 AI 协作者,充当前端。投资组合经理、分析师和风险官用自然语言提问,就能得到基于 Unity Catalog 的答案,无需使用 SQL 或 notebook。

通过 Model Context Protocol,同一受治理层可扩展到 Microsoft Teams、Excel 以及 Claude 和 ChatGPT 等助手。每个请求都会依据 Unity Catalog 权限进行授权并记录日志,生成的 SQL 和血缘关系可检查。交互界面会变。数字不会变。

Databricks Genie One 实用指南

赢得该奖项的治理技术栈

让 Databricks 在资本市场中脱颖而出的,不仅仅是速度或规模。而是治理架构,它使 AI 足够可信,能够适用于受监管环境:

Unity Catalog 为数据和 AI 资产提供统一的控制平面,用于访问、血缘和审计——因此每次查询、每个模型输出、每个仪表板都可追溯。

Unity Catalog Semantics 和 Glossary 中的业务语义让公司可以一次定义关键指标——VaR、预期短缺、流动性缓冲、敞口指标——并在报告、仪表板和 AI 应用中一致地复用。

MLflow 增加了结构化的模型治理:版本管理、实验跟踪、评估,以及受控部署到生产环境。

全面的日志记录会捕获谁在哪些数据上运行了什么查询或 AI 交互,从而为模型风险管理、内部审计和监管审查提供支撑。

这正是「演示效果出色的 AI」与「能够经受董事会和监管审视的 AI」之间的区别。

一种模式,而非个例

这一模式十分清晰:整个行业正在凝聚于同一个洞察——模型本身不是战略,平台才是。模型会不断更替。真正占据优势的企业能够在不造成业务中断的前提下采用最新最强的模型,因为它们专有的数据基础已经就位、已经完成治理,并且已经在为它们创造价值。

如果你是一位正在评估自身数据与 AI 战略的资本市场领域负责人,可以从以下几点入手:

阅读《A-Team AI in Capital Markets Awards 2026 Winners Report》,了解完整获奖名单以及行业社区所认可的成果。

关注 Databricks for Financial Services LinkedIn,获取面向金融行业的内容。

联系我们的资本市场团队,探讨像贵司这样的企业当下如何构建受治理的 AI 基础。

常见问题

原文较长,译文为节选;译文由英文原文机器翻译,术语按行业惯例统一,可在文末展开原文核对。

查看英文原文(Databricks Wins A-Team Award for AI in Capital Markets)

Leading institutions are already building on the platform. RBC Capital Markets generates draft research reports up to 95% faster with its Aiden QuickTakes AI agent, and LSEG showed AI agents screening 350,000+ investments in plain English.

Recognized for AI and governance, together.

Capital markets have no shortage of AI claims. Every vendor pitch deck promises intelligence, automation, and alpha. But the A-Team Group, whose awards are voted on by practitioners across the capital markets industry, chose to recognize something different: the combination of AI and governance on a single platform.

That distinction matters.

"Congratulations to Databricks for winning Best AI Solution for Data Intelligence & Governance at the 2nd annual AI in Capital Markets Awards 2026. Selected by our industry community, this recognition highlights true excellence and innovation in driving AI capabilities across capital markets. Databricks should be immensely proud of this achievement."- Angela Wilbraham, CEO of A-Team Group

This is Databricks' first A-Team AI in Capital Markets award. It arrives at a moment when the industry is drawing a hard line between experimental AI and production-grade AI capable of withstanding regulatory scrutiny.

Why governance is the real AI unlock in capital markets

The pressure on capital markets firms is acute. Intraday volatility remains elevated. T+1 settlement in North America is compressing operational windows. Regulators are sharpening expectations around AI governance, model transparency, and data lineage, from ESMA's supervisory briefing on algorithmic trading to the FSB's 2026 work program on AI sound practices.

Meanwhile, 94% of portfolio managers and analysts already use AI or machine learning in their alternative data research process, according to a 2026 survey of managers overseeing roughly $2 trillion in assets. But 71% say combining data from different sources is their most frustrating challenge.

The bottleneck isn't access to AI. It's the data foundation underneath it.

Large language models are powerful, but they can't see your firm's proprietary order flow, internal research, or risk positions. Getting there on cloud providers means stitching together separate services across storage, compute, orchestration, and governance. Terminal extensions can't reason across your internal systems.

The competitive edge in capital markets AI is not in the model. It's in the proprietary data layer. That's precisely where the Databricks Data + AI Platform lives.


What this looks like in production

The A-Team award recognizes a platform, but the proof is in what capital markets firms are building on it. At the 2026 Data + AI Summit, leading institutions shared how they're using Databricks to transform workflows across research and risk.

AI agents that accelerate research and portfolio analysis

Building that proprietary data foundation is exactly what RBC Capital Markets set out to do. The firm built Aiden QuickTakes, a proprietary AI agent on Databricks that accelerates research workflows while maintaining the governance and security standards required in regulated capital markets. By unifying data ingestion, AI-driven analysis, and content generation in a single governed environment, Aiden QuickTakes can auto-generate draft research reports up to 95% faster, freeing analysts to focus on the differentiated insights their clients value most.

"Every firm in capital markets has access to large language models. The difference is whether you can point them at your own proprietary data with the right controls in place. Databricks let us build that foundation once and scale AI across research workflows without compromising on security or governance."- Michael Tran, Managing Director, Head of AI & Digital Product, RBC Capital Markets

The same pattern extends to the buy side: firms are bringing market, fundamental, ESG, and alternative data together into a single governed lakehouse, then adding AI agents on top to speed up signal research, factor backtesting, and portfolio construction.

Research and screening: hours to minutes

LSEG partnered with Databricks to bring AI-ready financial data natively into analytics, AI apps, and agents. At the 2026 Data + AI Summit, LSEG demonstrated how AI agents can query 350,000+ investments using plain English, turning hours of manual fund screening into minutes. The partnership makes institutional-grade financial data available directly within the Databricks Data + AI Platform, eliminating the integration friction that slows most firms down.

Risk, security and compliance

Capital markets firms are also using Databricks to modernize post-trade operations. Institutions have built unified cybersecurity analytics platforms on Databricks, centralizing structured, semi-structured, and unstructured security data to power faster threat detection, forensics, and compliance reporting. Others have migrated core banking systems to a governed lakehouse architecture with OpenSharing and Unity Catalog to meet regulatory requirements, replacing fragile spreadsheets and manual reconciliations with governed, automated pipelines.

Across the lifecycle: One governed front end

Pre-trade research, at-trade screening, and post-trade reporting all converge on the same governed definitions, yet the people who need those answers rarely work inside a data platform. Databricks Genie One, the data-smart AI coworker for business users, is the front end. Portfolio managers, analysts, and risk officers ask in plain language and get answers grounded in Unity Catalog, with no SQL or notebooks involved.

Through the Model Context Protocol, the same governed layer extends to Microsoft Teams, Excel, and assistants such as Claude and ChatGPT. Every request is authorized against Unity Catalog permissions and logged, with the generated SQL and lineage inspectable. The surface changes. The number does not.

A Practical Guide to Databricks Genie One

The governance stack that earned the award

What sets Databricks apart in capital markets isn't just speed or scale. It's the governance architecture that makes AI trustworthy enough for regulated environments:

Unity Catalog provides a single control plane for access, lineage, and auditing across data and AI assets — so every query, every model output, every dashboard is traceable.

Business semantics in Unity Catalog Semantics and Glossary let firms define key measures — VaR, expected shortfall, liquidity buffers, exposure metrics — once and reuse them consistently across reports, dashboards, and AI applications.

MLflow adds structured model governance: versioning, experiment tracking, evaluation, and controlled deployment into production.

Comprehensive logging captures who ran what queries or AI interactions on which data, supporting model risk management, internal audit, and supervisory reviews.

This is the difference between AI that demos well and AI that can withstand board-level and regulatory scrutiny.

A pattern, not a one-off

The pattern is clear: the industry is converging on a single insight. The model is not the strategy. The platform is. Models will keep changing. The firms positioned to win can adopt the latest and greatest without disruption, because their proprietary data foundation is already in place, already governed, and already working for them.

If you're a capital markets leader evaluating your data and AI strategy, here's where to start:

Read the A-Team AI in Capital Markets Awards 2026 Winners Report to see the full list of winners and what the industry community recognized.

Explore the Databricks for Financial Services LinkedIn for industry-specific content.

Connect with our capital markets team to discuss how firms like yours are building governed AI foundations today.

Frequently asked questions


Who won the A-Team AI in Capital Markets Awards 2026?
Databricks won Best AI Solution for Data Intelligence & Governance at the second annual A-Team AI in Capital Markets Awards 2026, selected by the capital markets practitioner community.

What is the Databricks Data + AI Platform for capital markets?
The Databricks Data + AI Platform combines a unified lakehouse, Unity Catalog governance, and built-in AI tooling — including Agent Bricks and Genie One — so capital markets firms can scale AI across research, trading, risk, and compliance on a single governed platform.

Why is data governance critical for AI in capital markets? 
Regulators, including ESMA and the FSB, are sharpening expectations regarding AI governance, model transparency, and data lineage. Without a governed data foundation, firms cannot move AI from experimentation to production in a way that withstands regulatory scrutiny.

What is the Model Context Protocol and how does it apply to financial services?
The Model Context Protocol (MCP) is an open standard for connecting AI assistants to data and tools. Databricks uses MCP to extend its governed data layer, including Unity Catalog permissions, definitions and lineage, to Microsoft Teams, Excel and AI assistants such as Claude and ChatGPT. Every request is authorized and logged, so firms get conversational AI without creating an ungoverned path to their data.

Get the latest posts in your inbox

原文为英文,正文由自动抓取并保留原文。

本文由采集器自动同步

原文发布于 Databricks Blog(Databricks Wins A-Team Award for AI in Capital Markets),标题、摘要与正文由机器翻译,术语按行业惯例统一,未经人工逐句校订。

阅读原文 ↗