供资
Finpilots Secures $10M 系列 A为机构投资者构建AI
在 Google 上的首选来源中添加 Unite.AI
Finpilot联合创始人兼CEO拉克沙伊·乔汉宣布 2026年11月4日的回合：由游说资本带领的1000万美元A系列，由马德罗纳参与，为机构资本建设AI-native操作系统。
公司开发软件，供各小组监督大学、公共养恤金、基金会、医院系统和家庭的大型长期资金池。 财务驾驶员将西雅图儿童、达拉斯儿童健康、哈肯萨克Meridian健康、TIFF、蓝十字蓝盾和怀俄明退休系统命名为使用其软件的机构。 他们的投资工作有助于资助包括医学研究、保健和人们退休在内的活动。
投资办公室的记录
业务目标是使投资信息能够使用的经常性工作。 投资数据分散在不连接的旧系统，而许多报告是通过电子邮件作为PDF提供的。 工作人员必须阅读这些文件，提取数字，对照其他信息加以核对，并每月和每季度在正确的地点输入。 由此产生的数字为董事会和委员会的决定提供了依据，使准确性和明确的来源至关重要。
Finpilot将其平台组织为三层。 第一个机构保存着一个机构持有的股票和文件的现有记录。 第二幅地图绘制了记录之间的关系，包括每个数字的含义和来源。 第三个是agents，它读取收到的报告，更新数字，编写分析和报告，并跟踪变化。 当需要判决时，agents将问题交给一个人，每个数字仍然可以追踪到其来源。
用户可以描述他们想要的视图，让系统在实时数据上构建这些视图，然后保持这些视图的时态。 了解一个团队喜欢怎样做的工作就保存在系统里给更广泛的组织，而不是留在个人的聊天史上。
Agents 在定义的审查点停止
agent工作流程是配置的进程，而不是一次性的提示。 Agents监测核定来源，并根据任务规定、以前的指出和其他核定材料开展工作。 在个人必须决定如何行事时，指定的审查点将停止这一过程。 在季度监测的例子中，agent收到管理信件，起草更新，并在更新推进前升级一个实质性的按键人物变化。
支助的任务包括管理人员和组合监测、研究简报、投资委员会和董事会材料、客户报告、会议筹备和摘要，以及核准更新客户关系管理系统和其他记录。 会议工作流程可以将决定、所有者和未决问题与一般性讨论分开，然后提出后续建议，并记录只有在得到批准后才能取得进展的变动。
基本的连接数据层将现有的客户关系管理、研究、组合、文件和通信系统连接起来，同时为每个系统所拥有的记录留下权威。 实体决议将不同提及同一投资、组织、个人或账户的内容联系起来，并保留其来源。 不确定的匹配和相互矛盾的值仍然可见以供审查，而数值则保留其文档、页面和系统来源。 权限决定每个用户可以访问哪些记录。
安全控制和开放作用
Finpilot的安全文件报告SOC 2型的保证和在中途和休息时的加密。 每个机构的记录、文件和检索指数仍处在一个单独的界限内，每个请求都得到认证、作用和许可。 客户数据用于提供订约服务，不用于公司核准的政策和适用的供应商合同下的示范培训。 缺少、相互矛盾或因此产生的信息成为审查项目，由合格的工作人员负责判决。
该公司也在西雅图招募人员。 其职业生涯页面列出以西雅图为基地的软件工程师，产品经理，营销领头人，账户经理，业务开发代表，客户成功AI解决方案专家，以及业务业务协理。
资金支持一个将记录、分析和经常工作流程连接在一个系统中的平台。 其中心挑战是减少人工编制，而不混淆资料来源或削弱问责制。 产品的设计明确了取舍：agents准备和推进工作，而人们则保留对相应决定的批准。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Funding
Finpilot Secures $10M Series A to Build AI for Institutional Investors
Add Unite.AI to your preferred sources on Google
Finpilot co-founder and CEO Lakshay Chauhan announced the round on November 4, 2026: a $10 million Series A led by Lobby Capital, with participation from Madrona, to build an AI-native operating system for institutional capital.
The company develops software for teams overseeing large, long-term pools of money at universities, public pensions, foundations, hospital systems and families. Finpilot names Seattle Children’s, Children’s Health of Dallas, Hackensack Meridian Health, TIFF, Blue Cross Blue Shield and the Wyoming Retirement System among the institutions using its software. Their investment work helps finance activities including medical research, health care and people’s retirements.
A Record Built for Investment Offices
The operational target is the recurring work required to make investment information usable. Investment data is spread across older systems that do not connect, while many reports arrive as PDFs by email. Staff must read those documents, extract figures, check them against other information and enter them in the right place each month and quarter. The resulting numbers feed board and committee decisions, making accuracy and clear sourcing essential.
Finpilot organizes its platform into three layers. The first maintains a current record of an institution’s holdings and documents. The second maps relationships between records, including what each number means and where it originated. The third consists of agents that read incoming reports, update figures, prepare analysis and reports, and track changes. When judgment is needed, the agents refer the issue to a person, and each number remains traceable to its source.
Users can describe the views they want and have the system build them on live data, then keep those views current. Knowledge about how a team prefers work done is saved in the system for the wider organization, rather than remaining in an individual’s chat history.
Agents Stop at Defined Review Points
The agent workflows are configured processes rather than one-off prompts. Agents monitor approved sources and prepare work against mandates, previous notes and other approved material. Defined review points stop the process when a person must decide how to proceed. In a quarterly-monitoring example, an agent receives a manager letter, drafts an update and escalates a material key-person change before the update advances.
Supported tasks span manager and portfolio monitoring, research briefs, investment-committee and board materials, client reporting, meeting preparation and summaries, and approved updates to customer relationship management (CRM) systems and other records. The meeting workflow can separate decisions, owners and unresolved questions from general discussion, then propose follow-ups and record changes that move forward only after approval.
The underlying Connected Data layer links existing CRM, research, portfolio, document and communications systems while leaving each system authoritative for the records it owns. Entity resolution connects different references to the same investment, organization, person or account and preserves their origins. Uncertain matches and conflicting values stay visible for review, while values retain their document, page and system provenance. Permissions determine which records each user can access.
Security Controls and Open Roles
Finpilot’s security documentation reports SOC 2 Type II assurance and encryption in transit and at rest. Each institution’s records, documents and retrieval index remain within a separate boundary, with authentication, roles and permissions checked on every request. Customer data is used to provide the contracted service and is not used for model training under the company’s approved policy and applicable provider contracts. Missing, conflicting or consequential information becomes a review item, keeping qualified staff responsible for judgment.
The company is also recruiting in Seattle. Its careers page lists Seattle-based openings for a software engineer, product manager, marketing lead, account executive, business development representative, AI solutions specialist in client success, and business operations associate.
The funding supports a platform that connects records, analysis and recurring workflows in one system. Its central challenge is to reduce manual preparation without obscuring sources or weakening accountability. The product’s design makes that trade-off explicit: agents prepare and advance the work, while people retain approval over consequential decisions.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.