Why OpenAI and Anthropic Won't Win Finance

本次讨论探讨了人工智能对金融业的深远影响，强调了人工智能模型在某些分析任务中如何迅速超越人类能力。嘉宾Gabe，Rogo的创始人，提出在十年内，全球领先的投资公司和银行将有90%的企业价值来源于软件、数据和系统，而不再仅仅依靠人力资本。这要求战略性地转向将顶尖人才的“潜在智慧”嵌入到自主系统中。 Gabe回顾了Rogo的发展历程，指出在GB3（ChatGPT问世之前）出现之前，曾两次尝试构建人工智能驱动的金融工具但均告失败。Rogo的早期版本在概念上“神奇”，但在功能上受限。真正的突破来自O1 Pro等模型，它们为计算财务指标等任务提供了足够的可靠性；随后是Opus 4.5，它实现了与初级投资专业人士相媲美的能力。他强调，早期的应用型人工智能公司面临“先发劣势”，但如果它们能正确预测模型进展并以此为目标进行建设，就能取得成功。Rogo今天的成功归因于其对特定用户工作流程、合规需求以及用户体验（UX）细节的深入理解，例如允许董事总经理将批注通过电子邮件发送给人工智能分析师，并在20分钟内得到回复。 与最初认为由于数据可用性，公开股票将是主要应用领域的直觉相反，Rogo的核心市场目前是“交易撮合者”——参与买卖和协调交易的专业人士。这一重点利用了私募市场中现有“基础设施”的缺乏，在私募市场中，许多流程仍然由人工驱动且非标准化。Rogo与后端系统（CRM、投资组合监控）集成，以自动化不仅仅是智能的任务，旨在成为私募市场的基础设施。 Gabe概述了Rogo通过构建“垂直”解决方案来与Anthropic和OpenAI等大型人工智能实验室竞争的战略。当实验室专注于核心模型智能时，Rogo则深入研究金融领域特有的具体、复杂且监管严格的问题，例如可审计性、处理重大非公开信息（MNPI）以及构建安全数据室。他认为，人工智能模型周围的“配套体系”和基础设施与模型本身一样重要，它们能实现持久性、上下文保留以及与现有系统的集成。 Rogo作为一家传统的企业软件公司运营，按席位定价，这需要一套强大的市场推广战略。Gabe预计未来将转向基于结果的定价，即对有价值的洞察或成功执行的交易收费，而不是按代币消耗量收费。他设想了一个未来，资本市场将显著提高效率、流动性和透明度，通过让创业者更容易获得资金来加速创新。这一转型将要求Rogo从“副驾驶聊天机器人”发展为完全“自动驾驶工具”，并最终成为跨公司“代理进行交易的交易所”。 对于金融专业人士，Gabe建议未来的价值将在于收集喂养人工智能模型的独特数据和洞察，而不是常规分析任务。他强调，传统公司面临“创新者困境”，需要拥抱“AI原生”思维——不断重新评估和革新其业务的每个部分。在Rogo内部，公司使用一个广泛的“公司大脑”（昵称“史莱克”），它记录所有内部对话，并主动为员工提供上下文和洞察，从而推动赋能和效率。 Gabe反思了创办初创公司所带来的情感代价，将其描述为“吃玻璃”——应对拒绝、产品转型和人才挑战。他强调在拓展市场时需要“进取心”，尤其考虑到人工智能在企业中快速的采纳周期。尽管对潜在挫折存在内在的偏执，但他坚信Rogo有能力成为变革资本市场的重要力量，并将其与金融在经济发展中的历史作用相提并论。他还强调了“黑洞”策略的至关重要性——即吸引顶尖人才和资本，以在快速变化的格局中保持势头和竞争优势。

The discussion explores the profound impact of artificial intelligence on the financial industry, highlighting how AI models are rapidly surpassing human capabilities in certain analytical tasks. The guest, Gabe, founder of Rogo, posits that within ten years, the world's leading investment firms and banks will derive 90% of their enterprise value from software, data, and systems, rather than solely from human capital. This necessitates a strategic shift towards embedding the "latent minds" of top talent into autonomous systems. Gabe traces Rogo's journey, noting two previous unsuccessful attempts to build an AI-powered financial tool before the advent of GB3 (pre-ChatGPT). Early iterations of Rogo were "magical" in concept but functionally limited. The true breakthrough came with models like O1 Pro, which offered sufficient reliability for tasks like calculating financial metrics, and later Opus 4.5, which enabled capabilities comparable to junior investment professionals. He emphasizes that early applied AI companies faced a "first movers disadvantage" but succeeded if they correctly anticipated model advancements and built towards that future state. Rogo's success today is attributed to its deep understanding of specific user workflows, compliance needs, and attention to detail in UX, such as allowing managing directors to email markups to an AI analyst for a 20-minute turnaround. Contrary to the initial intuition that public equities would be the primary application due to data availability, Rogo's core market is currently "deal makers" – professionals involved in buying, selling, and coordinating transactions. This focus leverages the existing lack of "plumbing" in private markets, where many processes are still human-driven and non-standardized. Rogo integrates with backend systems (CRM, portfolio monitoring) to automate tasks beyond just intelligence, aiming to become the infrastructure for private markets. Gabe outlines Rogo's strategy to compete with large AI labs like Anthropic and OpenAI by building "perpendicular" solutions. While labs focus on core model intelligence, Rogo delves into the specific, messy, and regulatory-heavy problems unique to finance, such as auditability, handling material non-public information (MNPI), and building secure data rooms. He argues that the "harness" and infrastructure around the AI models are as crucial as the models themselves, enabling persistence, context retention, and integration into existing systems. Rogo operates as a traditional enterprise software business, pricing per seat, which requires a robust go-to-market strategy. Gabe anticipates a future shift towards outcome-based pricing, charging for valuable insights or successfully executed deals rather than token consumption. He envisions a future where capital markets become dramatically more efficient, liquid, and transparent, accelerating innovation by making it easier for entrepreneurs to access capital. This transformation would require Rogo to evolve from a "copilot chatbot" to a full "autopilot tool" and eventually to an "exchange for agents to transact" across firms. For finance professionals, Gabe suggests that future value will lie in gathering unique data and insights that feed AI models, rather than routine analytical tasks. He stresses that traditional firms face an "innovator's dilemma," needing to embrace an "AI-native" mindset – constantly re-evaluating and revolutionizing every part of their business. Internally, Rogo uses an extensive "company brain" (nicknamed Shrek) that records all internal conversations and proactively assists employees with context and insights, driving enablement and efficiency. Gabe reflects on the emotional toll of building a startup, describing it as "eating glass" – navigating rejection, product pivots, and talent challenges. He emphasizes the need for "aggression" in attacking the market, especially given the rapid adoption cycle of AI in enterprises. Despite inherent paranoia about potential setbacks, he maintains conviction in Rogo's trajectory to become a significant force in transforming capital markets, drawing parallels to the historical role of finance in economic development. He also highlights the critical importance of a "black hole" strategy – attracting top talent and capital to maintain momentum and competitive advantage in a rapidly evolving landscape.