AI Summary
→ WHAT IT COVERS Gabe Stengel, founder of Rogo, describes building an AI platform for capital markets that automates deal-making workflows across private equity, investment banking, and M&A. He covers the product's evolution through model generations, why private markets outpace public equities as an AI target, and how financial firms must restructure around AI over the next two to five years. → KEY INSIGHTS - **Vertical AI moat strategy:** The strongest applied AI businesses target industries with complex, unbuilt plumbing—proprietary data formats, regulatory compliance layers, and fragmented systems of record. Private markets fit this profile because coordination, standardization, and transaction execution remain entirely human-driven. Public equities, by contrast, are already heavily automated, leaving less infrastructure to own. Founders should prioritize verticals where foundational wiring doesn't yet exist. - **Harness over model intelligence:** Claude's run-up against ChatGPT came not from superior raw model capability but from a superior harness—the scaffolding around the model that enables long-running, complex tasks. Rogo rebuilds its core agentic harness continuously, treating it like Affirm rebuilds its ledger annually: to prevent ossification, retain engineering talent, and stay ahead of rapid model capability shifts every six months. - **Pricing evolution toward outcomes:** Enterprise AI businesses should plan two pricing transitions—from seat-based to usage-based, then from usage-based to outcome-based. Rogo's target is charging per investment idea generated or per deal simulation produced, because clients can directly measure that value. Skipping the intermediate token-pricing stage avoids misalignment where heavy usage costs the vendor money without clear customer ROI. - **Compaction as the unsolved frontier:** Current AI agents handle one-to-one conversations but fail at persistent memory across hundreds of interactions or across entire organizations. The compaction problem—condensing accumulated context into a usable token budget while maintaining coherent understanding of users—scales exponentially in multi-agent, multi-user environments. Solving this is what separates a useful tool from a true autonomous agent operating inside a firm. - **Institutional AI adoption bottleneck:** Individual bankers at Rogo-deployed firms report 100x efficiency gains—MDs producing five-page client materials in ten minutes without analyst support. However, firms struggle to convert individual productivity into measurable firm-level outcomes. The real strategic question becomes whether to use AI for cost reduction or market expansion, such as JPMorgan targeting SMB M&A deals previously too small to staff profitably. - **Talent black hole as competitive strategy:** The execution bar for vertical AI companies now exceeds prior software eras because labs and well-funded competitors recruit aggressively. Rogo acquired six fledgling financial AI startups to absorb former founders with domain expertise and product intuition. Building a reputation that attracts former founders—people who have already eaten glass, built product instincts, and survived failure—creates compounding talent density that generic engineering hiring cannot replicate. → NOTABLE MOMENT After 40 consecutive rejections from top-tier VCs including Sequoia, Benchmark, and Kleiner—many of whom advanced to dinners and IC meetings before passing—Stengel describes the experience as being broken up with by 40 people he had fallen for. The eventual lead investor, Keith Rabois, framed the contrarian bet by noting every peer had already said no. 💼 SPONSORS [{"name": "Ramp", "url": "https://ramp.com/invest"}, {"name": "WorkOS", "url": "https://workos.com"}, {"name": "Vanta", "url": "https://vanta.com/invest"}, {"name": "Ridgeline", "url": "https://ridgeline.ai"}] 🏷️ Vertical AI, Private Markets, Capital Markets Automation, Enterprise AI Sales, Agentic Systems, AI Product Strategy