Gabe Stengel - Building Investing Superintelligence - [Invest Like the Best, EP.492]

Gabe Stengel - Building Investing Superintelligence - [Invest Like the Best, EP.492]

At a Glance

Gabe Stengel is the co-founder and CEO of Rogo, an AI platform built for the financial industry. The core thesis of this episode is: Over the next 2-5 years, the best investors will reshape their firms and themselves around AI, and Rogo is building the infrastructure to make this happen. The most impactful insight comes from Gabe: "Auditability is more important than accuracy"—when AI outputs are traceable, even imperfect results can still be used; but accurate outputs that cannot be traced are worthless because they cannot be trusted.

---

Theme 1: Model Capability Leap and Rogo’s Product Evolution

Gabe Stengel believes that Rogo’s product capabilities are tightly linked to the iteration of frontier models, undergoing a qualitative shift from “completely unusable” to “changing the way work is done.”

• Historical Context: Gabe attempted to launch Rogo twice during high school and college, both times ending in failure. The real start came after the release of GPT-3 (before ChatGPT). Early products “looked magical in demos, but nothing worked well.”
• Key Milestones:
• o1 Pro Era: The model first achieved sufficient reliability, becoming “at least a good search tool”—capable of reliably calculating “financial metrics over the past 12 quarters,” making users feel that “doing it manually is more annoying.”
• Opus 4.5 Era (late 2025 to early 2026): Model capabilities reached a point where it could “basically do anything a junior investment professional or junior banker can do,” provided the right instructions and context are given.
• Gabe’s Reflection: “Application-layer AI companies face a first-mover disadvantage—you think you know where the world is going, but the model hasn’t gotten there yet. Users try it and say, ‘This is garbage.’ But if your judgment about the endgame is correct, when the model arrives, everything becomes magical.”

---

Theme 2: The Last Mile and the "Harness" Are the True Moat

Gabe Stengel emphasizes that in the financial AI space, the "harness" around the model and the engineering details of the "last mile" matter more than the intelligence of the model itself.

• Mechanism Breakdown: Gabe uses Claude Code vs. ChatGPT as an example—"The models themselves are essentially similar, but Claude's harness and presentation are far superior, allowing the model to leverage more long-range capabilities." He argues that humans are able to act with high agency not just because of IQ and knowledge, but also because the brain has various "microservices"—how to store knowledge, how to retrieve it, and how to trigger it. All of these need to be built out.
• Specific Details: Rogo has designed a workflow that allows Managing Directors (MDs) to send deck annotations via email—the MD simply annotates on an iPad as usual, sends it to Rogo's AI analyst, and receives results in 20 minutes (versus the traditional 2 days), while automatically notifying junior analysts and displaying a full audit trail.
• Compliance Barriers: When handling MNPI (Material Non-Public Information), pixel-level audit trails are required to ensure that "if one day a Delaware court judge demands that AI inputs and research be forensically admissible, you have done it the right way, with no information commingled."
• Gabe's Judgment: "For Anthropic, doing these things is like picking up pennies on the side of the road—they are on their way from $100 billion to $1 trillion in revenue, and the financial system has far too much depth to build beyond just intelligence itself."

---

Theme 3: Pricing Model Revolution — From Per-Seat to Per-Outcome

Gabe Stengel believes that AI software companies need to undergo two pricing revolutions: first shifting to usage-based pricing, then to outcome-based pricing. Rogo is attempting to skip the intermediate step.

• Current Model: Rogo currently charges per seat, as clients (large banks) are accustomed to this model, drawing comparisons to Bloomberg, FactSet, Capital IQ, and PitchBook.
• Future Vision: Gabe ultimately hopes to implement pay-per-outcome — "What if I charge per good investment idea? Or per perfect LP quarterly report?" He argues that clients can clearly perceive the value of a good idea (since they know how much money it made), while charging per token would lead to misaligned perceptions of value between both parties.
• Core Contradiction: If Rogo performs too well and client usage surges, Rogo's model costs also surge, turning clients into "worse customers." Gabe's response is: "We're only 1% of the way through the product roadmap, with 99% of innovation ahead. The key is to be a good partner so they want to work with us in the future."

---

Theme 4: The Future of Capital Markets — From Human Intermediation to AI-Native Trading

Gabe Stengel predicts that within 10–20 years, corporate financing, asset pricing, and trade execution will undergo fundamental transformation, similar to how Rocket Mortgage disrupted the mortgage market.

• Historical analogy: 15–20 years ago, every mortgage required a face-to-face meeting with a bank loan officer, and “no one thought you would ever want to remove the human element.” Today, 40–50% of mortgages are completed through online platforms like Rocket Mortgage.
• Extrapolation:
• KKR evaluates whether it can sell a portfolio company to another sponsor: reduced from 5 months to 5 minutes.
• Asset pricing time will decrease by an order of magnitude.
• Markets will become more transparent, more liquid, and more efficient, with a significant increase in activity.
• Required infrastructure: Gabe draws an analogy to Bloomberg’s strategy — “first give a little data to get in the door, build all the analytics and workflows, then provide a trading and communications platform.” Rogo’s path is: first use AI to get in the door, build a complete workflow (from copilot to autopilot), and then construct a communication channel between agents, allowing AI agents to negotiate transactions across institutions.
• Uncertainty: Gabe acknowledges that “the actual speed of transformation in private markets” is one of the biggest unknowns. The degree of standardization, regulatory forces, and whether “small business owners will be willing to click a button to sell their company instead of shaking hands” remain open questions.

---

Theme 5: The Current State of Enterprise AI Adoption and the Innovator's Dilemma

Gabe Stengel argues that the financial services industry is facing its first major "innovator's dilemma" in decades, and most companies remain stuck at the "personal productivity enhancement" stage, failing to translate it into "enterprise-level productivity."

• Current State: Most companies see significant personal productivity gains, but these have yet to translate into measurable enterprise productivity. Gabe describes: "Ask any analyst at any bank we've deployed in, and they'll say, 'Life is great, I'm 100 times more productive than before.' But the question is—where does this show up? Are you winning more deals?"
• Innovator's Dilemma: Over the past 10-20 years, investment firms and banks have had it easy. Private market funds enjoy natural inertia (raising one fund after another). Gabe assesses: "Now, for the first time, there's a shock that makes every investment firm and bank say, 'Wow, I need to completely rethink what I'm doing.' There will be hundreds of AI-native disruptors attacking my business model."
• Definition of AI-Native: "A willingness to constantly and completely reinvent everything you do, to be so 'addicted to AI' that you don't worry about what's possible or what seems far-fetched, but simply move toward integrating this alien foundational technology into everything. No part of the business is sacred."

---

Mentioned Positions

Position	Guest's Stance	Key Data
Rogo (itself)	Bullish, believes it is at the 1% stage of the product roadmap	Has acquired 6 financial AI startups; team of over 100 from top investment banks/institutions; rejected by 40 investors before Series A
Anthropic	Viewed as a model provider, not a direct competitor	Pricing model based on token consumption; Claude Code sees usage growth due to better "guardrails"
OpenAI	Same as above	ChatGPT is a personal 1-on-1 usage model
Bloomberg	Used as a business model analogy	Strategy: data entry → analysis workflow → trading/communication platform (Bloomberg Messenger)
Rocket Mortgage	Used as an industry transformation analogy	40-50% of mortgages completed via online platforms
JP Morgan	Used as a client innovation case	Announced it will attempt to do more M&A work for SMBs, as AI makes "one banker equals one deal team" possible
Harvey (founded by Winston)	Used as a peer reference	Gabe praised Winston's ability to avoid being distracted by "a hundred flesh wounds"
Molas (co-founded by John Montazzi)	Viewed as a cutting-edge industry practitioner	Wants digital clones of all the best bankers, allowing junior staff to leverage their expertise and relationships

---

Judgments Worth Remembering

1. “Auditability is more important than accuracy” (Gabe Stengel) — If output is traceable, it can still be used and debugged even if inaccurate; if untraceable, it is untrustworthy regardless of accuracy. This is especially critical in the heavily regulated financial sector.

2. “The model is already smarter than anyone I know. The problem is the pipeline.” (Gabe Stengel) — Connecting context, informing your investment thesis, and integrating into your workflow — these “pipeline” engineering tasks are the current frontier.

3. “Every PM can have 10,000 agents, let them talk to each other, read notes, debate for 24 hours, and then give you an idea.” (Gabe Stengel) — Because investors are willing to pay $50,000 for a good idea, this token consumption model is viable in finance but rarely seen in other fields.

4. “If every investment institution will make an AI purchasing decision in the next 18 months and will buy something anyway — then the only thing that matters is capturing the market as fast as possible.” (Pat Grady’s feedback to Gabe) — Based on this, Gabe adjusted his plan to be more aggressive, accepting a “30% increase in failure probability, but also a 20% increase in the probability of becoming a hundred-billion-dollar company.”

5. “In 10 years, 90% of the enterprise value of the world’s best investment institutions and banks will not reside in people, but in software, data, and systems.” (Gabe Stengel) — He advises every institution to start thinking now: how to extract the tacit knowledge from the minds of the best talent into systems owned and operated by the institution.

6. “Rogo has an internal company brain called Shrek — all internal conversations are recorded and filtered into it. It is both active and passive: you can query it for information, and it can proactively remind you, ‘You’re meeting this private credit firm on Thursday; here’s what you should know.’” (Gabe Stengel) — Gabe reviews the company-wide AI tool usage rankings monthly, and the last-place person gets a “dunce cap” posted in the office.

7. “Private markets have always been immune to standardization because of the vast amount of unstructured data. AI should be able to solve this problem.” (Gabe Stengel) — But the pace of transformation depends on regulatory and market forces, as well as “whether small business owners are willing to click a button to sell their company instead of shaking hands.”

8. “Applied AI companies will become ‘black holes’ for talent, capital, and brands.” (Gabe Stengel) — AI is a massive tailwind, but the execution bar is higher than ever. If you cannot become such a black hole quickly enough, you risk being crushed by labs or larger companies.