ExpertsFMGabe Stengel - Building Investing Superintelligence - [Invest Like the Best, EP.492]
Invest Like the Best with Patrick O'Shaughnessy · September 22, 2026
5 Selected Lessons:
framework · 3 heuristic · 2
Two pricing revolutions: usage-based then outcome-based
Source
Invest Like the Best with Patrick O'Shaughnessy — Gabe Stengel - Building Investing Superintelligence - [Invest Like the Best, EP.492] (September 22, 2026) · at 22:46 ▸
When this applies
when pricing any AI product whose cost structure and delivered value diverge from traditional seat-based software
The takeaway
Push pricing models to track realized customer outcomes rather than inputs (seats or tokens), because customers can verify outcome value but not token value.
Why it works
Seat-based pricing misaligns incentives once usage drives cost (more usage = more expense without more revenue); outcome-based pricing aligns price with realized customer value instead of consumption, solving the valuation-ambiguity problem of token costs.
One level up
Align pricing with the unit the customer actually values, and expect pricing models to evolve as a product's role shifts from informational tool to autonomous actor.
Where it stops holding
This stops working when outcomes are hard to define, attribute, or measure (e.g., diffuse, long-horizon value), or when buyers are structurally accustomed to a different purchasing model (e.g., enterprise seat budgets).
Detail
procedure:
- Stage 1: seat/enterprise-based pricing to match buyer habits (current Rogo state)
- Stage 2 (typical path): usage/token-based pricing
- Stage 3 (preferred, possibly skippable): outcome-based pricing — charge per good investment idea, per completed memo, per sim produced
benchmarks: {"examples":"charge per investment idea, per LP report, per banker sim"}
rationale: Token-based pricing creates disputes over whether $100k of tokens delivered $100k of value; outcome pricing sidesteps this because value is directly observable.
In their words
I think that every business needs to go through two different pricing revolutions. You need to move to some sort of usage-based, and then you need to move to some sort of outcome-based.
Company brain: continuous recording and retrieval for enablement
Source
Invest Like the Best with Patrick O'Shaughnessy — Gabe Stengel - Building Investing Superintelligence - [Invest Like the Best, EP.492] (September 22, 2026) · at 33:57 ▸ · at 36:30 ▸
When this applies
when organizational growth is bottlenecked by how fast new or existing employees can access institutional knowledge
The takeaway
Build a persistent, queryable organizational memory from routine work conversations so knowledge transfer isn't bottlenecked by who happened to be in the room.
Why it works
Recording and indexing all internal conversations creates a compounding knowledge base that any employee can query, eliminating dependence on tribal knowledge held by specific individuals across distant teams.
One level up
Growth bottlenecks caused by human ramp-time can be loosened by converting tacit knowledge into retrievable institutional memory.
Where it stops holding
This stops working when employees feel surveilled and withhold candor, or when the organization is too small for compounding knowledge-reuse to matter.
Detail
procedure:
- Record every internal conversation with transparency to participants
- Feed transcripts into a searchable company knowledge system
- Make the system prescriptive about company goals/values/North Star metrics
- Make it proactive (e.g., surfacing relevant context before a scheduled meeting) and reactive (answering queries on demand)
rationale: Core enablement problem is making a new hire productive fast; a company brain lets anyone pull in a relevant conversation a peer had weeks earlier, even across geographies, without ever having met them.
In their words
First off, every internal conversation that happens at Rogo is recorded. When anyone starts, we say, hey, just FYI, Patrick. That's how we do it. It's always being recorded, it's always being filtered into the company brain.
Diagnostic questions for incumbents facing an AI innovator's dilemma
Source
Invest Like the Best with Patrick O'Shaughnessy — Gabe Stengel - Building Investing Superintelligence - [Invest Like the Best, EP.492] (September 22, 2026) · at 50:18 ▸ · at 52:59 ▸
When this applies
when an incumbent firm with a historically simple, people-dependent business model faces a technology that can replicate its core expertise
The takeaway
Proactively audit which parts of your value chain rest on genuinely unique assets (data, relationships, context) versus merely perceived intelligence/expertise that AI can replicate, and systematize the former before competitors do.
Why it works
If enterprise value currently resides in tacit individual expertise but is migrating toward owned software/data/systems, firms that fail to systematize that tacit knowledge into an owned asset will lose their value to whoever does.
One level up
When a technology shift threatens to commoditize what currently sits in people's heads, the winning move is to convert that tacit value into an owned, operable system before someone else does.
Where it stops holding
This stops working when the firm's true value genuinely is irreplaceable relationship/trust capital that can't be systematized (e.g., small business owners wanting a handshake, per S21).
Detail
procedure:
- Assume 90% of enterprise value currently sits in people's tacit knowledge/relationships but will migrate to systems/data in ~10 years
- Start converting the latent knowledge of top performers into an owned, autonomously operating system
- Unpack every stage of the deal/value lifecycle and ask: is my advantage here a real unique asset (relationship, proprietary data) or just perceived superior intelligence that AI will obviate?
rationale: Cites John Momtazzi at Firm Moelis-adjacent example who built digital clones of top bankers so junior staff could leverage senior relationships/context directly — seen as prescient two years before it seemed credible.
In their words
and in 10 years, the world's best investment firms, best banks will have 90% of their enterprise value, not in but in software and data and systems. What would you start doing?
First-mover disadvantage reverses once models catch up
Source
Invest Like the Best with Patrick O'Shaughnessy — Gabe Stengel - Building Investing Superintelligence - [Invest Like the Best, EP.492] (September 22, 2026) · at 0:00 ▸ · at 3:09 ▸
When this applies
when building a product that depends on a rapidly improving underlying technology that isn't yet good enough
The takeaway
If you correctly bet on where an improving technology is headed and build for that end-state, being 'early' converts from a liability into a structural advantage once the technology arrives.
Why it works
Early demos look magical but fail in real use, creating a credibility deficit; but if the architecture was built correctly for the eventual capability level, performance jumps discontinuously once the technology catches up.
One level up
Position for where a trend is going, not where it is, but be prepared to look wrong for a long stretch in between.
Where it stops holding
This stops working when your read of the technology's trajectory is wrong, or when you can't survive financially long enough for the capability gap to close.
Detail
approach: Build the full-stack product for the capability level you believe is coming, accept early failure/criticism, keep iterating as models improve, and reap outsized gains when capability arrives.
signals:
- early users calling the product 'terrible/garbage'
- step-change jumps in underlying model capability
- product quality suddenly becomes 'magical' at a new model generation
rationale: Competitors who wait until models are obviously good enough arrive too late to have built the surrounding infrastructure.
In their words
But now, what we've seen is, well, if you were right about the end state and where the models were going and you were building towards that, when they get there, it's magical.
Target verticals with unbuilt plumbing, not just rich data
Source
Invest Like the Best with Patrick O'Shaughnessy — Gabe Stengel - Building Investing Superintelligence - [Invest Like the Best, EP.492] (September 22, 2026) · at 9:48 ▸
When this applies
when choosing which market segment to attack with a new automation technology
The takeaway
Prefer markets where the 'plumbing' (workflows, coordination, systems of record) is still unbuilt over markets that merely have abundant data, because plumbing is where durable value and moat accrue.
Why it works
Markets where coordination, standardization, and transacting are still done manually by humans offer more surface area to build defensible infrastructure than markets where the infrastructure already exists and the only lever is raw intelligence.
One level up
Look for depth of unaddressed operational complexity, not just abundance of raw material, when picking where to build.
Where it stops holding
This stops working when the unbuilt-plumbing market is too small, too regulated to enter, or when a well-funded incumbent can build the plumbing faster than a startup.
Detail
rule: Choose the vertical with the most un-automated coordination and transaction work, even if an adjacent vertical has richer data, because that's where you can build the full infrastructure stack.
threshold: Depth/TAM must be big enough to 'build a huge business just concentrating where I am today' before expanding ICP.
rationale: Private markets are attractive because coordination, standardization, and transacting are all still human-driven, unlike public equities where exchanges and infrastructure already exist.
In their words
I mean, part of the reason private markets are so attractive is because it's all done by humans, the coordination, the standardization, looking into things and the actual transacting. Whereas public equities, a lot of it has been automated.