ExpertsFMBringing AI to the Real Economy | Alexander Taubman
David Senra · September 30, 2026
5 Selected Lessons:
framework · 2 heuristic · 2 reasoning_pattern · 1
Horizontal AI platform shared across verticals (Nexus architecture)
Source
David Senra — Bringing AI to the Real Economy | Alexander Taubman (September 30, 2026) · at 14:39 ▸ · at 17:31 ▸
When this applies
when building any internal platform intended to scale across multiple business units or product lines
The takeaway
Separate a platform into a large reusable core and a small customizable integration layer, so that expanding scope strengthens rather than fragments the core.
Why it works
Roughly 70-80% of the technical infrastructure (model routing, security, orchestration) is reusable across industries; only the last 20-30% — integration into specific workflows and data sources — is bespoke, so adding verticals compounds platform value rather than diluting focus.
One level up
Scale advantage in talent and technology compounds when you architect for shared infrastructure, not vertical-specific rebuilds.
Where it stops holding
This stops working when the verticals you add are too dissimilar in workflow/data shape, making the 'shared 70-80%' assumption false and forcing near-full rebuilds each time.
Detail
procedure:
- Build a model-agnostic orchestration layer (routing, security, agentic orchestration) as the shared core
- Embed engineers in the field with operators to learn workflow-specific pain points (the 'last 20-30%')
- Port solved capabilities (e.g., email drafting) from one vertical to all others
- Each new vertical makes the shared core better, creating compounding returns
benchmarks: {"shared_infrastructure_share":"70-80%","bespoke_integration_share":"20-30%"}
rationale: Scale is increasingly important in an AI world because you're competing for talent and because every new vertical makes the core platform smarter.
In their words
about seventy-eighty percent of the infrastructure is shared across the verticals and that's kind of like our, you know, it's our platform where on one side we integrate with all the models, we're model agnostic
Deal sourcing: wide funnel, prepared mind, tight filter
Source
David Senra — Bringing AI to the Real Economy | Alexander Taubman (September 30, 2026) · at 10:58 ▸
When this applies
when allocating any scarce resource (capital, hiring slots, product bets) across a huge field of opportunities
The takeaway
Maintain deliberately broad optionality (stay open to many categories) while applying an extremely tight final filter, because breadth of search increases the odds of finding rare, high-quality matches.
Why it works
Reviewing a large volume of deals across many industries builds pattern recognition ('prepared mind') so that when the right asset appears at the right price, it's immediately recognizable, even though the funnel converts to only one or two new industries per year.
One level up
Optionality plus selectivity, not narrow focus alone, produces compounding access to the best opportunities over time.
Where it stops holding
This stops working when you lack the organizational capacity (sourcing engine, analyst bandwidth) to process thousands of deals, turning breadth into noise rather than pattern recognition.
Detail
procedure:
- Maintain a force-ranked list of 15-20 target industries, updated continuously based on market learning
- Review thousands of deals per year across those industries (sourcing engine)
- Convert only 1-2 new industries per year — a very tight filter
- Recognize a 'you know it when you see it' asset by markers like high logo retention, high dollar retention, long-tenured customer relationships
benchmarks: {"deals_reviewed_per_year":"4000-5000","new_industries_per_year":"1-2","example_asset_metrics":"high-90s logo retention, 100%+ dollar retention, 15-year average top customer relationships"}
rationale: Analogy to Buffett's 'bisexual and public' dating quote — being open to more options increases odds of finding the great match.
In their words
We've built a really powerful sourcing engine where we're seeing 4,000 or 5,000 deals a year across all these sectors and we'll literally do one or two new industries per year. So it's a big funnel, very tight filter
AI abundance thesis: productivity gains grow headcount, not shrink it
Source
David Senra — Bringing AI to the Real Economy | Alexander Taubman (September 30, 2026) · at 5:01 ▸ · at 23:43 ▸
When this applies
when introducing any major productivity technology into an existing workforce
The takeaway
Treat productivity-enhancing technology as a growth lever that increases the marginal value of labor, rather than purely a cost-cutting lever that reduces headcount.
Why it works
If productivity gains let each employee serve more customers with better service, retention and growth increase, which raises the value of each employee and creates demand for more of them, not fewer.
One level up
Positive-sum capture of a technology wave requires reinvesting efficiency gains into growth, not just margin extraction.
Where it stops holding
This stops working when the market for the service is saturated or fixed-size, so increased capacity per worker doesn't translate into more customers or revenue — in that case efficiency gains do turn into headcount reduction.
Detail
approach: Compare the counterfactual: would you rather have fewer 3x-productive salespeople or more of them given fixed quota ceilings are removed? Reason from unit economics of retention and growth, not just cost per unit of output.
signals:
- rising customer logo retention
- rising dollar retention above 100%
- need for more capacity to serve growing customer base
rationale: Most historical productivity-boosting innovations correlated with wage and job growth rather than net job loss, per Taubman's reading of history.
In their words
If you think about it, if your people are 50% more productive and can handle 50% more clients and with better customer service. You're retaining customers better, you're growing faster. So the people are much more valuable. Do you want more of them or less of them?
Choose investors/partners for mission-alignment, not maximal valuation
Source
David Senra — Bringing AI to the Real Economy | Alexander Taubman (September 30, 2026) · at 49:28 ▸
When this applies
when choosing partners (investors, co-founders, vendors) for a relationship expected to last decades
The takeaway
When a relationship is meant to be long-duration, optimize for partner alignment and trust over short-term transactional terms.
Why it works
In a multi-decade partnership, the compounding value of trust, speed of decision-making, and shared ambition from aligned partners outweighs the one-time gain from optimizing terms in a single financing round.
One level up
Short-term optimization (price, valuation) can destroy long-term compounding value (trust, speed, ambition) — weigh decisions by their time horizon.
Where it stops holding
This stops working when the relationship truly is a one-time, short-duration transaction, where terms genuinely are the dominant consideration.
Detail
rule: Pick partners who 'make you more ambitious' and whom you'll want to work with for 20+ years, rather than chasing the best headline terms.
threshold: Explicitly rejected outside offers with 'obscene terms' in favor of existing trusted partners.
rationale: From day one they got the thesis. They believed in us almost in a way before we even believed in ourselves — that trust is irreplaceable.
In their words
they made the good point where a lot of founders fuck up because they optimize for valuation. Zach and Rick made the point, it's just like, well, if we're going be working together for twenty years, this is a long term partnership.
Raise the average: incremental hire must improve the team
Source
David Senra — Bringing AI to the Real Economy | Alexander Taubman (September 30, 2026) · at 23:43 ▸ · at 47:26 ▸
When this applies
when hiring at scale and wanting to maintain or improve average talent quality as headcount grows
The takeaway
Set an explicit bar tied to the existing team's average, not an absolute minimum, to guard against quality erosion during fast scaling.
Why it works
Explicitly requiring each new hire to exceed the current average prevents the natural dilution that occurs when hiring speed increases, since average quality otherwise regresses toward the applicant pool's mean.
One level up
Growth-stage organizations must make quality standards explicit rules, because implicit standards erode under hiring-speed pressure.
Where it stops holding
This stops working when the organization's need for raw headcount/speed outweighs marginal talent optimization, such as in commodity roles where the 'average' metric doesn't matter.
Detail
rule: Our rule is the incremental hire has to make us better, raise the average.
threshold: ~3% hire rate from ~1000 applicants in a recent quarter; ~75% of hires still come from trusted in-network referrals even as headcount scales.
rationale: Stated directly as their explicit hiring rule while describing the team's rapid growth from the founding group.
In their words
Our rule is the incremental hire has to make us better, raise the average.