The Twenty Minute VCJake Saper, GP @ Emergence Capital: "We Sold Salesforce Early and Lost Out on Billions"
CHAPTERS
Emergence Capital’s track record and why this conversation matters
Jake opens with Emergence’s fund-level performance and “graduation metrics,” framing how the firm thinks about picking and helping winners. The stats set up the themes of concentration, power laws, and the craft of diligence that recur throughout the episode.
- •Emergence has deployed just under $2B and returned over $8B cash (plus private holdings)
- •Graduation metrics: follow-on rates, unicorn rounds, and IPO frequency
- •The power-law reality: a handful of companies drive most returns
- •Context for why process, ownership, and selling decisions matter
Zoom: thesis-driven entry, pricing nerves, and a churn diligence surprise
Jake recounts Zoom as his first deal at Emergence, rooted in a pre-built thesis that Webex was ripe for replacement. The diligence hinged on underwriting a very early, very expensive, highly concentrated check—and uncovering that the founder had miscomputed churn in a way that made the business look worse than it was.
- •Thesis: opportunity to replace a tired Webex; earlier diligence on competitor Fuze sharpened conviction
- •Zoom was ~ $2–3M revenue yet priced at ~$200M post (100x revenue in 2014) with a $20M check from a $250M fund
- •Diligence found churn was miscalculated (upgrades/pauses counted as churn), meaning the business was stronger
- •Integrity move: shared the finding pre-negotiation; helped win founder trust and the deal
- •Zoom’s PLG + later enterprise sales layering became the scaling playbook
Founder profiles, VC value-add, and the ‘prepared mind’ debate
Harry pushes on whether domain experts outperform naïve young founders, and whether “prepared minds” are real or just VC packaging. Jake argues both founder archetypes can win and that prepared minds help you get there faster—but most venture outcomes aren’t solely thesis-driven.
- •Domain expertise and youthful rule-breaking both produce great companies
- •Founders choose VCs to “bend the odds” via GTM, hiring, advice, and emotional support
- •Prepared minds aren’t the sole source of returns, but can accelerate recognition and conviction
- •Portfolio learnings can create future theses (e.g., early voice/AI learnings)
Fund outcomes and power laws: Zoom’s 16x DPI and what else mattered
They dig into how much Zoom contributed to a standout fund and what other outcomes supported it. Jake explains why even strong multiples can be immaterial in a large fund and how a fund can remain top-decile even without its single biggest winner.
- •Fund 3 achieved ~16x DPI; Zoom contributed the majority and returned the fund 10+ times
- •Other meaningful wins: SalesLoft (acquired for $2.3B; major multiple for software buyout), Chorus, and more
- •Why 5–10x exits may not “move the needle” in a large fund despite being great outcomes
- •Power-law dynamics with nuance: concentrated winners plus a strong supporting cast
Market structure and post-COVID reality checks: winner-take-all vs shared markets
Harry asks how Jake thinks about markets where several players coexist versus winner-take-all dynamics. Jake contrasts network-effect businesses (e.g., Doximity) with tool markets (e.g., no-code/app-building) where many incumbents can persist—then pivots to post-COVID whiplash and the risks of mistaking temporary pull for durable demand.
- •Winner-take-all tends to appear in network-effect businesses; Doximity as example
- •Tool/platform markets can support multiple large players; app-building/no-code as current case study
- •Expected outcomes for incumbents: consolidation, PE involvement, and a few reinventions
- •COVID-era bet example: “business-in-a-box” for fitness instructors had transient pull; founder returned capital
How to rank founder, market, traction—and what ‘market pull’ sounds like
Jake presents his ordering: market pull first, founder second, traction after—because pull indicates desperation and founder capability determines defensibility. He shares the specific phrases he listens for in diligence to validate true pull, especially in B2B.
- •Ranking: market pull > founder > traction (as a signal, not the driver)
- •Market pull definition: buyers are desperate; they’ve tried hacks, inferior tools, or waste hours without a solution
- •Diligence cues: “I’d quit if my boss stopped paying” or “I’d pay out of pocket”
- •Traction can be misleading; pull doesn’t guarantee endurance without defensibility
Valuation, competition, and the ‘What You Have to Believe’ diligence framework
They move from “are the best always expensive/competitive?” into Emergence’s internal diligence tool: identifying the 3–5 deal-specific beliefs required for fund-returning outcomes. Jake explains how factors like dilution, defensibility, competition, and fundraising needs are explicitly tested and documented.
- •Great companies are often expensive, but non-consensus opportunities can be cheap (e.g., Veeva, SalesLoft)
- •Competition isn’t universal; unique insight and timing can reduce competitive pressure
- •‘What You Have to Believe’: 3–5 deal-specific hypotheses that must hold to return the fund
- •Each belief is supported and challenged with data; the output is a pro/con evidence chart
- •Dilution and future capital needs are modeled as part of the underwriting
Why every partner does reference calls: Emergence’s ‘priority deal’ process under time pressure
Harry challenges Emergence’s unusually collective diligence model—every partner participates in calls, references, and often on-sites. Jake explains why this is possible (B2B-only focus and low deal volume), how it scales under time compression, and how it changes decision quality and post-investment support.
- •Emergence focus: only B2B software; each partner averages ~1 investment/year enabling deep collaboration
- •‘Priority deal’ means calendars get cleared; all partners do customer and reference calls plus backchannels
- •On-sites provide unstructured signal; multiple partners collect first-party data to ‘seek truth’ collectively
- •Time compression benefit: parallel processing—7+ people can run many calls in the same time window
- •Knowledge-sharing: recorded calls, detailed notes, nightly recap emails, and late-night synthesis calls
AI-era growth, retention risk, and margins: from ‘triple-triple-double-double’ to ‘quadruple 120’
They discuss how AI has changed growth expectations and why legacy SaaS heuristics may mislead founders. Jake argues the open question is retention and proposes a new mental model that pairs extreme growth with strong net dollar retention, while also addressing LLM cost structure and margin durability.
- •AI has increased market pull; some companies grow at unprecedented speeds (e.g., 2M to 100M in ~15 months)
- •New benchmark idea: ‘quadruple 120’ (very high growth plus 120%+ net dollar retention)
- •Key unknown: retention cohorts for many AI companies are still unproven and likely to disappoint on average
- •Gross margin concerns are tempered by model competition and improving open-source alternatives
- •Together.AI belief: open-source LLMs become a dominant-enough enterprise choice to support a huge business
Where value accrues in AI: focus beats incumbency, specialized models, and ‘coaching networks’
Jake explains how his view shifted: incumbents have data and distribution, but focused startups can outrun them. He also explores specialization—especially domain-specific copilots/coaches—where proprietary workflow data and feedback loops create durable advantage.
- •View change: less fear that incumbents capture most AI value; focus and speed remain decisive advantages
- •Three buckets: incumbents, startups, and ‘growthy-stage’ companies that can still pivot (e.g., Notion-like)
- •Specialized LLMs and domain copilots will proliferate, often built on open source
- •‘Coaching networks’ thesis: software as a domain coach that learns from outcomes and improves recommendations
- •Sticky workflows matter; entrenched daily use creates resilience even if the product isn’t ‘best’
Why AI won’t replace vertical SaaS: opinionated solutions, maintenance burden, and accountability
Responding to claims that companies will just build custom tools with AI, Jake gives three reasons vendors persist. He emphasizes that buying software is buying an opinionated approach, continuous maintenance, and—crucially—a responsible counterparty (“a throat to choke”).
- •Vendors provide an opinionated best-practice solution, not just code
- •AI lowers build cost but increases maintenance pressure as systems become obsolete quickly
- •Enterprises repeatedly learn ‘build vs buy’: building is possible; maintaining is the constraint
- •Accountability: buyers want uptime, support, guarantees, and someone to blame when things break
- •This pushes pricing toward usage and eventually outcomes-based models in some categories
Pricing in an AI-first world: usage-based today, outcomes-based later, and AI-enabled services
They unpack the evolution from seat-based pricing to usage and outcomes, including why outcomes-based is hard with human-in-the-loop workflows. Jake highlights that outcomes-based works best when a provider owns delivery end-to-end, blurring software and services.
- •Pricing spectrum: per-seat → usage-based → true outcomes-based
- •Outcomes-based friction: causality disputes, multi-touch workflows, and monthly renegotiation risk
- •Near-term reality: experimentation with usage metrics (tokens, interactions, tickets)
- •AI-enabled services make outcomes-based simpler by owning delivery (e.g., mainframe-to-cloud migrations)
- •Margin dynamics: provider takes execution risk; big upside if AI works as promised
AI adoption risks: trough of disillusionment, agents’ ‘FTX moment,’ and Salesforce vs IBM shorts
Jake predicts a classic adoption curve: experimentation moving to real budget, but many tools will be cut if they don’t deliver. He warns that autonomous agents could trigger a high-profile incident that slows deployment, and he defends Salesforce’s durability while identifying IBM as a more attractive short.
- •Expect a ‘trough of disillusionment’ as weak AI tools get culled from enterprise budgets
- •Agent risk: powerful action-taking systems may cause a major enterprise failure and prompt backlash
- •Salesforce bull case: massive embedded daily workflow makes displacement hard
- •Jake’s short pick: IBM due to AI enabling mainframe/COBOL modernization and cloud migration
- •Adoption timing: near-term volatility, long-term inevitability
Losing billions on Salesforce: when to sell public positions, LP pressures, and exit discipline
Jake explains Emergence’s post-IPO selling governance and admits the firm sold Salesforce very early after IPO, missing enormous upside. He describes a quarterly, board-informed decision process for public holdings and shares analysis showing their active approach has helped versus selling at lockup—while still leaving money on the table versus perfect timing.
- •Salesforce was sold shortly after IPO; a painful lesson on the importance of exit timing
- •Quarterly sponsor-led review after earnings; board-level insight informs hold/sell decisions within legal windows
- •Emergence sometimes buys more post-IPO (e.g., Doximity) when conviction remains high
- •LP dynamics: charitable/endowment LPs often prefer outcome maximization over immediate liquidity
- •Analysis (excluding Salesforce): selling at lockup would’ve returned ~$2B less; selling at peaks would’ve been ~$2B more
Quickfire: lost deals, craziest competitive tactics, carry structure, and what the next decade optimizes for
In rapid-fire, Jake covers the rare deals he lost, how he wins competitive rounds, and how Emergence avoids partner churn via internal promotion and founders forfeiting carry on retirement. He closes with a human-centric view of AI’s long-term impact—freeing people from rote work and emphasizing persuasion, emotion, and connection.
- •Lost deal example: Ironclad (initially lost to Sequoia; later joined at Series C)
- •Craziest win: mock board meeting to earn Assembled’s Series A amid heavy competition
- •Emergence retention: grow partners from within; retirees forfeit carry enabling true equal partnership
- •Career-building insight: relationships matter; Emergence knew founders ~13 months before investing on average in a recent fund
- •Long-term optimism: AI reduces rote work; future advantage shifts to human understanding, influence, and creativity
