Jake Saper, GP @ Emergence Capital: "We Sold Salesforce Early and Lost Out on Billions"

Summary

• Emergence’s numbers are the credential for everything else said: a little less than $2B deployed over 20 years, “a little over $8 billion in cash” returned, with fund three at about 16x DPI, by Saper’s estimate — Zoom alone returned that fund more than 10x, and SalesLoft’s $2.3B sale to [likely Vista] was “the highest multiple ever paid by private equity for a software company.” Graduation stats from the new-fund analysis: 9/10 early deals raise follow-ons, 1/5 raise at >$1B, 1/10 go public.
• The Zoom deal in 2014 was $20M from a $250M fund at $200M post on ~$2-3M revenue — 100x revenue when that was unheard of — and diligence found Eric (likely Yuan) was miscounting upgrades and pauses as churn: “Eric thought the business was worse than it was,” the only time in Saper’s career a founder underestimated his own business. Telling him before terms were final gave up leverage but won the deal.
• A possible “Quadruple 120” replacement for triple-triple-double-double: great AI-era companies should roughly 4x year-over-year with ≥120% net dollar retention — Bolt went 0→20 in two months, Together AI 2M→100M+ revenue in 15 months. But the unproven variable is retention: cohorts, Saper predicts, “will disappoint,” with outliers saved by sticky workflow wedges.
• Ranking for any investment: market pull > founder > traction. The diligence tell is a user saying “if my boss stopped paying for this I’d quit” — and the trap is “Mirage product-market fit,” where fast growth (a COVID fitness-instructor tool; AI services that grow because they’re cheaper without proving margin) masks the absence of durable demand or a business model.
• Tape calls: he wouldn’t short Salesforce — he’d short IBM, because 75% of the Fortune 500 still run core apps on COBOL mainframes and AI (via portfolio company Mechanical Orchard) is “the critical enabler” to finally migrate them. Buy Anthropic at $60B, sell [likely Grok] at $50B, OpenAI at $300B “feels expensive”; ten-year single-stock hold is Microsoft as a B2B software index.
• Biggest change of mind in 12 months: he feared LLM value would accrue to incumbents via data and distribution, but “what I underappreciated… is the value of focus” — narrow startups are outpacing incumbents, and buyers aren’t firing anyway: a voice-AI customer cut zero headcount but “grew my business three times with the same headcount.”
• Most venture partnerships are structurally broken: two-year checkbooks incentivize spraying, departures create “orphaned deals” that quietly hurt founders, and retiring founders hoarding carry drives the spin-out merry-go-round. Emergence’s founders forfeited carry on retirement — why no partner has ever left. Exit discipline matters too: they sold Salesforce right after IPO (“it was bad”), but case-by-case management of current public positions returned $2B more than selling at lockup.

Deep dive

1. Zoom: a prepared mind, an 8% fund bet, and a founder who didn’t know how good his business was

• Emergence had a 2013 thesis that WebEx was “a tired product” ripe for replacement — Saper’s interview case at the firm was diligencing Fuze, an early Zoom competitor (he correctly passed). So when Zoom arrived as his first deal, “they basically gave me the test before the test.” Partner Santi’s proof point was personal: calling family in Argentina, the rebuilt codec “works way better than everything else in the market.”
• The deal was anything but easy: a $20M check from a $250M fund — ~8% concentration — at $200M post on ~$2-3M revenue in 2014, the firm’s largest check at among its highest prices. Competitive partly because Zoom was profitable and didn’t need the money.
• The churn numbers wouldn’t tie in Saper’s model, so teammate Joe (PE background, “math Olympiad kid”) dug in and found Eric (likely Yuan) was counting tier upgrades and pauses as churn — “Eric thought the business was worse than it was.” They told him before finalizing negotiations, handing him leverage, “because we thought it was the high-integrity thing to do.” Eric’s response: “I want to work with you.”
• The winning commitment: layering a proper enterprise sales motion on top of product-led growth — they hired Dave Burman from RingCentral post-investment. Saper’s general law: “all PLG companies eventually need to layer on enterprise software motions to be sustainable and enduring.” Eric, he confirms with diplomatic timing, “was not an experienced salesperson.”

2. Why founders choose you, and what “prepared minds” are actually worth

• Harry’s Rabois-flavored pushback — the best founders don’t need you — gets a reframe: a founder picks a VC because “you’ll help them bend the odds of success,” via go-to-market, hiring, therapy, whatever. “The odds of success of these things are so low that if you can even bend the odds incrementally, it matters a lot.”
• On thesis-driven investing, Saper is blunt: “I don’t think any venture firm makes all their money on prepared minds — that is bullshit.” What a prepared mind does is get you there faster. And the best ones come from the portfolio: Chorus.ai — a seed deal, “the first voice AI company,” acquired by ZoomInfo for a little under $500M at a multiple Saper thinks was north of 5x — seeded the broader voice-AI thesis behind Bland, Regal, and Assembled.
• Harry’s honest jab lands: a 5x on a seed doesn’t move a fund of Emergence’s size. Saper concedes — which is precisely why the fund math below matters.

3. Fund three’s power law: about 16x DPI, and top-tier even without Zoom

• Fund three was at about 16x DPI, by Saper’s estimate. Zoom returned the fund more than 10x; SalesLoft — a non-consensus Series A — sold to [likely Vista] for $2.3B, “the highest multiple ever paid by private equity for a software company,” returning the fund a few times on its own.
• The striking claim: strip Zoom out entirely and the fund is still top-tier on SalesLoft, Chorus, DroneDeploy and a just-distributed multi-billion-dollar crypto outcome in Zapo (likely Xapo). Outliers drive the $8B returned, but the bench beneath them is real.

4. Market structure: both winner-take-all and crowded markets pay — and PE is no savior

• Doximity (“LinkedIn for doctors” — “a fucking beast and no one knows about it,” per Harry) was underwritten as winner-take-all and is “absolutely hitting” in public markets. Bolt/Lovable is the opposite case: many large incumbents (Webflow, Wix, Weebly, Squarespace) support multiple large next-gen winners. The incumbents’ fate: mostly consolidation and private equity, with a few nimble enough to reinvent.
• On PE as exit backstop: “savior is too strong a word” — SalesLoft’s 20x ARR was the exception; “most cases they’re paying whatever 3 to 5x.” Harry’s aside on Anaplan and Coupa: “how the fuck are they going to get that money back.”
• The $50-100M ARR SaaS cohort funded in the last five-to-seven years faces “this existential moment” — how much to invest in agents, whether customers adopt — and, “unfortunately,” most of them will make it. Counter-example: Guru, where Rick (Series A 2016, Saper on the board a decade) paired knowledge management with gen-AI search, cut to ~60 employees profitably, and reaccelerated growth — “in some ways these moments actually make the business.”
• Harry’s devil’s-advocacy: is a re-architected Guru venture-interesting when Glean is at $100M ARR? Saper’s reframe: fast-growing competitors prove market pull; Guru’s burden is offering something different that taps the same pull.

5. You don’t know your winners quickly — Bill.com and the Zenefits parable

• Bill.com was not up-and-to-the-right: it stalled through the financial crisis until Emergence helped crack the bank channel partnership strategy (Bank of America was, Saper thinks, the first) to sell a low-ACV product economically — then it “absolutely took off” and made fund one.
• The humility story: in 2015 a peer investor, possibly a principal, announced “I just made my career-defining investment” — Zenefits. Emergence had just backed Gusto, growing slower. Gusto endured; Zenefits didn’t (no knock on Parker, “a monster”). The lesson, verbatim: “just because something is a breakout right after you invest doesn’t guarantee it’s going to win. The breakout can indicate market pull… but it doesn’t necessarily indicate an enduring company.”

6. Market pull is everything — and how to detect the fake kind

• The definition: desperation. “It’s not a desperate problem if your buyer hasn’t tried to hack together something on their own to solve it, or bought an inferior product… otherwise it’s nice-to-have.” The diligence tells: “if my boss stopped paying for this I’d quit” or “I’d pay for this out of pocket.”
• Harry’s ranking question gets a clean answer: market pull, then founder, then traction — because it’s the founder’s job to convert a market-pull wedge into defensibility. Bolt’s underwriting hypothesis was WebContainer technology, though “it’s still very early, so we have no idea how that’s going to play out.”
• The confession: a ~$9M Series A into a COVID-era business-in-a-box for exercise instructors (Saper’s mother is a nearly-70-year-old likely Jazzercise instructor, hence the soft spot). Market pull existed “briefly,” then gyms reopened. The founder, to her credit, shut down and returned most of the money. Harry piles on — even success was maybe a $500M outcome — and Saper owns it: COVID forced assumptions about a permanently remote economy, and “our job is to call the future.”

7. The best deals are often — not always — expensive, and the “what you have to believe” chart

• From the portfolio: Gusto, Zoom, Yammer, Ironclad were expensive; Veeva and SalesLoft were non-consensus and not expensive. Recent proof it’s still possible: a partner led [likely Federato] (AI underwriting for insurers) before the zeitgeist at a good price, pushed through the doubts, and the Series B came at a much higher mark. “You want to be in both.”
• The internal framework: identify the 3-5 deal-specific “what you have to believes” for the investment to return the fund — dilution and future capital needs, defensibility, market, competition, team — then build a chart of data supporting and negating each, refined through every diligence call, “and we can all stare at this and say, on balance, do we believe it.”

8. The broken-partnership antidote: everyone diligences everything, one deal per partner per year

• Emergence is focused three ways: only B2B software, ever (Salesforce → Veeva → Together, Bolt, Bland, Unify); one investment per partner per year on average; and growing partners from within. “Priority deal” are “holy words” — every partner’s calendar blows up, every partner does customer calls, management references, backchannels, and many do on-site visits. The contrast with the industry: instead of an associate-and-partner defending a deal “against an onslaught of questions and doubters,” it’s “a process of seeking truth collectively.”
• The logistics objection (time compression) gets a real answer: seven partners plus principals can run seven diligence calls in one slot, coordinated by a quarterback, with recorded calls, detailed notes, nightly summary emails, and deliberately late-night synthesis calls — “after kids go to bed… you have theoretically an unlimited amount of time on the back end.”
• Harry’s needle: Saper has never had a zero — “that means you’re not taking enough risk.” Saper half-agrees (“some of them may still go to zero”), crediting recurring-revenue downside protection, but insists the team model bends outcomes post-investment: 9/10 deals raise successful follow-ons, 1/5 raise at north of $1B, 1/10 of early-stage investments have gone public.
• The signature war story: Regal’s hot COVID Series A (Saper thinks six term sheets and 0→$1.5M in a year, pre-AI). Saper and CEO Alex Levin each flew to Denver and took a four-hour walk in a field outside the airport — Saper, sunscreen-less at altitude, “horribly red” — while co-founders did 1:1 Zooms with every partner. Harry’s pushback on process-heavy selling — “I don’t fucking want to miss Revolut” — draws the counter: they win young hot founders too, co-leading UniFy’s “founding round” with Austin Hughes (ex-Ramp, co-founder ex-Scale AI), terms agreed before the company was incorporated.

9. Quadruple-120: the possible new benchmark — and the retention rooster that hasn’t come home to roost

• Triple-triple-double-double “is just not true anymore” because market pull itself has exploded — the LLMs are genuinely good and “everyone’s boss is saying go buy AI.” Evidence: Bolt 0→20 in two months; Together AI from $2M to north of $100M revenue in ~15 months (Harry, deadpan: “they have shit marketing” — the growth is that organic).
• The proposed replacement: “quadruple 120” — roughly 4x year-over-year growth plus net dollar retention of 120%+. Hold both and “these are generational companies.” But the caveat is loud: retention is the metric these companies are too young to have proven, and Saper thinks cohorts “will disappoint” on average, with outliers that find sticky wedges.
• On being “revenue funnels for OpenAI/Anthropic,” he’s unbothered for two reasons: closed-model price competition means app-layer gross margins are rising, and open-source LLMs are a credible fallback — if OpenAI ever 10x’d pricing, “you now have a credible ability to spin up an open-source model” (which is Together’s whole business).
• Together’s clearest “what you have to believe”: open-source LLMs become a dominant part of the enterprise market. His honest status check: “trending positively… but it’s still frankly a little TBD” — 80/20 vs 60/40 matters. Harry’s counter: even 10% of every company in the world is huge; Saper’s concession: it also depends how painless Together makes the spin-up — “that 10 could be the whole world.”

10. Why software vendors survive the build-it-yourself era — and how AI gets priced

• His investing lens comes from Mike Maples’ Pattern Breakers: back founders with a unique insight on an external inflection — which is why he “cares a little bit less” about replacement vs. new markets. Meta-detail: he hand-wrote his synthesis of the book — “I retained more by doing the synthesis myself versus asking an AI to do it.”
• The 2017 “coaching networks” thesis (partner Gordon’s — “poorly branded but I think correct”): AI shows up as a coach, learns from real outcomes across a network, and improves recommendations. “Copilot is the term that took off.” The kicker: domain-specific models on that data yield insights “even an OpenAI won’t be able to have.”
• Against the likely Clay-style “we’ll build all our SaaS ourselves” thesis, three reasons vendors endure: you’re buying “an opinionated perspective on how to solve a problem,” not code; the same tools that make software easy to build make it hard to maintain — “that thing becomes out of date in six months,” which is why enterprises boomerang from build back to buy; and third, most important, “the buyer wants a throat to choke.” Harry adds creation and scale: most enterprises don’t know what Slack is, and the average company runs 172 tools — self-maintaining them is “absolutely moronic.”
• Pricing is a spectrum — per-seat → usage → true outcomes — and the market currently lives in usage. Pure outcomes pricing (à la Fin’s resolution-based model) is “hard for now”: multi-touch tickets make causality contestable, and monthly outcome negotiation turns the vendor-buyer relationship antagonistic. The exception where outcomes pricing already works: AI-enabled services — Mechanical Orchard doesn’t sell its “cursor for mainframes,” it sells the migration itself: half the time, 80% of the incumbent’s price, “and if it doesn’t work, you don’t pay.” Pricing on labor is “taking a bet on yourself: how good is my AI” — margin catastrophe if it fails, outsized margin capture if it works.

11. The change of mind: focus beats incumbent data and distribution

• His biggest reversal of the past 12 months, stated as such: “I was fearful when the power of LLMs came out that most of the value would accrue to the incumbents because of their data and distribution advantages. What I underappreciated — which is just the recurring lesson of startups — is the value of focus.” A likely Unify running narrowly at one go-to-market problem simply outruns Salesforce. Growth-stage companies are “a mixed bag.” Harry’s counter-observation stands alongside: incumbents like Adobe are shipping faster than the old caricature.
• The answer to “Google could just build this”: start narrow, because Google won’t solve the narrow problem as well as you will — then expand. The canonical case: Veeva’s pharma-CRM TAM was $400M globally at investment; the company is a $35B market cap today, having become “the board-level vendor” to big pharma, which is what unlocks the upsell machine. Corollary both agree on: “you continuously underestimate how big the landing wedge is” — especially now that AI wedges capture labor spend, not just software spend.
• On Sarah Tavel’s pay-for-the-work thesis, his empirical hedge: buyers so far aren’t firing anyone — they’re not hiring. His new healthcare voice-AI investment’s customers reduced zero headcount: “I love it because I’ve grown my business three times with the same headcount.” Vendors should capture some of that labor value, but the substitution is emotional and slower than the pitch decks assume.

12. Tape calls: short IBM not Salesforce, buy Anthropic, and the coming “FTX moment in B2B”

• Everyone on 20VC shorts Salesforce; Saper won’t — tens of millions of weekly users on core workflows is real incumbency. His public short is IBM: “75% of the Fortune 500 still run their core applications on refrigerators in their closet,” written in COBOL “that no one writes anymore,” while IBM bills billions annually in maintenance and servers. AI is “the critical enabler” to migrate that spaghetti code — if the Mechanical Orchards succeed, “the IBMs of the world are in trouble.”
• Rapid-fire book: at anthropic $60B / [likely Grok] $50B / OpenAI $300B — sell [likely Grok] (“I don’t yet know what niche they’ve carved out”), buy Anthropic if they figure out the app layer, OpenAI “feels expensive but they have a strong consumer brand.” Ten-year single stock: Microsoft, explicitly as “the best index” on B2B software — “you don’t want some super risky GameStop shit.”
• Near-term adoption: expect “a little bit of a trough of disillusionment” — experimental budgets are converting to real ones, but non-delivering vendors get cut. And the darker scenario, his phrase: a possible “FTX moment in B2B” when some enterprise agent “does something really bad” — rogue emails, rogue purchases — triggering backlash. “We do need to figure out the guardrails so that they’re deployed safely.”

13. Orphaned deals, forfeited carry, and why Emergence has never lost a partner

• His anatomy of the venture merry-go-round: firms hire ex-CEOs or “an army of junior people” with a two-year checkbook — incentivizing check volume — and two years is never enough to judge. High performers then look up, see founders who “retain carry after they depart,” and leave to start their own firms. The under-discussed casualty is founders: “orphaned deals” where the sponsoring investor departs, pro-rata odds drop, a less constructive board member arrives — “they’re getting something that’s not what they bought.”
• Emergence’s structural answer: an equal partnership where retiring founders forfeited their carry — “not something that’s talked about in venture,” and something Saper didn’t know when joining 11 years ago. “I have no reason to leave… I can look you in the eye recruiting you as a principal and say you have a real chance to be an equal partner alongside me.”
• On seed signaling risk — real or a seed-investor talking point? Real, when multi-stage firms run seeds as option programs: small checks, junior people, tracking-to-double-down. Emergence’s counter: treat seed “like a core bet” — partner-led, double-digit ownership so the founder can raise the next round elsewhere without signaling damage. Ownership floors flex via the framework — “we did Zoom at 10%” — as long as the deal can still return the fund.
• Harry challenges the return-the-fund fixation itself (half a fund is great; winners are underestimated). Saper’s data: the bulk of $8B+ returned on <$2B deployed came from a handful of companies — but, back to fund three, some funds stay top-tier even without them.

14. “We sold Salesforce too early” — and the $2B case for managing publics from the board

• The scar: Emergence sold Salesforce shortly after IPO. “It was bad.” His broader point: exit matters, and it doesn’t get talked about enough in venture.
• The framework: every earnings quarter, the deal sponsor — usually still on the board, which is the only “defensible reason to hold” — updates the partnership with inside information, and decisions happen inside legal windows. Sometimes they buy: Doximity, purchased more at IPO, has already returned 3x from there.
• The self-audit for LPs (Salesforce excluded — “we made such a bad decision”): had they dumped everything at lockup expiry, they’d have returned $2B less; had they sold every position at its (often 2021) peak, $2B more. Actual case-by-case management sits squarely between — “we made you $2 billion more than if we’d just given you it all right away.” Their LPs, mostly charitable foundations and endowments, want maximized outcomes over immediate liquidity.
• On reserves, his failure taxonomy is “Mirage product-market fit”: fast growth that fools you — either a diverse customer base all using the product for different things, or AI-enabled services growing “because it’s cheaper and faster” without proving a high-margin model — “you pour more cash in, and you don’t [have PMF].” Against Maples’ “99% of bridges are a bridge to nowhere”: Intacct — the #2 cloud ERP — was saved by an Emergence bridge on favorable terms, cracked an accounting-firm channel, and sold to Sage for what Saper thinks was about $1B. “Obviously a cherry-picked example.” Pay-to-plays “suck… but sometimes necessary.” IPOs return “beginning of next year,” on his guess.

15. Losses, mock board meetings, and Sam Altman’s parenting advice

• Deals lost: I think two or three in a career. The first was likely Ironclad, to Jess Lee at Sequoia — Saper recused himself from the Series A over a portfolio conflict (Simple Legal had contract management on its roadmap), the B came fast, and “to Jess’s credit she ran fast.” He got back in as I think the only external investor in the next round, priced somewhere in the $300 millions, maybe $400M; the company is north of $100M revenue with Jurist, a fast-growing Harvey competitor — he thinks materially more than the ~5x Harry sketches.
• Craziest thing done to win a deal: Assembled’s frothy Series A (three Stripe founders; I think Stripe’s first-ever seed investment). After making the top three, the CEO called at 9pm: materials in an hour, then a mock board meeting to judge how Saper showed up as a board member. They won — he thinks over Index. Related stat from fund five: Emergence knew founders an average of 13 months before investing.
• Worst exit: Comfy (energy-efficiency software) — seven-figure bookings from Salesforce and others, but the people deploying it didn’t care: “a huge incentive issue between the buyer and the implementor,” big bookings, weak deployed ARR. Sold to [likely Siemens] for a small gain; the CEO’s thank-you was a French Laundry gift certificate he still can’t get a reservation to use.
• The close: likely Sam Altman, three months ago, on raising kids amid GPT-2→3.5-scale gains likely to be dwarfed in the next two-three years — “don’t teach them to code… teach them how to understand how people are thinking and feeling and how to influence that.” Saper’s optimism: the rote work goes away and “we can hopefully be more human” — his wife now teaches sales and persuasion at Stanford Business School, so “you’re the future.”