Topics Pioneers Insight Method Author
Back

How Salesforce Chooses the AI Companies It Wants to Buy

·The a16z Show

Rendering the long image…

Topics Legacy Software Buyouts and AI Overhauls Autonomous Agents Reshaping Enterprise SaaS Early-Stage VC Investing Dynamics SaaS Valuation Reset Under AI Disruption AI Venture Funding and Bubble Risks
Pioneers John Somorjai

Summary

  • Salesforce’s acquisition logic pairs scarce capabilities with distribution at scale. Somorjai pitches access to 15,000 sales reps and faster global reach—“as widely as possible and as quickly as possible”—but buying and integrating a company can still take six to nine months. Cheaper software development therefore strengthens the need to compare acquisition timelines against internal delivery.
  • Outcome pricing gives Salesforce a way to capture AI value as customers need fewer seats. Somorjai reports that Fin deployed internally in 14 days, handles over 40% of help.salesforce.com traffic and delivers 65% resolutions without a human. Through “outcome-based pricing,” software can capture more of the labor budget while reducing customers’ overall costs, as Immerman frames it.
  • Slack’s acquisition thesis gained an AI role Salesforce had not anticipated. Somorjai says revenue is almost three times its acquisition level and Slack is Salesforce’s fastest-growing product, though integration initially took time. The original distribution thesis—“great technology that was starved for distribution”—now benefits from an interface where humans, applications and agents work together.
  • Salesforce Ventures seeks future enterprise markets, with acquisitions representing a small subset of its investments. Salesforce has acquired 29 of more than 750 companies backed; its $1 billion AI fund, established at the beginning of 2023, sits alongside an expanding interest in robotics. Asked about investing in potential Salesforce competitors, Somorjai says, “I think we probably should.”
  • AI CRM remains a disagreement between incumbent confidence and startup adoption potential. Somorjai has yet to see enough novelty to support a large multibillion-dollar challenger and argues, “There are too many of them.” Immerman counters with faster deployment, adoption further downmarket and Stripe’s example of growing alongside initially small customers.
  • High private valuations can leave companies with too few viable exits. Somorjai estimates “probably one percent or less” go public and says the buyer universe becomes very small above $10 billion; a reckoning is “inevitable,” although he does not know when. Immerman pushes back that today’s largest outcomes are substantially bigger, while Somorjai distinguishes rare generational companies from ordinary businesses receiving similar valuations.
  • ARR growth can conceal the problems that kill an acquisition. Somorjai focuses on customer attrition, financial controls and security, which has veto power across Salesforce’s 14 diligence functions. One penetration test reached source code, the CEO’s laptop and employee information; the exposure was “unrecoverable.”

Deep dive

1. Salesforce buys time, reach and entrepreneurial talent

  • Immerman puts acquisition activity during Somorjai’s tenure at nearly $100 billion, with 12 deals in the past year. Customer requests and Benioff’s “beginner’s mind” inform three tests: strategic fit and customer success; whether Salesforce can accelerate the target with compatible technology and people; and value for shareholders and customers, including valuation and time to accretion. Somorjai calls the last test hardest because “everything is expensive.”
  • The distribution pitch is concrete: Salesforce has 15,000 sales reps and more than 8,000 ISV partners. Founders can partner, sell or keep building, but recreating sales coverage across small businesses, enterprises and global markets takes years, additional fundraising and dilution. Somorjai’s question is whether to undertake that work or reach customers “as widely as possible and as quickly as possible.”
  • Cheaper development does not settle build versus buy. Product teams map three-year roadmaps against available expertise and scarce AI engineers; an acquisition can take six to nine months from idea through integration, so internal development might still win. Listen Labs supplied the opposite case: Somorjai’s team launched a survey in ten minutes with “no enablement whatsoever,” then used the product to survey its customers during diligence.
  • Integration also refreshes management: acquired founders have become Salesforce’s chief product officer four times. Somorjai’s example is David Schmaier’s Velocity, which brought six industry verticals; Salesforce subsequently added seven, reaching 13. The acquisition changed how Salesforce understood industry-specific use cases, while Schmaier moved through chief product officer to chief strategy officer.

2. Outcome pricing connects AI efficiency to Salesforce revenue

  • Salesforce’s headless platform raises Immerman’s seat-count question: customers using AI may need fewer seats. Somorjai points to consumption pricing—already used by MuleSoft—and a shift toward paying for outcomes, such as resolved support requests. Immerman’s economic framing is that software can capture more of the labor budget while lowering the customer’s overall cost of selling.
  • Service Cloud is almost $11 billion in AOV and rests on a 2008 acquisition, but Somorjai says Salesforce could not build the new combination of AI applications, outcome pricing and fast deployment quickly enough. Fin supplied that capability and an AI team in Dublin. He also values founder Eoghan’s willingness to challenge the incumbent: he is “really trying to shake us up.”
  • Fin’s average implementation is 16 days; Salesforce’s internal pilot took 14. Somorjai reports that it now handles over 40% of help.salesforce.com traffic and delivers 65% resolutions without a human. Qualified extends labor substitution into sales: its Piper agent answers website visitors through a lifelike avatar and books meetings automatically. It “can replace your sales development rep organization, essentially,” he says.

3. Slack’s distribution thesis gained an unforeseen AI upside

  • The 2020 Slack thesis was distribution and a common interface for Salesforce products; agents came later. Slack was “losing to Teams,” Somorjai says; a bad quarter and stock decline helped trigger discussions about combining a beloved product with Salesforce’s reach.
  • Somorjai acknowledges the outcome was uncertain: pandemic diligence happened largely through Slack, without many face-to-face meetings, and integration “took a little bit of time” to hit its stride. He now says revenue is almost three times the acquisition level, roughly five years and a quarter later, and calls Slack Salesforce’s fastest-growing product. His explanation: “great technology that was starved for distribution.”
  • The AI payoff came later: Slack is “the perfect conversational interface for AI,” in Somorjai’s view. Immerman calls it the “unofficial org chart,” containing reporting relationships, interactions and decision-making context. Somorjai connects that work history to richer Slackbot answers; the collaboration layer originally intended for humans and apps now also accommodates agents.

4. Ventures follows future enterprise value beyond software

  • Salesforce Ventures began in 2009 with ecosystem investments such as DocuSign, nCino and Twilio, plus regional services partners. The mandate widened to Snowflake and cybersecurity through Wiz, then a $1 billion AI fund at the beginning of 2023. Somorjai names Anthropic, Mistral and Cohere, ElevenLabs in voice and Runway in video; the through-line is future enterprise value.
  • Physical AI pushes that mandate beyond software. Somorjai highlights Figure AI, Physical Intelligence, Dyna Robotics and Zipline’s drone delivery, linking robotics progress to improved dexterity and mobility. Immerman adds that early signs of scaling laws and more vertical go-to-market strategies are beginning to translate into revenue growth. Customer introductions and operating advice remain Salesforce Ventures’ main pitch to founders.
  • Asked whether he would back a future Salesforce competitor, Somorjai says, “I think we probably should.” Staying close can lead to a partnership, an acquisition or neither. Of more than 750 companies funded, Salesforce has acquired 29—a relatively small subset. The ventures team’s thousands of annual company meetings also feed information about traction and technological direction back into Salesforce’s decisions.

5. AI CRM challengers face a disputed opening

  • Somorjai’s skepticism is sharper on new CRMs: “I have yet to see something so novel that it’s going to turn into a large, multibillion-dollar company in that space.” Existing products are strong, Salesforce keeps improving, and small differences are insufficient. Immerman counters with Lightfield and Monaco, then suggests fast deployment and bottom-up adoption could create an opening further downmarket.
  • Somorjai concedes, “I could see that,” but calls it a small opportunity. Immerman’s stronger counterexample is Stripe: start with small companies, then grow alongside them. Somorjai does not resolve that analogy; he returns to market crowding—“There are too many of them”—and says venture capital should fund new, important categories rather than numerous niche CRM variations.

6. High valuations can exhaust the buyer pool

  • The valuation objection is exit arithmetic: Somorjai estimates “probably one percent or less” of companies go public, while valuations above $10 billion leave very few potential buyers. He calls a reckoning inevitable but explicitly says, “I don’t know when.” He also thinks unsustainable valuations will hurt employees; maximizing the next-round price is “usually not the best decision.”
  • Immerman tests whether power-law outcomes justify the pricing, positing Anthropic above $1 trillion and citing SpaceX at $2 trillion. Somorjai calls such businesses “generational companies” and “very rare.” Salesforce invested in six Anthropic rounds because its first meeting with Dario and Daniela convinced the team they were exceptional; he contrasts that conviction with ordinary ideas receiving $10 billion-plus valuations.
  • Immerman argues that today’s top-1% outcome is larger than a decade ago, citing Wiz and Cursor. Somorjai agrees he is an optimist but sees thousands of companies that don’t make it; many come to Salesforce seeking acquisition, a part of the industry the press doesn’t cover. Those unpublicized cases, he argues, belong in founders’ fundraising and valuation decisions.

7. Churn and security can outweigh headline ARR

  • Deal trouble often appears after price and major terms are agreed, during deep diligence. Somorjai flags missing customer-contract data, weak financial controls and fixation on ARR without understanding attrition. High churn can indicate unhappy customers and product-quality problems; ARR growth alone does not answer whether customers will renew. Organized data can materially accelerate the process.
  • Security has an explicit veto across Salesforce’s 14 diligence functions: it can say, “I’m killing this deal.” In one abandoned transaction, penetration testers accessed the company’s source code, the CEO’s laptop and employee information. Salesforce shared its findings, but the exposure was “unrecoverable.” Somorjai’s preparation advice is to red-team the technology before the buyer runs its own tests.
  • Not every vulnerability kills a deal. Salesforce can distinguish P0, P1 and P2 problems, require some fixes before signing and others before closing. Somorjai says the outcome depends on how severe the problems are.