The Twenty Minute VCAriel Cohen: The Death of Salesforce; How OpenAI is Changing the Travel Industry | 20VC #975
CHAPTERS
- 0:00 – 0:35
AI “eating software” and the urgency to adapt
Ariel opens with a broad claim: AI is now the force reshaping software itself, and companies that don’t adapt will disappear. This frames the rest of the conversation as a story about reinvention, not incremental optimization.
- •AI is the next wave after “software is eating the world”
- •Non-adopters risk becoming irrelevant
- •Sets up a worldview of continuous rebuilding and disruption
- 0:35 – 1:51
How Ariel learns: curiosity-driven skills over formal education
Ariel explains why traditional schooling didn’t work for him and how self-directed learning shaped his approach. He and Harry discuss how fast-changing technology makes much of formal curriculum feel quickly outdated.
- •Learns best only when deeply interested
- •School performance was “spiky”: A’s in passions, failures elsewhere
- •Advice: focus on discovering what genuinely interests you
- •Belief that AI/ML will make many “things we learn today” obsolete
- 1:51 – 3:10
Advice to his kids: delay the pressure, explore, and travel
Ariel shares how he guides his teenagers on education and career decisions. He emphasizes taking time to explore before committing, and highlights travel and meeting people as a key way he forms conclusions.
- •No need to force college at 18
- •Travel and real-world exposure as education
- •Give yourself time to find authentic interests
- •Long-term planning is harder in an AI-accelerated world
- 3:10 – 5:17
StreamOnce lesson: don’t invent a problem nobody needs solved
Reflecting on his earlier company, Ariel describes why it didn’t create lasting impact despite a financially OK outcome. The core mistake wasn’t management—it was choosing an unproven, small market and over-indexing on “cool tech.”
- •Small team, early sale, limited opportunity to scale management
- •They built impressive technology without strong demand
- •Market size/proof matters more than technical elegance
- •Contrast with desire to build a multi-decade company at Navan
- 5:17 – 6:40
The founding mission: “magic” between company controls and employee delight
Ariel explains the original TripActions thesis: build enterprise software people actually like, without sacrificing compliance, reporting, and efficiency. He positions legacy enterprise giants as cautionary examples of stagnation.
- •People-centric enterprise software as the category
- •No compromise between CFO needs and employee experience
- •Salesforce/SAP cited as examples of innovation decay
- •Fear: Navan someday innovates only via M&A and market power
- 6:40 – 8:50
Staying innovative at 3,000 people: the courage to obsolete your own work
Ariel details how Navan tries to maintain speed at scale by repeatedly “killing” prior approaches and rebuilding with new tech. He describes internal ML systems (like “Katie’s IQ”) that continually learn to improve service automation.
- •Innovation requires recurring self-disruption
- •Shift from travel-only to payments and expense management
- •ML-driven automation for hard operational workflows (refunds, cancellations)
- •Internal learning systems that improve with usage signals
- 8:50 – 13:03
Prioritizing product vs distraction: change fast, then lock focus
Ariel walks through a last-minute design overhaul two weeks before launch and how leaders decide when to embrace change vs stabilize execution. The deeper point: teams often resist change by default, and great product orgs manage that tension explicitly.
- •Example: late-stage redesign decision for the unified app
- •Leaders must make decisive calls and align the team quickly
- •After major change, intentionally freeze scope to ship
- •Fear of change is a core enemy of innovation
- 13:03 – 16:25
Killing projects with discipline: hypotheses, metrics, and strategic importance
Ariel describes how Navan evaluates initiatives using lean-startup logic: define a hypothesis, measure honestly, pivot or kill. He also explains when to “double down” despite early failure—if the bet is foundational to the strategy.
- •Product work exists to test hypotheses, not just “build features”
- •Hard metrics prevent self-deception
- •Navan Rewards initially failed; they iterated UI/algorithms until adoption
- •Time horizon depends on strategic centrality vs optionality
- 16:25 – 20:44
COVID zero-revenue moment: cost cuts, layoffs, and doubling down anyway
Ariel recounts the shock of losing product-market fit overnight in March 2020 and the decision tree: kill, hibernate, or continue. He argues they kept going because the underlying beliefs (T&E matters, travel returns, people-centric software wins) remained intact.
- •Revenue went to zero; PMF disappeared overnight
- •Rapid mental reset: old plan no longer relevant
- •Cut costs and did significant layoffs
- •Simultaneously doubled down on innovation and go-to-market
- 20:44 – 24:49
Raising through the crisis and scaling into a $9.2B valuation
Ariel explains the fundraising narrative: the core investor bet was simply whether travel would come back before the company ran out of cash. He also argues the long-term opportunity is massive given $1T travel spend and poor legacy adoption due to bad user experience.
- •Pitch: “Bet on travel returning; we’ll execute”
- •Most investors said travel would never return; a few leaned in
- •Large customers signed during downtime because change management was easier
- •Thesis: legacy tools force usage; modern tools win by being loved
- 24:49 – 28:52
From TripActions to Navan: rebrand, super-app ambition, and unified experience
Ariel ties the name change to a category shift: from a travel tool to a broader navigation layer across travel, payments, and expenses. The challenge is not bundling features, but delivering one coherent experience that creates value for both employees and CFOs.
- •Why rebrand: signal a new category and broaden scope
- •Navan name meaning (navigation + avant) and “inviting” brand idea
- •Super-app goal: one workflow across travel, payments, expenses, personal trips
- •CFO value: real-time visibility across spend, especially in downturns
- 28:52 – 36:02
Competition and incumbents: why Salesforce/Concur/AmEx GBT can’t “just respond”
Ariel argues legacy enterprise players have lost the ability to innovate and rely on M&A plus sales channels—an approach that fails during platform shifts like AI. He cites Slack’s decline inside Salesforce as an example of innovation and product quality decaying post-acquisition.
- •Incumbent playbook: M&A + distribution channels
- •Platform shifts (cloud, now AI) break that model over time
- •Claim: forcing employees to use workflows will become unacceptable
- •Acquisitions often fail to retain the innovative teams that built the product
- 36:02 – 48:45
Navan’s AI stack: OpenAI-powered Ava, automation gains, and margin impact
Ariel details how Navan uses ML/AI in search and support, then explains why OpenAI is a leap forward. He predicts the new Ava will handle the majority of support interactions, allowing human agents to focus on high-stakes edge cases—and supports this with strong margin economics.
- •Ava already handles a large share of support; OpenAI accelerates capability
- •Example: LLMs infer context better (e.g., recognizing “The Westin” as a hotel)
- •Forecast: bots replace most interactions; humans handle the hardest 10%
- •Navan gross margins cited at ~75% trending higher via automation
- •Moat: operational infrastructure + culture of self-disruption, not the model alone
- 48:45 – 51:09
Loyalty and retention: defining true adoption via “attainment”
Ariel explains how Navan measures loyalty differently because it can detect out-of-channel bookings through the expense side. He defines retention as consistent, near-total “attainment” (employees actually using the platform for all applicable spend) and uses feedback loops to fix drop-off.
- •Retention measured by whether employees book outside the system
- •“100% attainment” as a practical definition of true retention
- •Higher segments reach near-total attainment; downmarket is lower but still strong
- •Proactive outreach to understand non-usage and build features like loyalty clubs
- 51:09 – 54:19
Quickfire: success, fatherhood, hiring, and the next five years of AI
In closing, Ariel shares personal definitions of success and leadership lessons from scaling. He also reiterates his view that AI will broadly increase efficiency—and may make people smarter rather than simply replacing them.
- •Success = making his kids proud; fatherhood as a top priority
- •Regret: would have invested even more aggressively during COVID with hindsight
- •Scaling lesson: avoid “hype-joiners”; use resets to correct culture
- •Hiring: expect mistakes and correct quickly
- •2028 outlook: AI as a “new calculator” that amplifies human capability
