Feb 7, 2026 · 54m · 20vc
20VC: From $6.2BN Market Cap to $2.8BN: What Is Not Translating About Navan's Public Story | Are Any Public Company CEOs Actually Happy? | Why Navan Built It's Own Customer Service AI and What it Could Mean For Customer Service AI with Ariel Cohen
0:00 / 0:00
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of TwentyVC, host Harry Stebbings interviews Ariel Cohen, co-founder and CEO of Navan, exploring the operational realities behind public market valuation shifts, proprietary AI architecture, and corporate strategy. Cohen shares lessons on navigating public market scrutiny, building domain-specific AI models, maintaining software retention, and preserving CEO focus amidst stock volatility.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 28.4% of the talking time here. How this is scored →
Harry as informed peer 5.5 Guest teaching 5.1 Guest disagreement 3.9 Harry pushing back 5.4
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 20:59 Dismissing generic AI agent demos
Ariel forcefully dismisses third-party AI agents and generic LLM demos as completely unsuited for complex travel execution where hallucinations lead to lawsuits.
Hardest push from Harry ▶ 16:54 Direct pushback on Salesforce thesisHarry explicitly rejects Ariel's framing that Salesforce is obsolete, citing distribution power and AI sales rep ROI as reasons why it remains a screaming buy.
Biggest teaching moment ▶ 20:59 Vertical travel API integration realitiesAriel educates Harry on why building bespoke vertical AI is necessary, detailing the strict zero-tolerance requirements for flight changes and credit applications.
Harry holds his own ▶ 25:12 Citing founder data on model performance shiftsHarry demonstrates high domain knowledge by bringing up recent interviews with founders of Harvey and Legora to refute Ariel's claim that LLM infrastructure is commoditized.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Reflections on Navan's Journey and the Pre-IPO Process | 5 | 4 | 3 | 5 | Harry asks a pointed question about whether Navan was forced to go public due to its debt structure. Ariel calmly reframes the narrative around payments business advantages and enterprise customer demands. Harry shows good preparation by bringing up capital structure details. | |
| IPO Pricing Dynamics and Long-Term Market Strategy | 6 | 5 | 4 | 6 | Harry invokes Bill Gurley's IPO pricing framework and challenges Ariel on whether private competitors like Ramp have an unfair burn advantage. Ariel rejects the premise of worrying about named private competitors, highlighting historical examples of well-funded startups that failed. | |
| Software Obsolescence and Public Market Perception | 7 | 5 | 5 | 8 | Harry directly rejects Ariel's assertion that legacy enterprise software like Salesforce is obsolete, arguing forcefully for distribution moats. Ariel counters by insisting end-user sentiment and retention matter more than enterprise distribution. | |
| Real-World AI Infrastructure: Building Ava vs. Buying Third-Party Agents | 5 | 6 | 4 | 4 | Harry asks why Navan built its own AI agent Ava rather than buying solutions from specialized vendors like Sierra or Decagon. Ariel educates Harry on the zero-hallucination domain complexity of travel rebooking that generic third-party agents cannot execute. | |
| AI Paradigm Shifts, Model Wars, and Vibe Coding Engineering | 6 | 6 | 4 | 5 | Harry challenges Ariel's view that LLM infrastructure is a commoditized layer by citing firsthand insights from recent podcast guests. Ariel details internal engineering shifts including how his co-founder vibe-coded an expense product over a weekend. | |
| AI Cost Structures, Margin Realities, and Enterprise Market Expansion | 6 | 5 | 4 | 6 | Harry presses on AI margin degradation and inference costs, arguing that public markets lack patience for long-term payback periods. Ariel explains direct airline integrations and value creation model to maintain margins. | |
| Public CEO Morale, Equity Fluctuations, and Talent Retention | 6 | 6 | 4 | 5 | Harry brings up Jason Lemkin's claim that no public CEO is happy and relates a previous interview with Robinhood CEO Vlad Tenev regarding stock price correlation to morale. Ariel compares public market stock scrutiny to navigating the COVID travel shutdown. | |
| Quickfire Round: Mindset, Personal Life, Parenting, and Optimism | 3 | 4 | 3 | 4 | In a quickfire wrap-up, Harry playfully traps Ariel in a logical paradox when Ariel expresses regret over missed time with his kids while admitting the company's success required that sacrifice. Ariel acknowledges the dilemma thoughtfully. |
Statements from this episode (23)
Stebbings: Navan went public at a $6.2B valuation
“So last month, Nevan went public. They went public with an IPO price, or a market cap, of over 6.2 billion dollars.”
Ariel Cohen plans to build Navan for another 20 years
“11 years in this journey, and probably at least for me, another 20, 20 years to go.”
Cohen: Public status gives payments companies a major capital advantage
“In the payments business, I think there is huge, huge advantage to be a public company in the way that you are actually raising capital for that part of the business compared to being a private company. I think if you kind of go too long in the payments busine…”
Cohen: Enterprise customers require public company transparency and stability
“We have been starting to take, share in the enterprise segment, assigning more and more and more customers, and for them, for that type of segment, it is actually important that you're going to be public. They want the transparency. They want to understand you…”
Cohen: Unmanaged corporate travel is larger than the managed market
“There is the managed side, you know, Amex and Concare. There is the non-managed. Everybody that are doing whatever they want to. It's actually a bigger market, and we think that we have an opportunity and the right to win in both markets, and that's what matte…”
Cohen: Upside Travel raised $200M versus Navan's $4M in the same year
“When we started Navan, There was this company, Upside Travel. We raised four million dollars. They raised 200. Same year, ok?”
Cohen: Avoiding margin scrutiny in private markets builds weak businesses
“This mechanics of maybe it's better to be in the private market because nobody will look at how I'm defining revenue, or what's gross margins, or what contribution margin, I think it doesn't create a good business.”
Cohen: Navan fears unknown early-stage disruptors over visible competitors
“I tell you what is the only competitor that I'm worried about. The one that I don't know about. The one that sits, like, the ones like Navan, 1011 years ago, that are ignorant enough about the problem, but are really, really, really good, that will create some…”
Cohen: Form-based software like Salesforce will be obsolete within 10 years
“I think I called out Salesforce back then, as I cannot imagine people using something like this in the next 10 years. I think AI accelerates this.”
Stebbings: Salesforce is a screaming buy due to its distribution advantage
“I think they're an incredible buy, because I think the only thing that matters is distribution. And when you have an AI sales rep that works, hey, sales reps powered by AI that make you money while you sleep. Everyone will sign up. Yes, please. And all that ma…”
Cohen: Navan has only lost six enterprise customers ever; five returned
“In the history of Navan, we lost six enterprise customers. Five of them are back, ok?”
Cohen: Navan's AI handled 55% of support chats during JFK shutdown
“55% of the chats were run through Eva, ok? Our call center, in the worst time during this event, The wait time was 16 minutes, one six.”
Cohen: Sierra, OpenAI, and Anthropic Fail at Complex Vertical AI Needs
“There is nothing even remotely closed to what is needed when you go to a complex vertical out there. Okay. And it's definitely not in the kind of Sierra level, but even if you take it to the level of the APIs, If I'm kind of taking an anthropic and an open AI,…”
Cohen: LLM Travel Demos Are Nowhere Near Production Readiness
“It's all of the cute demos that you sometimes see by LLMs, like you can book your trip, whatever, or you can, it's really nice cute demos. Trust me. It's not even remotely close to being there.”
Cohen: Market obsession with LLM infrastructure is misdirected as models commoditize
“I think if we'll move from the LLM infrastructure discussion, which I think people are obsessed there, which I actually don't think it's an important discussion. It's also a very commoditized kind of thing.”
Cohen: Navan's internal AI routing is shifting significantly toward Google models
“We've seen our agentic platform, Cognition, plays between models based on who will give you the best answer, and actually we saw movement much more to Google recently. A lot of the transactions with Edge, which we didn't release yet, but you know, we have it i…”
Cohen: Navan's deep travel ticketing plumbing creates an unassailable moat
“An area that I actually think it's almost impossible to disrupt us is the travel side. To create the infrastructure that we've created to bring the content into the platform, the plumbing there, I don't think that people fully understand the plumbing that you …”
Cohen: Navan's co-founder vibe-coded their expense product in six hours
“I gave the team a challenge, and actually Ilan, my co-founder, obviously took it first and did it over the weekend to vibe code our expense product. And it took him six hours to do that.”
Cohen: Navan will ultimately capture the entire corporate travel market
“And I think as the, as long as we will service our customers and our users, the business will continue to gain share. We're going to take the entire market. I really believe in it.”
Stebbings: Passing on a $10M seed round cost me a $1B gain
“I passed on deal at the seed round at a ten million dollar valuation to put in a large check. I think about that every single day because that's a billion dollar gain.”