Ariel Cohen on Navan, Travel Software and AI

Ariel Cohen says he learns intensely when a topic interests him and struggled to engage with subjects that did not. He describes himself as an average student whose performance varied sharply with his interest, and says rapid technological change makes him question how much of today's curriculum will remain relevant.

His advice to his children is not to feel compelled to enter college immediately at 18. He encourages them to take time to explore their interests, including through travel and meeting people; this is his personal view rather than a universal prescription.

Cohen says his earlier startup StreamOnce was sold after nine months with only two customers. Although the sale could be financially positive, he believes the company had pursued impressive technology in a small, unproven market without solving a sufficiently important customer problem.

With TripActions, later renamed Navan, he wanted to reconcile company needs such as reporting and efficiency with an experience employees actually want to use. He says his fear is that Navan could eventually become a company that relies on acquisitions and market power rather than continuing to innovate.

Cohen criticizes SAP and Salesforce as insufficiently innovative and predicts incumbents could be displaced. These are his competitive opinions as Navan's CEO, not established facts about those companies or their future.

At roughly 3,000 employees in the interview, Navan still needed to reconsider its own prior decisions, he says. The founders moved from travel into payments and expense management, and later rebuilt parts of service around machine learning rather than assuming the first design had to last.

On deciding what to prioritize, Cohen describes seeing a better design shortly before a planned app release. The team chose to make the change, then deliberately pause further design work so it could focus on other projects; he sees the willingness to make and communicate such trade-offs as important.

During the pandemic he shifted designers, product managers and developers from travel toward expense management. Cohen says many members of the product team disliked the change and left, but he believes it was necessary for the company. He acknowledges that leaders cannot always bring everyone along.

To decide whether to stop a project, he invokes a lean-startup approach: state a hypothesis, define measurements and check whether the expected outcome occurs. A first implementation may need revision, but teams should be honest when evidence does not support the original idea.

How long to persist depends on strategic importance. Navan's employee rewards for booking below company policy initially attracted little use, he says; because the idea was central to the strategy, the team changed the interface, explanation and algorithm rather than abandoning it immediately.

Cohen recalls that travel revenue disappeared in March 2020 and the company cut costs and made substantial layoffs. He nevertheless believed business travel would return and that travel and expense software remained important, so he chose to continue developing products and selling.

His pandemic sales argument was that a pause in travel offered companies a chance to change systems without disrupting active trips. Cohen says customers including Netflix, Heineken and Adobe signed up during that period; this is his account of those sales.

He told prospective investors that the company had no revenue and was burning cash, but argued that its plan would work if travel returned before the money ran out. He recalls most investors he approached disagreeing with that premise, while some backed the company.

Cohen distinguishes the later $9.2 billion valuation from the earlier pandemic financing: he says a roughly $5 billion round came as vaccines emerged, while the higher valuation came more recently amid a market correction. These are his descriptions of historical financing, not current valuation claims.

Asked how the company might justify a much higher value, he points to large annual travel spending and an opportunity in managed travel and expense. His estimate of market size and expectation that Navan can capture a broad share are investment arguments, not guaranteed outcomes.

Cohen predicts that incumbent travel and enterprise software will lose users as people demand easier mobile and AI-enabled tools. His sweeping claims about Salesforce, SAP, Concur and competitors are forecasts and criticisms from a founder with a direct commercial interest, not verified outcomes.

The rebranding from TripActions to Navan, he says, accompanies an ambition to combine travel, payments and expenses in one experience. He wants the product to serve both employees and company finance teams rather than simply placing unrelated features inside one app.

He gives the example of a frequent traveler repeatedly booking similar flights, hotels and restaurants. In his proposed experience, the system could use past preferences to prepare a trip, subject to current availability and pricing, while finance teams gain more visibility into expenses.

Cohen argues that older vendors built around workflows and approvals may struggle to adapt. He predicts that technology-led rivals will displace them; Stebbings challenges whether established sales channels, acquisitions and switching costs could preserve the incumbents' position.

Cohen says acquisitions and enterprise sales channels are strengths, but questions whether incumbents can retain the people and culture needed to keep improving products. He offers his own dissatisfaction with Slack's experience as an example; it is a subjective assessment, not independent evidence of product decline.

He says Navan is not for sale and has declined acquisition opportunities because the founders believe they can build something enduring. His own stated concern is that Navan could eventually become complacent in the same way he believes incumbents have.

Cohen recalls discussing machine learning in travel years earlier and says Navan used it to speed booking and support. At the time of the interview, he says the company's existing bot handled about 30% of support interactions and could help with changes, cancellations and receipts; these are company-reported claims.

He says OpenAI's technology represented a major improvement over the company's prior approach and that Navan planned to use it in its assistant, Ava, while relying on Navan's own APIs to carry out tasks such as changing a flight. He describes a February release of the first OpenAI-powered version as planned at the time.

Cohen forecasts that the assistant could handle 50% of contact-center interactions by the end of that year and 90% by the end of the following year, leaving agents to focus on difficult cases. These were projections in early 2023, not verified results or a promise that staff would be unaffected.

He attributes the improvement in part to models trained on large data sets. His example is a system recognizing that the Westin is a hotel and asking for dates, whereas he says the previous bot needed more explicit guidance; the interview does not establish an independent benchmark.

Cohen says Navan's gross margin was about 75% at the time and might reach 80% by year-end; he speculates about 85% in two to three years. He contrasts this with his estimate of competitors' margins, but the interview does not independently verify the figures or ensure they use comparable accounting.

He argues that automation can improve unit economics and predicts companies that do not adopt AI will struggle. This is his view of a future transition, not a demonstrated outcome for particular companies.

Asked about relying on an outside AI provider, Cohen says Navan's defensibility comes from operational infrastructure, supplier relationships, global service and the ability to act on a request, not only the language model. He also sees a cultural advantage in adopting new technology and replacing prior work when appropriate.

He says the rebrand includes opening the product more widely, offering free access at the lower end of the market and introducing a loyalty program. In his account, efficiency gives the company room to invest in access and customer incentives; these are business claims from its CEO.

Cohen recounts a delayed family trip as an example of the service: he says Navan's technology and human travel agents helped rebook him while other passengers faced disruptions. The story illustrates his experience, not an independently tested comparison with rival services.

He says the loyalty program is intended to reward employees for saving company money while preserving airline and hotel loyalty benefits. His goal is to make employees choose the corporate tool voluntarily rather than rely on enforcement.

Because Navan also handles expenses, Cohen says it can compare bookings inside and outside the platform to estimate how much of a customer's travel is captured. His reported capture rates vary by customer segment; the comparison with industry rates is his own estimate and is not independently verified here.

In quick-fire answers, Cohen says success means making his children proud and spending time with them. With hindsight, he says he would have invested even more aggressively during the pandemic had he known when and how travel would recover.

He says rapid pre-pandemic hiring taught him to be more careful about people attracted by hype, and advises founders to accept that hiring mistakes happen and correct them promptly. He also predicts American Express Global Business Travel will not remain a leading competitor; that is his competitive forecast.

Cohen would like venture investors to look beyond a Silicon Valley echo chamber. Looking toward 2028 from early 2023, he hopes AI will make people and businesses more capable and efficient; the outcome remains uncertain.

Stebbings thanks Cohen for the discussion, and Cohen closes the interview.