If Navan co-founder and CTO Ilan Twig is right, AI agents are poised to replace standalone LLMs in enterprises.
“My advice to CXOs, COOs and CIOs is that they need to embrace this technology. There is no other way.”
Twig shared his perspective in an interview with InformationWeek after his presentation at the recent Momentum AI conference in New York City. During the interview, Twig mentioned that Navan, an enterprise travel booking SaaS platform, used OpenClaw to build its recently announced service. TravelClaw agent tier.
OpenClaw is an open source autonomous AI agent. Navan’s TravelClaw is currently only used by Twig himself during development and testing. Once TravelClaw is released at scale, we expect it to proactively contact users via email or app to address their travel issues, rather than waiting for a prompt like a chatbot would. Twig said TravelClaw works continuously in the background, monitoring travel details.
The wave of agents that Twig foresees is certainly bullish comment Nvidia CEO Jensen Huang is urging CEOs to embrace the OpenClaw vision for their companies. Navan’s aggressive pursuit of agents and other AI technologies is part of Twig’s continued strategy to expand what relatively small teams can accomplish with AI tools.
And despite the “LLM is dead” rhetoric, Twig acknowledged that the introduction of ChatGPT was a seismic shift.
“Many people in the company know that I lost eight pounds researching this technology. I didn’t eat anything. This is a true story,” Twigg said.
He explained that the first step to unlocking ChatGPT’s capabilities in late 2022 was sourcing a “super expensive GPU,” which took some effort. “I reached out to my co-founder and was like, ‘I need $30,000 to build a PC.’ And he was like, ‘What?!’
Twig said he eventually secured the hardware himself after trying to explain the need. The reason is that, as a CTO, I wanted to experience ChatGPT first-hand before deploying it in production within Navan.
Twig likened the approach to vetting chatbots to interviewing job candidates before hiring them. This included switching to an uncensored model of ChatGPT that has not gone through the final stages of training on what is right and wrong or how to be politically correct. “I didn’t want to get involved in politics, but I wanted to understand the essence of politics,” he said.
These experiences have helped us find ways to make technology safer and more secure to protect organizations and their customers.
It also solidified his bullish view on agent AI.
All hands on deck for the rise of AI agents
The current agent craze is relatively new in the AI hype cycle. Twigg said that initially no one was seriously discussing AI agents, but the term was popularized at the World Economic Forum’s annual meeting in January 2025. “I was in Davos and that’s what everyone was talking about: agent-like, agent-like, agent-like,” he said.
Wayfair CTO Fiona Tan also We discussed AI agents at the Momentum AI eventand how they are being deployed in retail consumer goods and other areas.
Twig encouraged his colleagues to become familiar with technologies such as OpenClaw, even if it means working on separate machines. This is how Navan’s TravelClaw is developed. “Don’t use LLM, use the agent system.”
He believes that OpenClaw could be the beginning of a substantial AI revolution.
“For me, it’s as big as ChatGPT. [first] It came out. “I don’t think most people realize that… Interview me in two years and you’ll see that everything is in the spirit of OpenClaw,” he said.
There are many benefits and advances that Twig has discovered while exploring OpenClaw. First, he explained that OpenClaw will make LLM more proactive and autonomous, moving its capabilities from functions to responsibilities.
That “sense” of responsibility is evident in the fact that Twigg runs Navan’s TravelClaw agent on his phone, which monitors Navan’s travels, itineraries and locations. “It knows everything. It doesn’t wait for me to prompt it to do something on my behalf. It prompts me,” he said.
For example, TravelClaw knew that Twig needed to arrive at the airport within a few hours of the interview, but was actually concerned about whether he would arrive on time, so to speak. “It’s a shift in responsibility. I want to make sure I get there on time and don’t miss my flight.”
Weigh the freedoms and powers given to agents
TravelClaw’s autonomy includes launching itself whenever it wants to perform functions according to Twig’s needs. Navan’s AI-powered travel assistant, Ava, also speaks to TravelClaw without his knowledge to resolve issues on his behalf. For example, if you requested an aisle seat but it wasn’t initially available, TravelClaw will continue to process the reservation on its own.
“Every day I automatically wake up and talk to Eva and see if there’s an available aisle seat on the plane,” Twigg said.
If such seats become available, TravelClaw will confirm with users via messaging apps such as Telegram, WhatsApp, iMessage, and Slack.
Twig also added a little challenge by asking TravelClaw to make restaurant reservations for her, even though Ava personal assistant doesn’t have a restaurant reservations feature yet. TravelClaw then attempted to contact a live human agent to complete the task, prompting engineers to intervene.
“It was really interesting. It’s about how far the agent will go to solve my problem. I don’t care about that. The agent is responsible,” Twigg said.
However, its aggressive nature must be controlled. He warned that TravelClaw could destroy his computer, so he is not giving any freedom to TravelClaw company-wide at this time. As the world learns to work with agents, Twigg said he believes Navan’s early start on this technology has prepared it for future possibilities. “I’m ready for the next one.”
