Edge 2.0 Demo — NNY

Technical and product walkthrough with Navan's Edge engineering team. April 26, 2026.

Meeting Context

Farid and Ofir met with Navan's Edge engineering team for a full technical walkthrough of Edge 2.0. The goal: give our UX team enough context to run an independent design exercise grounded in real architecture and constraints.

In the transcript, "Me" = Farid. "Them" = various Navan speakers (primarily Snir, Danny, and Yaya). Tom was not present.

Name
Role
Affiliation
Notes
Farid Chaouki
UX / Design lead
Our team
"Me" in transcript
Ofir Kedmi
UX / Design
Our team
Present, less vocal
Snir (Yael)
Engineering lead, Edge
Navan
Primary technical presenter
Danny (N)
Architect, growth org projects
Navan
Yaya
Product / Engineering
Navan
Product context, demo
Langarov, Drabin, Dkoren
Engineering
Navan
Tel Aviv office + Zoom
Nbenjano, Yshraybom, Srozenbaum
Unknown
Navan
Tel Aviv office + Zoom

Engagement Scope

Navan's ask: Run a UX exercise covering conversation design, homepage, and search results. They're migrating from a proprietary platform called "Cognition" to a LangGraph-based architecture.

SCOPE AGREEMENT

Standalone POC

We build an independent prototype. No contributing to their codebase. No API keys. Fake data, real frameworks. LangSmith for agent prototyping. Farid: "We don't want to work in a vacuum."

DELIVERABLES

Prototypes, Not Figma

Documents + working prototypes in a GitHub repo. Real front-end code Navan can pull from. Farid: "Figmas for this type of product are not helpful at all." The team agreed on this format.

EXPECTATIONS

Two Tiers of Proposals

Navan wants multiple approaches: one audacious / ahead-of-the-curve, one industry-standard. Research-backed, structured output that tells a story.


Three Workstreams

Three distinct areas of work crystallized during the meeting. Each requires its own research, prototyping, and proposal track.

WORKSTREAM 1

Homepage & Navigation

What lives outside the chat. Trip surfacing, contextual components on the home screen, "where you left off." Navan confirmed components can render anywhere—not just in chat. Major navigation problem: users lose their conversation when they navigate away.

component surfacing trip lifecycle navigation

WORKSTREAM 2

Search & Discovery

How to help users who don't know exactly what they want. The "homework problem"—current results dump too much on the user at once. Flight/hotel result presentation, browsing vs. booking modes, multi-city complexity.

discovery mode result density browsing UX

WORKSTREAM 3

Conversation Design

Onboarding, tone, pacing, iteration vs. dump, personalization cold start. Users don't see value until ~5th session—most never get there. Navan wants proposals from conservative to audacious. Dia browser cited as onboarding reference.

onboarding tone of voice cold start
We don't want to work in a vacuum. And we don't want to suggest things that will make your life completely impossible.
— Farid, setting the engagement tone

Key Takeaways

The eight things that matter most for Tom coming into this engagement cold.

Prototypes, not Figma Working code in a GitHub repo. LangSmith for agent prototyping, real front-end for UI. Navan explicitly endorsed this.
Three workstreams Homepage/navigation, search/discovery, conversation design. Each needs its own research and prototype track.
TypeScript only Placeholder-based component injection. No stop/cancel mid-run. Mobile parity required. Web view fallback is the escape hatch.
Components live anywhere Not just in chat. Navan confirmed full control over rendering components on homepage, notifications, trip lifecycle surfaces.
1.0 is still better Hotel flows are better in 1.0. Look at both versions and design for the combined best-of-both.
Onboarding is the biggest open question Current "lazy" approach means most users never see personalization value. Proposals needed from conservative to audacious.
Proactive communication Disruption handling, "we already rebooked you"—stated goal, separate workstream. Relevant to conversation design principles.
Farid's Lastminute.com experience Prior art designing complex booking flows (~2001). The team's credibility on flight complexity is established with the client.

Action Items

Current status of all commitments from the meeting.

Owner
Action
Status
Snir
Share Edge 2.0 access link; add team to split if needed
Pending
Navan eng
Share coded components (JSX) for prototype reuse
Pending
Snir / team
Explore opening anonymized user conversations
Pending
Navan legal
Finalize contract (was expected 2026-04-28)
Pending
Farid
Call with Dana to agree on next steps
Pending
Farid
Look at Dia browser onboarding as reference
Pending
UX team
Review both Edge 1.0 and 2.0
Pending
UX team
Product deep-dive session (to be scheduled)
Pending
UX team
Prepare multiple approach options (audacious + standard)
Pending

Deep Dives

Detailed breakdowns from the meeting, organized by domain.