Navan : AI Assistant for Seamless Business Travel & Smart Expense Management

Hospitality & Travel

Navan – AI Agent for Business Travel & Expense Automation

We worked with Navan to build Ava, an AI-powered travel assistant that streamlines everything from bookings to expense filing. Ava automates hotel suggestions, rebooking missed flights, and handling policy-based approvals in real time. Integrated with expense tracking, it uses OCR and NLP to read receipts and categorize expenses instantly. The platform also offers real-time analytics to monitor spending trends and ROI. This all-in-one assistant helps companies cut travel chaos, stay compliant, and give their employees a stress-free travel experience. It’s now used by teams globally for smarter, AI-led business travel.

Project Overview

Client: Navan (ex-TripActions | $300M+ annual revenue in travel tech)
Challenge: Complex travel booking processes + slow, manual expense tracking across large teams
Goal: Build an AI-powered agent to:
- Automate bookings, rebookings, and travel policy approvals
- Instantly categorize expenses using OCR + NLP
- Offer real-time analytics for budget control and compliance
Team: 8 (2 NLP Engineers, 2 OCR Specialists, 2 Backend Devs, 1 Product Strategist, 1 QA Analyst)
Timeline: 6.5 months (Prototype → Pilot with Fortune 500 clients → Global rollout)

“Ava isn’t just a chatbot—it’s an entire travel and expense department in your pocket. GenX helped us set a new standard in enterprise mobility.”

Head of Product Innovation, Navan

The Challenge

Critical Pain Points:

Travelers manually tracked expenses, leading to errors and lost receipts
Flight changes or disruptions created support overload during peak hours
Finance teams lacked visibility and struggled to enforce policy compliance at scale

Technical Hurdles:

Real-time decisioning for flight and hotel rebooking with corporate policy alignment
Parsing unstructured data from receipts and invoices in various formats
Scaling OCR + NLP pipeline to handle thousands of concurrent claims per day

Tech Stack

Component	Technologies
Chat Assistant & NLP	GPT-4, Dialogflow CX, Rasa, LangChain
Expense Management	Tesseract OCR, spaCy, Python NLP, AWS Textract
Data & Workflow Orchestration	AWS Step Functions, Lambda, Kinesis, DynamoDB
Backend + Integrations	Node.js, FastAPI, GraphQL, PostgreSQL
Analytics + Dashboards	Metabase, Looker, Redshift
Monitoring & Reporting	Datadog, Mixpanel, GA4

Key Innovations

Navan’s AI system managed instant trip changes, authorization workflows, and receipt-to-expense automation. It used policy-matching logic to cut delays and manual reviews. Users received full travel and expense support through one intelligent interface.

24/7 Smart Rebooking Bot

Ava auto-handled flight disruptions with zero human involvement

Result: 53% reduction in support tickets during peak travel times

OCR-Powered Expense Automation

Receipts processed and categorized in under 2 seconds

Result: 4x faster reimbursement cycles

Policy-Aware Booking and Approval

Built-in rules engine approved/rejected requests based on roles and budgets

Result: 37% drop in policy violations without added friction

Our AI/ML Architecture

Core Models

Conversational Travel Assistant (Ava):
- NLP-driven chat assistant using Dialogflow CX + custom GPT prompts
- Recommends, books, cancels, and rebooks with flight/hotel availability tracking
Smart Expense Categorizer:
- OCR + NLP pipeline using Tesseract + spaCy
- Auto-classifies spend (e.g., meals, lodging, transportation) based on company policy
Compliance & ROI Insights Engine:
- Rule-based filtering + anomaly detection models
- Generates dashboards for finance visibility, trend analysis, and fraud detection