Personal Project • 2025

NYC Tour AI Chatbot

Designing a Conversational Planning Experience

Role Nirkhunan Kuppuram
Disciplines Product Design • Generative AI • UX Strategy • Interaction Design

Overview

Tour operator websites are typically brochure-heavy and decision-light. Users are presented with long package descriptions and PDF itineraries, creating cognitive overload.

This project reframes the problem: instead of just improving filters, what if the interface behaved like a calm, experienced travel consultant? The resulted conversational experience helps users understand why a package fits them, reducing bounce rates and increasing inquiry conversion.

The Challenge

"I don’t know where to start, and I’m not confident I’m choosing the right option."

The Industry Gap

Travel planning is stuck between two extremes:

  • OTAs optimize for search and comparison, assuming users already know what they want.
  • Traditional Operators rely on static content, placing the cognitive burden on key users.
  • Emerging AI Tools are often generic and destination-agnostic, lacking the contextual depth of a real operator.

Core Insight

People don’t struggle to find information. They struggle to interpret it and feel confident in their choice. The challenge is sense-making, not access.

Experience Strategy

The interface was reframed from information delivery to a guided conversation. The chatbot was not designed to "answer everything," but to guide, explain, and adjust.

Guide before selling

Users should understand the destination and experiences before being asked to choose a package. Education builds confidence.

Ask before recommending

Even minimal context (traveler type, duration) dramatically improves perceived relevance and trust.

Reduce cognitive load

Information is introduced progressively, avoiding long lists. Inspire confidence, not urgency.

Gen AI Strategy

Context-Grounded Generation

Generic LLMs often hallucinate. To act as a trusted tour operator, accuracy is non-negotiable.

I utilized Context-Grounded Generation to anchor the AI's responses in Enterprise Knowledge—the operator's specific tour catalog and brand guidelines. While the prototype uses static context, the architecture is designed for Retrieval-Augmented Generation (RAG) to scale with live inventory and pricing.

UX & Interaction Design

Conversation Mechanics

The conversation flows like a human consultation: Warm introduction → Light qualification → Highlight explanation → Recommendation with reasoning.

Key Decisions

  • Justification-First Instead of just listing packages, the bot explains why a recommendation fits the user's constraints (e.g., "Because you prefer a relaxed pace...").
  • Quick-Reply Buttons Reduces typing effort and guides the user through the happy path while allowing free text when needed.
  • Progressive Disclosure Detailed information (policies, logistics) is only shown when relevant, preventing overwhelm.
1
Welcome

User is greeted with a welcoming interface and selects trip duration using quick-reply buttons

Step 1: Full application interface showing welcome screen
Step 2: User expresses preferences
2
User Query

User expresses interest in iconic spots and nightlife

3
AI Recommendation

Bot explains why the package fits user's preferences

Step 3: AI provides tailored recommendation
Step 4: User confirms selection
4
Confirmation

User confirms selection with quick-reply buttons

5
Completion

Friendly sign-off with clear next steps

Step 5: Conversation complete

Visual Language

The interface is designed to feel calm, friendly, and non-salesy. It looks more like a "guided conversation" than a standard chat app. Visual hierarchy inside the chat uses cards and bullets to improve readability.

Value & Outcomes

For Business

Reduced time to inquiry, higher qualified lead submissions, and fewer repetitive agent questions.

For Users

Reduced decision anxiety, improved clarity between similar offerings, and a scalable expert guidance without pressure.