FreshBot for FreshDirect: Designing Conversational Experiences for Everyday Grocery Needs (A Concept)

Designing a conversational chat-based assistant to help FreshDirect customers track orders, discover deals, and manage returns through guided interactions, enhancing their online grocery experience.

⏰ Timeline

Feb’25 - Apr’25 (2 Months)

👥 Team

Roshni Ganesh

Eric Lopez

Julia Maloof

👩🏽‍💻 My Role

  • Conversational UX Design

  • Conversational UI Design

  • Usablity Testing

The Chosen Service

FreshDirect (founded in 2002) is an online grocery service delivering fresh, high-quality food in the New York tri-state area with a focus on convenience and healthy options.

Currently relying solely on human support, FreshDirect presents an opportunity to enhance customer experience with a fast, seamless chat-based assistant.

Our Goal

To design a conversational chat-based assistant to help FreshDirect customers quickly track their orders, discover deals, and manage returns, enhancing their online grocery experience.

The Chatbot Design Process

Identifying Task Flows

We started by analyzing FreshDirect's app flow to identify key user intents and subtasks.

Key User Intents > Subtasks

1

To look up Grocery Items

    • Browse using Departments

    • Browse using Filters

    • Browse using List (Ingredients List)

    • Browse using Recipes

    • View Item Details

    • View current discounts/sale deals on items

    • View current coupons

    • Add discounts/coupons to the order

    • Reminder for expiring discounts/coupons

    • View expired discounts/coupons

3

To resolve Return & Refund requests

    • Grab the Order History

    • Select the Order for Return/Refund

    • Reason for Return/Refund

    • Help with eligibility checks for the same

    • Assist with submitting/declining the request

Try using FreshBot yourself!

Defining the Chatbot’s Personality

  • Uses light, playful humor in speech to help users stay engaged and joyful.

  • Uses a natural, stress-free, and friendly tone.

  • Shares relevant insights on deals, policies, and products for users.

  • Helps deliver fast and clear responses, ensuring that users have a smooth experience.

Designing Conversations

We mapped user intents to natural utterances, allowing the chatbot to understand different ways people ask for the same thing and trigger the right flow instantly.

  • Designed for both free-text input and guided paths

  • Created response variations to avoid repetitive conversations

  • Balanced flexibility with clarity in every flow

The focus was to make interactions feel natural, but still purposeful.

Testing & Iteration

We tested the chatbot with users to understand how they naturally interact with it- what worked, what didn’t, and where the experience broke.

What we learned

  • The experience felt quick and easy to use

  • Conversational tone made interactions more engaging

  • The system struggled with off-script inputs, highlighting the need to account for and more flexibility

  • Clearer navigation and labeling were essential for smoother flows

  • Users needed more actionable, guided interactions and simpler, step-by-step information, along with the option to reach human support when needed

2

To Find Discounts / Coupons

Designing natural user utterances to trigger key flows by integrating an LLM to enhance the chatbot’s contextual understanding, specifically tailored for Fresh Direct and our key intents

Handling Uncertainty (Error Pathways)

We designed for moments when users go off-script—asking unexpected questions or requests outside the chatbot’s capabilities. Instead of breaking the experience, the system guides users back on track while still being helpful.

How it works:

  • Recognizes unsupported or unclear inputs

  • Provides helpful fallback responses (not generic errors)

  • Redirects users to the closest relevant action or page

  • Offers clear next steps or alternative options

The goal was to make the chatbot feel reliable and supportive, even when it doesn’t have a direct answer.

CUI Footprints

The chatbot is designed to sit seamlessly within FreshDirect’s app and website through a floating, collapsible interface, making it accessible without disrupting the browsing experience.

Enter Voiceflow - An artificial intelligence (AI) startup that offers a no-code platform for designing, prototyping, and deploying conversational AI experiences like chatbots and voice assistants.

Managing edge cases

Improved clarity in navigation

Mapping intents to conversation flows

🥬 Lettuce begin with the FreshBot Demo!

Hi 👋🏼, I'm FreshBot, your grocery sidekick - check out my fruitful capabilities!

Behind-the-Scenes

Exploring how Voiceflow’s backend commands power conversation flows from handling logic, intents, and responses to create smooth, responsive user interactions.

Key Learnings

  • This project pushed me to think beyond screens to design conversations that feel natural, responsive, and intentional.

  • It also made me realize how much conversational design is rooted in language and linguistics- understanding how people phrase the same intent in different ways, and designing systems that can interpret and respond meaningfully.

  • I learned how important it is to balance structure with unpredictability, especially since users don’t always follow the “expected” path, and conversations rarely happen in perfect flows.

I also wrote a blog about Writing for the “Screen”— What Bollywood Can Teach Us About Conversational UX

Read

Key Learnings