Sutter Plan
Appointment-Scheduling Chatbot Design

Disclaimer

This is a conceptual project created for portfolio purposes and is not affiliated with or endorsed by Sutter Health.

Overview

Sutter Plan - a user-friendly chatbot

Scheduling a doctor’s appointment through voice systems often leads to friction, especially for non-native speakers and users with speech impairments.

Sutter Plan is a conversational chatbot designed to improve accessibility, reduce task failure, and provide a structured alternative to voice-first interactions.

Research

Patients struggle with voice-based scheduling

I surveyed 37 participants to identify common pain points when booking appointments using voice technology.

Key findings:

• High failure rates in voice-only scheduling flows

• Pronunciation barriers for non-native speakers

• Speech-impaired users frequently unable to complete tasks

• Repetitive loops leading to frustration and escalation to human agents

Insight: Voice-first systems alone are not inclusive. Users need structured, flexible interaction options to successfully complete tasks.

Problem Statement

Patients fail to schedule appointments due to limitations of current voice interfaces

Existing Sutter Health VUI systems do not accommodate all patients. Users with different native languages or speech impairments often cannot follow the “happy path,” resulting in task failure and negative experiences. 

Proposed solution

Provide an inclusive, easy-to-use chatbot

To address patient frustration, I designed a simple chatbot that:

• Reduces ambiguity through structured options

• Supports multiple communication preferences

• Minimizes input errors and misunderstanding

Design Process

Designing the conversation flow

I began by mapping the user flow, incorporating research findings to create a minimum viable conversation.

Key design considerations:

• Clarity over flexibility: guided choices instead of open-ended input

• Error prevention: pre-defined responses and confirmations

• Accessibility: reduced reliance on speech or typing accuracy

The flow was prototyped in Voiceflow, allowing for live interaction testing.

FINAL DESIGN

“Your appointment has been scheduled.”

The Sutter Plan MVP allows users to book appointments efficiently and inclusively. The prototype demonstrates:

• Alternative to voice scheduling

• Pre-selected options for users with language or speech challenges

• Confirmation checkpoints to prevent errors

Evaluation & Success Metrics

To measure how well the chatbot performs and where improvements are needed, the following metrics would be tracked:

• Task completion rate: how often users successfully schedule an appointment

• User drop-off points: where users abandon the conversation

• Fallback / misunderstanding frequency: how often the bot fails to understand user input

• Time to complete scheduling: efficiency of the interaction

• Escalation to human support: instances where the bot cannot complete the task

Iteration Strategy

Conversational design is never static. The system is intended to evolve based on real usage data.

• Refining unclear prompts and responses – improving clarity and comprehension

• Reducing friction points in the flow – streamlining steps where users get stuck

• Improving guidance where users drop off or hesitate – adjusting prompts to guide completion

Summary

Outcome & Conceptual Impact

By applying these evaluation and iteration strategies, the chatbot aims to deliver:

• Increased accessibility for diverse user groups: supporting non-native speakers and speech-impaired users

• Reduced task failure compared to voice-only systems: more users successfully complete scheduling

• More consistent and scalable interaction model: easier to maintain and extend across different scenarios

What I Learned

This project reinforced important lessons in conversational design:

• Structured flows are as important as natural language for success

• Guided interactions and robust error handling significantly improve outcomes

• Continuous evaluation and iteration are critical for scalable AI systems