Alice Chatbot Case
- Role: Product Designer
- Timeline: 8 months
Designing Alice Chatbot to Streamline Front Office Operations
Context: ALICE is a comprehensive platform that improves operational efficiency and communication within the hospitality industry, contributing to better productivity and higher guest satisfaction. The front office department benefits from ALICE's chat solution, using it to communicate directly with guests.
Users: Front Desk Agents and Concierges
The first one are the guest's initial point of contact at the hotel and ensure a seamless stay by overseeing check-ins, check-outs, guest inquiries, and various guest services.
The second one are the hotel's experts for guest assistance and they specialize in arranging on-site and off-site activities, from dining to events and attractions, providing valuable recommendations and reservations.
Depending on the hotel's setup, both Concierges and Front Desk Agents may use Alice Messaging to efficiently address guests needs.
Hypothesis: Front Desk Agents and Concierges receive a large volume of basic questions via Chat, such as 'When is check-in?' and 'What is the wifi password?'. They spend at least one hour per day answering these questions and would like to be able to make better use of their time by dealing only with more complex conversations.
Quick peek at Alice for some examples of repetitive questions
Proving the hypothesis: In order to validate our hypothesis, we worked on some important iniciatives, like Data Analysis, Qualitative and Quantitative Research.
Data Analysis: We wrote a query to find out the percentage of guest messages containing predefined keywords related to common questions, such as 'check-out' and 'check-in. The results show that 26% of guest messages fell into this category.
Qualitative Research: During research trips to Austin and Las Vegas, we visited 5 hotels and conducted 12 in-depth user interviews with Concierges and Front Desk Agents. One of our findings was that some Front Office desks close late at night, potentially causing frustration for guests seeking assistance during those hours. A solution that we heard was to create an ‘Out of Office’ feature.
Quantitative Research: We also conducted preliminary quantitative research and usability test.
As a result, we were more confident in the project direction and that we would solve a real problem.
Iniciatives to prove hypothesis
Problem Definition: Based on our research, we created a user story to define the problem from the perspective of the end user. We used the format As a {User}, I want {Functionality}, so that {Value} because it helps us keep a user-centric approach and outlines the desired outcome.
As a Front Desk Agent, I want a messaging solution to handle FAQs and basic requests, so that I can save time and handle more complex conversations without being distracted by the low-value conversations.
Concept Exploration: That's when I began sketching ideas and creating a basic UX flow, along with potential UI components.
Basic UX Flow and UI Experiments
Challenges:
Toggle positioning: This element holds significant importance within the feature. Although initially considered placing it close to the text field, we decided against it to prevent any visual confusion for the user. Instead, it was positioned separately in the upper-right corner of the page, where it gained more prominence.
Design of Conversation Cards: We designed distinct UI cards for each stage of the flow. For example, when the chatbot successfully provides a response, the card displays only a bot icon with the label 'We're good' on hover.
Development of Escalations Logic: As the chatbot is trained to answer specific questions, we needed to design how the chatbot would respond when a guest asked non-basic questions. This led to the concept of Escalations. When the chatbot is unable to answer, the card background turns yellow, drawing the user's attention and prompting them to engage with the guest.
Implementation of Buffer Time Logic: Following an Escalation and a response manually sent by the user, it made sense to introduce a time period before the chatbot would be reactivated, allowing the user to maintain control until finishing a conversation. That's why we added the Buffer Time.
UI Elements: Toggle, Conversation Cards and Snack Bars
Concept Validation: Next, I finished the user flow, making sure all user interactions across various permission levels were covered, on both the backend and frontend. This gave to the team a clear visual representation of the user journey on the feature.
The final two initiatives before feature development were the UX Writing review and the a new usability test.
Our UX Writer worked to ensure that our writing was not only clear and consistent but also aligned with our tone of voice. This was essential because well-crafted messaging improves user understanding and engagement.
In our usability test, which got over 100 responses. Users could navigate through tasks and write down comments at the end. Most users completed tasks with ease. The results provided valuable insights from three distinct groups that we was able to categorize into
The largest group think the feature will simplify their day-to-day and relieve their workload. They are totally bought into the idea and don't question the chatbot capabilities.
The 'skeptical but optimistic' group of people who are open to trying the chatbot. However, they have concerns about the chatbot's ability to correctly answer to and fulfill guest requests.
The smallest group don't think the chatbot can be good enough and would rather keep direct communication with guests to deliver personalized guest experience.
Maze - Remote Usability Testing
Impact: Measuring the impact is not feasible as the feature is in Beta with Graduate Hotels. However, based on our data analysis and projections, we expect the following outcomes:
A 40% reduction in active guest messages, allowing hotel staff more time for complex tasks that requires their expertise. Front Desk Agents and Concierges will no longer need to handle common questions, as these are efficiently addressed by the chatbot.
Guests will experience a better experience, with the average conversation completion time reduced from 4 minutes to just 1 minute.
Team:
Product Designer: Fernando Chaves
Product Manager: Kenya Puig
UX Writing: Catherine Montesdeoca
Tech Lead: Fernando Silva
QA Engineer: Nupur Nangare
This case was internally presented to the entire product team, as a way to illustrate an example of our Design Process.