BMW Intelligent Personal Assistant (IPA) - Chitchat Enhancement
Role
Lead Conversation Designer
Multi-modal UX Designer
Deliverables
Conversation Design
Competitive analysis
User research
Usability tests and findings
IPA Persona
Sample dialogues
Project Specifications
Duration
12 WEEKS
Tools
Figma
Excel
NOTE: Many images are blurred and cannot be expanded for confidentiality purposes.
Overview
In a pioneering effort, I led the team to enhance BMW’s Intelligent Personal Assistant (IPA) for the Chinese market, focusing on cultural adaptation and conversation diversity. Over 12 weeks, as the Lead Conversation Designer, I spearheaded the development of 300 new and 100 enhanced conversations, making the IPA more engaging and culturally relevant.
Challenges & Constraints
Time Constraints: Rapid development of a vast number of conversations within a limited timeframe.
Technical Limitations: Achieving natural voice prosody and intonation was restricted by project deadlines.
Persona Development: The absence of a pre-existing Chinese IPA persona required quick, research-backed creation to inform design.
Results
The project significantly improved the IPA's relevance and usability for the Chinese market, as evidenced by positive feedback from the headquarter.
Enhanced User Engagement and Satisfaction
Recognition and Adoption by Global Teams
Increased Investment in Conversational AI
Background
BMW's IPA plays a crucial role in enhancing the driving experience for new BMW car models. Our project aimed to revitalize the IPA's Chitchat feature, making it more relevant and appealing to the Chinese market by expanding its conversational capabilities and cultural appropriateness.
The old version of Chitchat (starts from 1.01 min in the video on the right) could only support very limited questions, whose answers that did not have variations.
Problems
The ‘global’ IPA was a German character, which didn’t fit other countries that well, especially Asian countries. The conversations, which were designed according to that ‘global’ IPA, caused confusion in China and led to the low frequency of usage of the IPA in China.
Some of the conversations could be applicable for Chinese users, but didn’t get translated in the correct context which brought about confusion to users.
Goals
Make Chitchat conversations culturally appropriate for the Chinese market.
Broaden the scope of Chitchat to cover more scenarios.
Increase the usage frequency of the IPA.
Advance the IPA towards becoming a fully-fledged AI assistant, displaying more intelligence and human-like interaction capabilities.
Research
During the research phase, I looked at the following pieces of research in addition to conducting my own. These pieces of research helped to ground my design.
Research Insights
Through competitive analysis, customer studies, and exploration into Chinese user preferences, we uncovered key insights:
Chinese BMW owners are younger and more tech-savvy compared to global averages.
Preferences for a conversational assistant that balances humor with emotional connection, avoiding a machine-like demeanor.
These findings informed our ideation, focusing on creating an IPA character that embodies BMW’s brand values while resonating with the Chinese audience.
Summaries Of Research Findings
Persona Development: A Foundation for Relevance
With a clear understanding of our audience, we developed a persona for the IPA that exemplified the qualities most valued by Chinese users: intelligence, reliability, and a touch of wit. This persona guided the creation of dialogues, ensuring consistency in tone, style, and character throughout the interactions.
Legal Guidelines
I was able to get touch with the legal team and studied the guidelines the legal specialist provided. The guidelines laid out the do’s & don’ts, and the safety area for designers to design the conversations in.
I managed to use the results of the research/studies to design Chitchat conversations.
Initial Chitchat Content design
First, I created 30 new conversations and refined 10 conversations based on the IPA values & the IPA persona. The IPA’s responses to some questions were designed to change randomly, or change depending upon the logic conditions.
The content development phase was marked by a highly collaborative approach, involving cross-functional teams from design, technology, legal, and linguistic backgrounds. This collaboration ensured that every piece of content was not only creatively engaging but also technically sound, legally compliant, and linguistically accurate. Regular workshops and review sessions facilitated a dynamic exchange of ideas and feedback, driving the continuous improvement of the content.
Examples Of The Conversations I Created
Iterative Design and User Testing
I conducted an internal test on 40 people in the office for our initial design iteration and received some feedback.
The positive feedback validated the values of the IPA and the IPA Persona for the Chinese Market. This was very meaningful and helpful to the rest of my work, while also contributing to the team.
I refined the total of 40 conversations with my team, from the initial design iteration based on the feedback I received. Next, I then created 270 more new conversations for a total of 300 while refining the rest of the old Chitchat conversations. When I was designing the voice conversations, I reviewed the dialogue designs with different stakeholders across different countries to make sure that there were no problem with understanding, acceptance, consistency, feasibility, copy rights and other important matters.
The Prototype For Testing & Feedback
validation testing
Besides gathering people in the office to test them on each iteration of the design, I led the designers to conduct UX validation testing and worked with speech specialists to conduct speech tuning. (In the Chinese language, if the emphasis or pause in one sentence is in different positions, it could have totally different meanings. Since the prompt is generated from text, this issue happens quite often.) Based on the feedback I received, I did the final refinements.
Deliverables
In the end, my team and I delivered 300 new conversations with up to 6 answers for each question, and 100 enhanced conversations for the Chitchat feature.
Feedback from users
“Discovering this level of connection with the IPA was unexpected. It goes beyond following commands – it’s as if it gets my mood and what I like, making every drive better than the last.”
“This is the first time I’ve ever felt so in tune with technology. The IPA isn’t just taking in what I say; it genuinely understands me, offering a kind of friendship that’s both surprising and soothing.”
Results
The project significantly improved the IPA's relevance and usability for the Chinese market, as evidenced by positive feedback from the headquarter.
Enhanced User Engagement and Satisfaction
The most immediate and measurable result of our project was a marked increase in user engagement with the IPA. Users reported a significantly improved experience, highlighting the IPA's enhanced conversational abilities, cultural relevance, and intuitive interactions. This positive feedback was quantified in our post-launch metrics, which showed a substantial uptick in the frequency of interactions with the IPA, demonstrating its value addition to the overall BMW driving experience for Chinese customers.
Recognition and Adoption by Global Teams
A testament to the quality and innovation of our work, the redesigned IPA received high praise from BMW's global assistant team. Elements of our conversational design were recognized for their creativity and effectiveness, leading to their adoption into the global IPA solution. This not only validated our approach but also set a new standard for how localized content can enrich global products, fostering a more inclusive and versatile conversational AI platform.
Increased Investment in Conversational AI
The project's success catalyzed further investment in the development and enhancement of the IPA's capabilities. Leadership's commitment to expanding the IPA's functionality was a direct result of the demonstrated impact of our work, paving the way for future innovations in conversational AI. This additional funding and support underscore the strategic importance of our project, positioning the IPA for continuous improvement and evolution.
Things I would do differently
Reflecting on the project's journey and its outcomes, there are always insights to be gained and lessons learned that can inform future initiatives. In the spirit of continuous improvement and innovation, here are some aspects I would approach differently in hindsight:
Greater Emphasis on Multimodal Interactions
Focusing predominantly on voice interactions, we may have overlooked opportunities to enhance the user experience through multimodal interactions, combining voice with visual or tactile inputs. In future projects, exploring and integrating these additional dimensions could lead to a more immersive and intuitive user experience, especially in the context of an in-car assistant.
Leveraging Advanced AI and Machine Learning Techniques
The project successfully implemented a wide range of conversational enhancements. However, incorporating more advanced AI and machine learning techniques could offer dynamic personalization and learning capabilities, enabling the IPA to adapt more fluidly to individual user preferences and contexts over time. Investing in these technologies could significantly elevate the intelligence and user-centricity of the IPA.
Metrics
Due to limitations with the local speech solutions, we couldn't obtain direct metrics. However, had we had access, the primary metric of focus would be Quality of Conversational Experience. This metric is crucial as it directly influences user engagement by measuring how natural and relevant the interactions are, reflecting the IPA's effectiveness in delivering a seamless and culturally resonant user experience.