Empowering Conversational AI Designers to create culturally adaptive AI to improve accessibility and increase engagement
MY ROLE
Interaction Design, User Experience Design, Visual Design, Dev Hand-Off
TEAM
1 Product Manager
1 Front-end Engineer
1 Back-end Engineers
TIMELINE
2 weeks
PROBLEM
Inaccurate chatbot content translations caused by semantic gaps in the Google Translate API increased bounce rates, customer support calls, and operational costs, negatively affecting user satisfaction and business profitability.
THE SOLUTION
Provide a way to check the accuracy of translated content and edit translations for multilingual chatbots before going live, eventually avoiding errors and enhancing usability
WHAT IS ENGAGELY.AI ↗ ?
A B2B2C enterprise-grade conversational AI platform for creating multilingual chatbots, voice bots, and social media bots. It serves companies globally, enhancing customer engagement through automated AI interactions
CONTEXT
The need for culturally adaptive conversational AI
The platform initially supported only English, but its adoption across Asia highlighted the need for multilingual support, covering languages like Mandarin, Arabic, and Indian dialects.
RESEARCH: IDENTIFYIING WHY BEHIND THE PROBLEM
Interviews with conversational AI designers and understanding their needs
The interviews revealed that our current users expect to have a workflow developed where:
Check the accuracy of individually translated content from Google API without needing to test the whole bot journey.
Edit translated content incase of inaccuracy.
Ability to Assign the translation to a Content Translator hired by their company who would be translating each individually inaccurate content manually.
CHALLENGES AND SOLUTIONS
🤕 How do we display the large volume of content, including multimedia, without making it overwhelming to use?
✅ Modular design for content translations
Separated chatbot content into logical sections (failure messages, forms, user journeys).
🤕 How do we translate multimedia?
✅ Enabled multimedia content uploads in regional languages.
User can view all the multimedia in the same place and can upload the corresponding translation in the adjacent column
🤕 How do we allow checking and editing an automatic translation?
✅ Displayed original English content alongside translated versions for editing.
Assign translation to an internal team member (content writer, conversation designer, language specialist)
Relevant filter to quick search for content
USABILITY TESTING
5 conversation AI designers of clients
SOLUTION
✅ Simplified workflows by introducing a new filter 'Since your last visit'.
This helps the user to instantly check translations of any new added content since their last visit without going through the entire list of contents
✅ Refined media translation and replacement flow to have better affordances
IMPACT
The multilingual solution was a one-of-its kind in the conversational AI enterprise space and is used across industries in the Asia Pacific region
70%
decrease in bounce rate for a non-English end customer
45%
increase in multilingual engagements