Aishwarya Iyengar
Service Designer, User Researcher & Design Educator
CHATBOT UX STANDARDIZATION
The Challenge
Chatbots are not the most preferred channel to get support, given easy dead-ends and non-organic approach, defying natural human conversation behaviour. In a multinational company there are hundreds of chatbots that deliver an experience that is heavily siloed, and not unified in the way it communicates with the user, in the type of information it delivers or the way it delivers it! This is an added friction in the user's journey who wants to get "Fast support".
Our Client from the Centre of Excellence team of chatbots, approached us to redesign the Chatbot UX that would then become the standard for all chatbot teams to comply with.
How might we ensure that users of chatbots have a unified and seamless experience while interacting with any of the Company's chatbots?
The Approach - Lean UX
Our client insisted that he is seeking an agile approach, where we would provide design recommendations in sprints of 3 weeks, and the development and data science team could work out the logic behind chatbot functions!
So Voila! The Design team decided that we would not compromise on research, but 'Optimize Research' to direct conversational UX design. So we followed the principles of Lean UX, and included chatbot teams and users as and when we iterated.
think
make
test
1. Primary Research
2. Secondary Research
3. Personality principles
4. Brand image
1. Design principles
2. Personality guide
3. Conversational flow
4. Conversational guide
5. Small Talk package
6. Small talk guideline
7. Trail of usage
8. Demo video
1. Co-design workshop
2. Individual User testing
Top Findings
We categorized the findings in different aspects of the Chatbot Experience
Access | Bots are not strategically placed for users' reach in time of need, defeating the purpose of bot- to save time and effort in mundane tasks
Profiling | Bots do not seem to recognize patterns even within the same conversation, hence they end up annoying the user asking the same details over and over, rather than building on latest knowledge.
Visual Indication | Bots do not give an indication of the task and do not keep the user engaged while performing a task, especially when the task to be completed in expected to take a long time.
Overloading response | Bots often give long answers or multiple answers for a single query, thus making it difficult for users to comprehend the solution.
Threshold | Bots do not identify human-threshold for non-understandable intents/ unknown tasks and thus keep hitting dead ends in a conversation, without knowing when to hand-over to human support.
The guidelines
Bot Pitfalls
& Principles
Identifying shortcoming of existing chatbots and set design principles that will guide chatbot teams
Conversational guide
Translate the defined design principles into conversational flows.
Personality guide
Go beyond functional traits and define behavioral traits to create consistency in conversational experience.
Small Talk
guide
To identify contextual application of small talk with various use-case examples.
We designed a set of guidelines that were based out of pitfalls identified in current chatbots and user experiences, combined with design principles for both functional and behavioral traits of chatbots.
Each guideline is crafted based on use-cases from various 'relatable workplace functions', that will guide the reader to understand flaws in current chat flows and thereby how specifically to improve them in terms of content, language and information architecture.
*As they have been designed for a multinational company, these cannot be shared publicly, If you are interested to know more, excerpts of the content can be shown in a private conversation!
The co-design workshop
We invited 4 chatbot teams from Finance & IT functions to the co-design workshop and created scenarios that could be tested by role-play. Each team was given 2 scenarios and chatbot conversational flows, and they reviewed various aspects including 'intent recognition', 'flow', 'tone of language', 'dead-ends', 'human-handover' & 'small talk'.
NEW IDEA!
The enthused team not only gave us excellent feedback on conversational flows, but also chirped that they missed seeing simple UX checklist for project teams that are just starting out chatbot conceptualization.. This was our AHA moment!
UX check tool
A check-list for Chatbot teams that are starting out or they are in the middle of a chatbot development cycle and need guidance with scope & maturity levels of chatbots.
The impact
After the UX Library of Standards was launched in a chatbot conference, they were adopted by several chatbot teams to implement the Conversational UX guidelines.
This project helped us crack a chatbot design project with a new customer. This project cannot be published, but can be discussed in a private conversation.