Final Designs
Designing a unified AI chat pattern and centralised component library in JDS — enabling consistent, proactive conversational experiences across iOS, Android, and Web for all Jio and RIL businesses.
"To revolutionise the Jio Chat platform, our design vision is to create a centralised repository of multi-domain business-specific components in the JDS for native apps & Web chat, which isn't currently supported well by Haptik. It will enable consistent experiences across Jio & RIL businesses, reducing learnability for our customers."
Jio's customer support was fragmented across separate Haptik-powered chatbots for each business unit — no shared components, no consistency, no scalable JDS backing. Only 32% of users reaching the support section were actually engaging with chat.
1.25 lakh users hit the support section monthly. Introducing quick actions, cross-sell flows, and a JDS-backed component library would convert passive support traffic into active engagement — and deflect costly agent calls.
The chatbot is a self-service response system — in many cases the first line of support for Jio customers through digital channels. It's designed to quickly address incoming inquiries through human-like conversation without needing an agent's intervention.
For Jio, it handles common, predictable queries representing a significant share of total inbound volume. While individually straightforward, these queries collectively consume enormous time and resources when handled manually. The problem: Jio's chat was fragmented, inconsistent across verticals, and deeply underused — with next-generation AI-driven chatbots now offering context-specific experiences and elevated personalisation, the gap was widening.
Data from past 6 months revealed a sharp drop-off funnel. Only 32% of users reaching the support section were engaging with chat.
20% churn on accessing the Product Section — Discovery is the issue. Opportunity loss from missing quick actions. Solution: Introduce Quick Actions and cross-sell for better discoverability.
Desk work conducted with the business to establish second-hand data and validate the purpose before moving to primary research. The Chat component is used to display real-time and past chat logs between service agents and customers, with sub-components covering different chat item types — across B2B messaging, live chat for Retail, EdTech, and other Jio businesses.
Decrease in chatbot usage in the Jio app over 6 months — and a 12% decrease in Product Support Engagement, signalling a UX and discovery problem.
Users reach the support app monthly. 83% visit chat, 63% try using chat support, but only 40% are actively using it — a steep engagement cliff.
Majority of users had not used Jio chatbots. Most-used chat within Jio businesses: MyJio and Ajio. JioMart (18), JioCinema (3), Ajio (6), MyJio (45), JioMeet (6), JioSaavn (13) users surveyed.
After secondary research, primary research was conducted — in-person and telephone interviews as part of qualitative research — working closely with target users to uncover hidden information about their Jio chat usage (web + mobile), issues, and flaws.
Understand user pain points and emotions when using the Jio Chat Application.
Understand mental models, needs, and frustrations of users from ground zero.
Access behaviour and beliefs to produce data on user preferences across the platform.
Responses from diverse geographical, educational, and professional backgrounds — capturing usage patterns, challenges, and chatbot preferences.
6 Platforms were selected for competitive analysis as we compare features, strengths, weakness relative to Jio as an organisation.
|
Websites/Tools →
Features ↓
|
Haptik | Freshchat | Velocity | Intercom | Zendesk | ChatBot |
|---|---|---|---|---|---|---|
|
Templates
Chatbot Stories that let you launch task-specific chatbots in just a few clicks
|
✅ | ✅ | ❌ | ✅ | ✅ | ✅ |
|
Customisation
Built-in ready-to-use, personalised message, with or without human interface, no code solution, modify response time
|
✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
|
Omni-Channel Support
Stitch conversations from all channels like WhatsApp, Facebook, Slack, Instagram & Telegram etc. and bring them under 1 roof
|
✅ | ✅ | ❌ | ✅ | ✅ | ✅ |
|
Interactive Flow Builder
Compile & structure questions and their subsequent answers or various replies to certain customer queries
|
✅ | ✅ | ❌ | ✅ | ✅ | ✅ |
Multi-Language Capabilities |
✅ | ✅ | ❌ | ✅ | ✅ | ✅ |
|
Emotional Intelligence
Reads whether the customer is angry, confused, or happy. Underlying emotions & intent of customers, responds appropriately
|
✅ | ✅ | ❌ | ❌ | ✅ | ✅ |
Type of Chat System |
Menu Base + AI | Menu Base + Conversational | ChatGPT-powered AI chatbot | Menu Base + Keyword | Menu Base + Keyword | Menu Base + Conversational |
Demo |
✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
Target Industries |
FinTech E-Commerce Education Healthcare |
FinTech E-Commerce Education Healthcare |
E-Commerce | FinTech E-Commerce Education Healthcare |
FinTech E-Commerce Education Healthcare |
FinTech E-Commerce Education Healthcare |
The main purpose of the heuristic evaluation is to determine how current systems are working and establish usability principles like Nielsen's heuristics or others specific to the project.
Apps evaluated — JioMart, JioCare, Ajio, JioCinema & JioMeet
A high-level overview of the key findings from the evaluation — areas of the interface that performed well and areas that need improvement. Each finding is linked to a specific heuristic or usability principle.
| SR | VIOLATION | PRINCIPLE VIOLATED | RECOMMENDATION | INSTANCE | SEVERITY |
|---|---|---|---|---|---|
| 1 | The system provides a read status but it is very small to observe. The system also replies instantly without wasting a second. |
Visibility of system status
Keep users informed about what's going on, through appropriate feedback within time.
|
Tick marks can be increased in size. For the chatbot to be not too robotic, we can have a "..typing" before every message. | JioCare Chatbot | 2 |
| 2 | The system asks for authentication then says type "Hello" to proceed. This differs from usual behaviour where a user is shown app value after authentication. |
Match between system and the world
Follow real-world conventions, making information appear in a natural and logical way.
|
Users can have the menu directly after entering OTP — that is the expected behaviour after successful authentication. | JioCare Chatbot | 2 |
| 3 | Quick buttons disappeared by themselves when no click action was taken. The user might have to repeat all steps to reach the same point again. |
User control and freedom
Users should leave the unwanted state without having to go through an extended dialogue. Undo and redo.
|
Buttons should remain so the user picks up where they left off without having to retype the same query. | JioCare Chatbot | 3 |
| 4 | There is no way a user can change their number in case of a wrongly entered number, or when using two Jio SIMs. |
Error prevention
Eliminate error-prone conditions or present users with a confirmation option before they commit to the action.
|
If history is stored and the user creates another session, a way to change number should always be provided. | JioCare Chatbot | 3 |
| 5 | After completing a number of steps, there is an error from which the system doesn't help the user recover. |
Help users recognise, diagnose, and recover from errors
Error messages must be clear, precise, and offer solutions.
|
Fallout scenarios must be mapped out well, reducing the steps in each direction. | JioMart | 3 |
| 6 | No usability issue here. The system provides the user with options in every step where it understands the user's query and intent. |
Recognition rather than recall
Minimise the user's memory load by making elements, actions, and options visible.
|
NA | JioCinema, JioMeet, Ajio & JioMart | 0 |
| 7 | The system is consistent throughout in the way the UI is designed, the choice of words and their meanings. |
Consistency and standards
Users should not be left guessing about the meaning of words, situations, or actions.
|
NA | JioCinema, JioMeet, Ajio & JioMart | 0 |
| 8 | Users are not readers. A user wouldn't know more options exist if they fail to see the hamburger menu or know what a text-box is. They may be clueless if they don't understand English. |
Help and documentation
Users should not be left guessing about the meaning of words, situations, or actions.
|
Show users all options upfront, or show a button with "All Menu" written. Show the hamburger menu exists via micro-animation. Before authentication, show the language selection option. | JioCare Chatbot | 2 |
The problem is not awareness — users understand chatbots and want to use them. The problem is trust, discoverability, and perceived usefulness. Users abandoned because every path looped back to unhelpful responses. The opportunity: quick actions, personalisation, and a consistent JDS-backed component system across all Jio verticals.
Based on interview responses, 5 user personas were designed for the chat platform. Insights for user needs, pain points, and general behaviour were greatly aided by personas — allowing more effective and targeted solution design.
Separate user flows were mapped for the two primary verticals — Retail and EdTech — reflecting distinct intents, task types, and escalation paths in each domain.
Constructing HMW questions generated creative solutions while keeping the team focused on the right problems to solve.
Various themes were defined for the chatbot pattern strategy — exploring how the chatbot can serve each persona across all Jio verticals.
Personalised recommendations based on preferences, chat history, likes & dislikes. Immediate suggestions on products or better versions of selected products.
Offering different products based on what customers are interested in or have already bought. Suggestions on pre & post purchases to increase basket value.
Payments, cart history, apply & redeem points, schedule appointments, live status on complaints, recommendations, agent connect — all within one interface.
Seamlessly transferring to a human agent based on user intention and query complexity. Agent receives full conversation context — no dead ends, no abrupt cuts.
Knowing the user's buying intent and offering product reviews, discounts, and brand comparisons. Feedback loops build continuous improvement in conversational data.
Domain-specific flows for Retail — mapping the taxonomy of intent categories to clear resolution paths, with graceful fallbacks at every decision node.
"To revolutionise the Jio Chat platform — create a centralised repository of multi-domain business-specific components in the JDS for native apps & Web chat, which isn't currently supported well by Haptik. It will enable consistent experiences across Jio & RIL businesses, reducing learnability for our customers."
The conversation taxonomy was mapped across the top intent categories. Each branch required clear intent detection, a defined resolution path, and a graceful fallback to agent escalation.
Quick lookup + live status + delivery updates
Payment flows, refund initiation, points redemption
Plan info, quick-select, deep link to payment
Auto-diagnostics + ticket raise + network support
Post-interaction survey, complaint logging, escalation
Context-aware agent transfer when confidence drops
Together with Monotype, Jio created its corporate typeface — JioType. The forms are largely geometrical but feature friendly, open curves — designed to work across analog and digital environments. For us it was important that the typeface works in both contexts.
The colour system was inspired by India's vibrancy — from the explosion of Gulal at Holi to vibrant herbs & spices. Colour conveys meaning, directs attention, creates mood, and brings digital experiences to life.
Final UI covered two primary verticals. Each screen was designed to JDS spec and built to be configurable as a no-code embeddable widget across Jio business properties.
A screen responsible for showcasing multiple tasks achievable within the chatbot — categorised into data insights, visualisation, and personalised custom chat messages. Users can track orders, view cart history, apply points, and connect with agents without leaving chat.
Users can meticulously create new mock tests, access results and areas of improvement within chat, and connect with Tutors/Mentors directly. Performance data (e.g. "21% decrease in percentile — impact of wrong answer") is surfaced conversationally for targeted improvement.
Moderated usability testing across retail and EdTech archetypes — 2 rounds of iteration between sessions with 8 participants each.
We underestimated the edge-case complexity in the fallback-to-human handoff flow. Conversational dead ends found in QA could have been caught earlier with more diverse user testing. Next time: stress-test failure states as rigorously as the happy path — and involve real support agents in QA from the start.
Post-launch metrics measured across MyJio and Jio Retail over the first 90 days in production.
