AI Assistant + Feature Guide
A unified in-product guidance system for legacy enterprise software.
Designed a scalable In-Product System, addressing steep learning curves and feature adoption through both AI-assisted discovery and guided feature navigation.
My Roles: UX Researcher, UX Designer, UI Designer
Softwares Used: Figma (Design, presentation and research)
‼️ This project is covered under NDA. ‼️
Visuals and product names have been anonymised, and data has been modified to preserve confidentiality.
01
Context
-
Legacy enterprise mainframe products with decades old workflows and steep learning curves
-
Multiple products built on different technologies with inconsistent UI patterns
-
Heavy reliance on external documentation and support for task completion
-
Goal to improve usability and feature adoption through in-product guidance
03
Why existing system failed?
-
Static help documentation, disconnected from real user workflows
-
Users had to leave the application to learn how to perform tasks
-
One-off UI cues did not scale across products or evolving features
-
Lack of contextual guidance increased dependency on support teams
02
Constraints
-
No shared design system or common interaction baseline across products
-
High risk actions requiring accuracy, traceability, and user verification
-
Solutions needed to be reusable and scalable across multiple products
04
Core Idea
-
Shift learning from external documentation to contextual, in-product guidance
-
Introduced two complementary solutions: Feature navigation system and AI assisted discovery
-
Design both solutions as plug-and-play frameworks usable across products
-
Balancing flexibility with safety through human-in-the-loop (HITL) verification for AI assistance
Problem Exploration
Observation
After working on these legacy enterprise products for over a year, I firsthand experienced the steep learning curve caused by the absence of in-product guidance.
Assumptions
Long-time users were already deeply familiar with the tools. However, conversations with managers revealed increasing challenges for new users and employees adopting these products.
Investigation
Discussions with Support and Marketing teams showed that many usability-related questions were handled through support tickets which were effective for bugs and technical glitches, but inefficient for feature understanding.
Problem statement
This led to the insight that traditional onboarding was insufficient for long-standing products, creating the need for contextual, on-demand feature guidance.
Solutions
Since a single solution could not address all learning needs. Users required structured, step-by-step guidance for repeatable tasks, while also a flexible discovery support when they didn’t know what to do or where to start.
Solution 1:- Feature Guide (Navigation)
Check the list and play recording
What?
Feature Navigator is a deterministic, in-product walkthrough system that guides users step by step through specific features and workflows using pre-recorded instructions within the application.
Why?
- Many users knew what they wanted to do but struggled with how to do it due to the absence of in-product guidance.
- Traditional onboarding was ineffective for mature products with numerous features and returning users.
- A guided, on-demand approach allowed users to learn only what they needed, when they needed it.
How?
- Users select a feature they want to learn.
- The walkthrough guides them through each step directly within the product interface.
- Guidance is contextual and tied to real UI elements, eliminating the need to switch to external documentation.
Design intent
Feature Navigator was designed for clarity, predictability, and safety, making it ideal for repeatable workflows and feature adoption without introducing automation risk.
Tool tips guiding the user about the feature's potential and how to use it.
Designed in way that components can be reused in other products with minimal UI changes.
User flow diagram for Feature Guide
Solution 2:- AI Assistant (Discovery & Task Support)
Plug-and-play design framework, usable across products with minimal UI change requirement
What?
The AI Assistant is a conversational support system designed to help users discover features, understand workflows, and perform tasks when they are unsure where to start.
Why?
- Some users didn’t know which feature or workflow to use to achieve their goal.
- Static guidance and walkthroughs could not address exploratory or open-ended questions.
- There was also a growing need to align with market trends and leadership direction toward AI-enabled products.
How?
- Users interact with the assistant through natural language.
- The assistant provides contextual guidance and suggests next steps or actions.
- For task execution, the assistant follows a human-in-the-loop model, requiring user confirmation at each step.
Design intent
The AI Assistant was intentionally constrained to assist and guide, not act autonomously, ensuring trust and safety in a high-risk enterprise environment.
Although both solutions address the same core problem i.e. steep learning curves and feature adoption, they were intentionally designed as separate features to respect different user mental models.
-
Feature Guide supports users who want structured, step-by-step certainty for known tasks.
-
AI Assistant supports users who need discovery, clarification, or help deciding what to do next.
Keeping them separate avoids cognitive overload and allows users to choose the type of help that best fits their situation.
Together, they form a scalable in-product guidance system that balances predictability and flexibility without forcing AI into every interaction.
Reflections
-
Gained deeper understanding of designing AI boundaries and user control, in high-risk enterprise workflows.
-
Recognised the importance of designing for scalability early when working across inconsistent systems.
.jpg)
