Enhancing Self-Service Support Through a Conversational AI Assistant
This project focused on creating an intuitive, brand-aligned AI experience that guides users through complex tasks and answers their questions in real time.
Role
Product Designer
Company
UserGuiding
Tools
Figma
Miro
Year
2024
Website
panel.userguiding.com
Project Overview
UserGuiding is a no-code platform that helps SaaS companies onboard and engage users through product tours, tooltips, and in-app guides. As the product grew, we saw a need to provide faster, more scalable user support. To address this, I worked on designing an AI assistant that could guide users, answer common questions, and help them navigate the product more independently.
Problem Statement
How might we support users throughout their journey by helping them quickly find the information or guidance they need, so they can confidently adopt and use the product without relying heavily on support teams?
This project has two sides: the admin panel experience and the end-user experience.
What should be taken into account when designing an AI assistant?
Designing an AI assistant requires balancing usability, trust, and functionality. Key considerations include clarity of responses, visual consistency, relevance of suggestions, accessibility, and seamless integration into the product experience.
Chat Experience & Usability
The chat interface should be intuitive, responsive, and user-friendly.
- • Timestamps
- • Chat history
- • Loading states
- • Rich responses (images, links, buttons)
- • Suggested prompts
Admin Dashboard & Control
The assistant should access accurate, relevant information from:
- • PDFs
- • Internal documents
- • Help articles
Admins should be able to:
- • Upload and organize content easily and test the source
Localization & Personalization
The assistant should support multiple languages and adapt to user context, such as their time zone, role, or region. Personalization enhances trust and makes responses feel more relevant and human.
Research
Wireframing
After completing the benchmark analysis, I created an information architecture to define the structure, screens, and key interactions for the AI assistant. This helped align user needs with business goals and set a clear foundation. I then moved into wireframing to explore layout ideas, prioritize usability, and quickly iterate. The wireframes served as a visual blueprint, aligning stakeholders and streamlining the transition to high-fidelity design. better understanding of the various market players.
User Testing
I tested my wireframes by conducting feedback sessions with three key stakeholders to evaluate the AI assistant’s usability and alignment with user needs.
Component Library & Design System
I created a modular design system to ensure visual and functional consistency across all user-facing experiences. It included reusable components, layout patterns, and interaction rules, helping us maintain a shared design language across the AI assistant and future projects. This approach also improved collaboration and made the design-to-development process more efficient.
Design
After completing the research and stakeholder interviews, I clustered all insights and translated them into refined, high-fidelity designs. The final version of the AI assistant reflects user needs, business goals, and a cohesive experience across both admin and end-user interfaces.
Results
Within just two weeks of implementation, we began to see real impact. Users engaged more confidently with the AI assistant, and internal teams reported fewer repetitive support requests, validating the assistant’s role in improving self-service and overall product adoption.
490
chat resolved for 241 unique users.
51%
resolution rate with AI compared to 46% across total conversations.
40-48 hours
Saved between 40-48 hours by resolving issues faster and smarter.