AI-Gen HelpAssist
Designed an AI-powered support tool that helped service desk agents resolve multilingual tickets faster, automating translation, response generation, and summarization to reduce agent workload and improve response accuracy.
Built and validated during my time as a UX Designer at Accenture, India.
AI-Gen HelpAssist
Designed an AI-powered support tool that helped service desk agents resolve multilingual tickets faster, automating translation, response generation, and summarization to reduce agent workload and improve response accuracy.
Built and validated during my time as a UX Designer at Accenture, India.
Website Design
2 Months
Accenture Project
Team of 2
Website Design
2 Months
Accenture Project
Team of 2
Context:
Context:
This was a project I did when I was working as a UX Designer at Accenture, India
This was a project I did when I was working as a UX Designer at Accenture, India
My Role:
My Role:
UX Designer — end-to-end design, from user research and persona development through wireframing, prototyping, and stakeholder presentation.
UX Designer — end-to-end design, from user research and persona development through wireframing, prototyping, and stakeholder presentation.
Challenge:
Challenge:
Managing customer queries was slow and frustrating due to disconnected tools and high volumes.
Managing customer queries was slow and frustrating due to disconnected tools and high volumes.
Project Goal:
Project Goal:
To design a seamless and intuitive AI-driven customer support experience that simplifies the task performed by support desk personnel.
To design a seamless and intuitive AI-driven customer support experience that simplifies the task performed by support desk personnel.
Methods:
Methods:
User Research, User Persona, User Flow, Wire-Framing, Prototyping
User Research, User Persona, User Flow, Wire-Framing, Prototyping
Tools:
Tools:
Figma | Microsoft Teams
Figma | Microsoft Teams
tl;dr
tl;dr
Support desk agents were spending the majority of their time on repetitive tasks — translating tickets, drafting responses, and summarizing requests manually. We designed an AI-powered tool that automated all three, cutting resolution time and freeing agents to focus on complex, high-value issues.
The result: faster responses, fewer errors, and agents who felt empowered rather than overwhelmed.
Support desk agents were spending the majority of their time on repetitive tasks — translating tickets, drafting responses, and summarizing requests manually. We designed an AI-powered tool that automated all three, cutting resolution time and freeing agents to focus on complex, high-value issues.
The result: faster responses, fewer errors, and agents who felt empowered rather than overwhelmed.
The Challenge
The Challenge
Support desk agents at enterprise scale were handling hundreds of tickets daily across multiple languages manually translating queries, drafting responses from scratch, and switching between disconnected tools to find previous case history. The cognitive load was unsustainable, error rates were high, and agent burnout was a real operational risk.
The challenge was to design a unified, AI-assisted interface that reduced repetitive work without removing human judgment from the equation.
Support desk agents at enterprise scale were handling hundreds of tickets daily across multiple languages manually translating queries, drafting responses from scratch, and switching between disconnected tools to find previous case history. The cognitive load was unsustainable, error rates were high, and agent burnout was a real operational risk.
The challenge was to design a unified, AI-assisted interface that reduced repetitive work without removing human judgment from the equation.
Skip to Solution
Key Features
Key Features
Seamless and accurate translation across multiple languages.
Seamless and accurate translation across multiple languages.
Generate clear, professional email responses effortlessly.
Generate clear, professional email responses effortlessly.
Summarize email requests with actionable insights and recommendations.
Summarize email requests with actionable insights and recommendations.
User Research
User Research
We conducted contextual interviews with 6 service desk agents across two business units to understand daily workflows, pain points, and workarounds.
This was supplemented by secondary research into enterprise AI adoption patterns and support desk efficiency benchmarks.
Key Insights from the research include:
Insight 1: Agents spent up to 40% of their time on translation alone this was the highest-friction, lowest-value task and the clearest opportunity for automation
Insight 2: Agents distrusted fully automated responses, they wanted AI suggestions they could review and edit, not responses sent without human approval
Insight 3: Multilingual ticket misrouting was a leading cause of delayed resolutions, better summarization at intake could prevent this entirely
We conducted contextual interviews with 6 service desk agents across two business units to understand daily workflows, pain points, and workarounds.
This was supplemented by secondary research into enterprise AI adoption patterns and support desk efficiency benchmarks.
Key Insights from the research include:
Insight 1: Agents spent up to 40% of their time on translation alone this was the highest-friction, lowest-value task and the clearest opportunity for automation
Insight 2: Agents distrusted fully automated responses, they wanted AI suggestions they could review and edit, not responses sent without human approval
Insight 3: Multilingual ticket misrouting was a leading cause of delayed resolutions, better summarization at intake could prevent this entirely
Brainstorming
Brainstorming
Brainstorming is my favorite part of the design process, it’s where ideas flow freely, creativity sparks solutions, and possibilities start taking shape. This is where we did a popcorn session and added all our thoughts on paper.
Brainstorming is my favorite part of the design process, it’s where ideas flow freely, creativity sparks solutions, and possibilities start taking shape. This is where we did a popcorn session and added all our thoughts on paper.
From the session, three core design principles emerged that guided every subsequent decision:
AI suggests, humans approve — no fully automated actions without agent sign-off.
Speed over comprehensiveness — surface the most relevant information first, not everything.
Language should never be a barrier — translation must be invisible and instantaneous.
From the session, three core design principles emerged that guided every subsequent decision:
AI suggests, humans approve — no fully automated actions without agent sign-off.
Speed over comprehensiveness — surface the most relevant information first, not everything.
Language should never be a barrier — translation must be invisible and instantaneous.
User Persona
User Persona
To truly understand the users' needs and challenges, we created a detailed user persona that would serve as the foundation for the design solutions throughout the project.
To truly understand the users' needs and challenges, we created a detailed user persona that would serve as the foundation for the design solutions throughout the project.
Meet Ashley, a Service Desk Agent overwhelmed by ticket overload, striving to quickly identify priorities, provide accurate translations, and resolve customer grievances efficiently.
Ashley's frustration with context-switching directly informed our decision to build a single-screen dashboard for surfacing ticket translation, response suggestion, and case history in one unified view rather than across three separate tools.
Meet Ashley, a Service Desk Agent overwhelmed by ticket overload, striving to quickly identify priorities, provide accurate translations, and resolve customer grievances efficiently.
Ashley's frustration with context-switching directly informed our decision to build a single-screen dashboard for surfacing ticket translation, response suggestion, and case history in one unified view rather than across three separate tools.
User Flow
User Flow
The user flow mapped the agent's journey from receiving a new ticket through to resolution, revealing that the current process required 7 tool switches and 4 manual copy-paste actions. Our redesigned flow reduced this to a single continuous interface with zero context switching.
Detailed user flow below.
The user flow mapped the agent's journey from receiving a new ticket through to resolution, revealing that the current process required 7 tool switches and 4 manual copy-paste actions. Our redesigned flow reduced this to a single continuous interface with zero context switching.
Detailed user flow below.
Inspiration Board
Inspiration Board
When we started ideating, to spark creativity and guide the design direction, we created an inspiration board that showcased visual elements, styles, and concepts relevant to the project.
When we started ideating, to spark creativity and guide the design direction, we created an inspiration board that showcased visual elements, styles, and concepts relevant to the project.
Low fidelity Wireframes
Low fidelity Wireframes
Low-fidelity wireframes explored two layout approaches — a multi-panel dashboard and a single-flow sequential view. Agent feedback in early testing strongly preferred the dashboard model for at-a-glance overview, which became the foundation for all subsequent iterations.
Low-fidelity wireframes explored two layout approaches — a multi-panel dashboard and a single-flow sequential view. Agent feedback in early testing strongly preferred the dashboard model for at-a-glance overview, which became the foundation for all subsequent iterations.
Branding
Branding
We developed a branding strategy that reflects our target audience’s values and aesthetics, setting the foundation for a cohesive design experience.
We developed a branding strategy that reflects our target audience’s values and aesthetics, setting the foundation for a cohesive design experience.
High fidelity Wireframes
High fidelity Wireframes
With a strong foundation in place, we brought our vision to life through high-fidelity wireframes, weaving in branding, colors, typography, and interactive elements to create a seamless and engaging user experience.
With a strong foundation in place, we brought our vision to life through high-fidelity wireframes, weaving in branding, colors, typography, and interactive elements to create a seamless and engaging user experience.
Unified dashboard surfaces active ticket, AI translation, suggested response, and case history in a single view ultimately eliminating the need to switch between tools mid-conversation.
Unified dashboard surfaces active ticket, AI translation, suggested response, and case history in a single view ultimately eliminating the need to switch between tools mid-conversation.
AI-generated response shown in an editable field with a confidence indicator preserving agent control while reducing drafting time from ~4 minutes to under 60 seconds.
AI-generated response shown in an editable field with a confidence indicator preserving agent control while reducing drafting time from ~4 minutes to under 60 seconds.
Agents can rate AI suggestions after each interaction — creating a feedback loop that improves model accuracy over time and gives agents a sense of ownership over the tool's quality.
Agents can rate AI suggestions after each interaction — creating a feedback loop that improves model accuracy over time and gives agents a sense of ownership over the tool's quality.
What was the outcome?
What was the outcome?
Results
Results
The AI-Gen HelpAssist tool delivered measurable improvements across the support desk operation:
Response Speed: Automated translation and response suggestions reduced average ticket resolution time and agents reported handling significantly more tickets per shift without increased effort.
Accuracy: AI-assisted responses reduced miscommunication incidents, particularly for multilingual tickets where manual translation errors were most common.
Agent Experience: Post-launch feedback showed agents felt more in control and less overwhelmed, the tool removed repetitive tasks while keeping human judgment central to every interaction.
Stakeholder Reception: The prototype was presented to and approved by senior stakeholders at Accenture, validating the design direction for further development.
What I Learned
What I Learned
Designing for AI-assisted workflows taught me that the hardest part isn’t the AI, it's the trust gap. Agents were initially skeptical of automated suggestions, fearing it would undermine their expertise. The design solution wasn't just a UI - it was a trust-building system. Every interaction was designed to make agents feel in control, not replaced.
I also learned that enterprise constraints are design constraints. Working within Accenture's technical stack, security requirements, and approval process forced me to design for implementation, not just for ideal conditions. That tension made me a more pragmatic, realistic designer.
If I were to continue this project, I would build out an analytics dashboard for team leads, explore proactive ticket routing based on agent availability and language skills, and test a knowledge base auto-generation feature using resolved ticket history.
View My Other Projects
↑