USE CASE Analytics Assist Leveraging Gen AI to Retrieve Insights
Project Overview
Client: ADP
Role: UX Director, UX Designer
Time Frame: 2023 - 2024
Scope: Vision/New Product Design
Industry: Enterprise Software UX
Problem:
HR teams face challenges in deriving actionable insights from complex data due to difficulties in defining and articulating problems, as well as the need for extensive filtering and specialized expertise. This creates a barrier to effective decision-making. Generative AI offers a solution by simplifying data interaction, enabling more intuitive and efficient access to insights.
Objective:
Introduce an NLP-driven, prompt-based experience that would enable HR and Payroll practitioners to retrieve insights by asking natural language questions about their data.

This solution aimed to simplify the process of interacting with complex datasets and facilitate the creation of data visualizations and dashboards, ultimately improving user engagement and efficiency in the Analytics platform.

Phase 1
Discovery/Ideation

The goal of our discovery process was to define the MVP, with a product strategy centered around an alpha release to enable real-time testing with live data. At the alpha stage, the focus is on evaluating the product's functionality rather than achieving a perfect solution, giving UX the opportunity for extensive ideation and collaboration with Product, Data Scientists and Prompt Engineers.

Internal Ideation Sessions

Since this was a vision project, we conducted idea-generation research activities, including affinity mapping and card sorting exercises. During this time, we also focused on educating ourselves about prompt engineering and gaining a deeper understanding of how our current analytics models were structured, along with the related opportunities and challenges.
CAB Sessions

Through internal sessions with Product Owners, we gathered ideas that we presented to our Customer Advisory Board (CAB) to solicit input and prioritize needs against strategic objectives. We then held additional sessions with CAB clients to assess their understanding and sentiment towards AI-driven insights and assistive features.

Competitive Analysis

The competitive analysis focused on a select group of key players in the HR industry, including Visier, DataGPT, and Intuit, as well as major technology leaders like Google and ChatGPT.

Research Outcomes
Opportunities

Out of the multiple rounds of engagement with product and our CAB we were able to define some opportunities to improve the user experience with our People Insights offering when it came to these 3 key exercises:

Enhancing Insight Discovery with NLP

Implement natural language capabilities for question-and-answer interactions using NLP. The existing curated, themed dashboards provided an effective starting point for insight exploration, but they lacked the necessary customization and flexibility to adapt to diverse business needs.

Streamlining The Dashboards Creation Workflow

The dashboard creation process was highly labor-intensive, requiring significant time to find and customize metrics, resulting in a repetitive and time-consuming workflow.
Reducing The Time to Value

Most tasks required multiple steps to achieve the desired outcome. While traditional UIs followed an imperative approach, a declarative approach offered the potential to significantly reduce time to value.


Phase 2
Concept Design

The first concept aimed to create a dedicated "Assistant" tab for an immersive experience, akin to ChatGPT, prioritizing the perception of this feature as a destination rather than focusing on specific capabilities. Our objective was to gather practitioners' sentiments and experiences with artificial intelligence, exploring use cases, identifying potential pain points, assessing the impact on their roles, and collecting preferences between two design options.


What:

Presenting visual ideas to users helped establish early direction by focusing on high-level concepts rather than detailed interactions, distinguishing this approach from traditional usability tests.

Why:

This process aimed to validate early-stage solutions, assess whether the design intent resonated with users, and inform product improvements, ultimately reducing the risk of creating a design that didn’t meet user needs.

Evaluative Research
Research Outcomes & Themes

Users envisioned using the assistant for both quick questions and deeper analysis. They preferred a hybrid design that included a separate Assistant tab while maintaining access to the assistant from any page, allowing them to quickly ask questions while working and easily navigate to the Assistant tab for more detailed insights.
Trust and Security:

Users need assurance that their data remains confidential and is not shared outside of the ADP environment.

Effective Communication:

Users seek guidance on formulating their questions effectively to achieve better responses.

Respect for User Settings:

Users want confirmation that their individual settings and permissions are upheld.

Response Optimization:

Users appreciate tips on how to interact with the assistant to maximize the quality of the responses received.
Source Verification:

Users require reassurance that the information provided by the assistant is sourced from ADP and has been properly vetted.


Phase 3
Concept Refinement

After evaluating user feedback and considering what was viable within our current UI and tech stacks, we chose to place the access point on the overview page. This decision also aligned with long-term plans to integrate with the broader Gen AI strategy across the ADP ecosystem.


Wireframe Exploration
During the subsequent iteration cycle, we focused on developing the chat UI component, experimenting with a floating and draggable version, a docked version, and a version that could be pinned.

Data Visualization
Incorporating data visualizations into the query responses was essential, as they provided context for existing metric calculations and offered deeper insights into trends and comparisons. It was also necessary to account for various chart types and scaling concerns when users exited the immersive experience.


Phase 4
Detailed Designs

We deployed concept refinement changes to clients for engagement and data collection, shifting our focus towards a North Star solution that delivers intrinsic value and fully leverages the potential of Gen AI.


North Star Designs
Focused on the “So What” Factor

Enhanced the analytical responses by addressing the "so what" factor, ensuring that each insight provided tangible value and relevance.

Analytics and User History Integration

Developed methods to capture user history and analytics, informing future interactions and enhancing personalized experiences.

Refined Visual Data Representation

Users want confirmation that their individual settings and permissions are upheld.

Guided Interaction Flows

Designed UI elements that suggest options, guiding users through complex decision-making and ambiguity.
Added Contextual Relevance to Results

Improved the contextual presentation of AI-generated results, making them more actionable and insightful.



Outcomes & Learnings


As a nascent technology, AI analytics presents numerous unknowns and challenges related to user trust. This uncertainty is compounded by the tendency of AI systems to hallucinate and heavily rely on probabilistic assessments. Users often expect data analytics to yield definitive answers—either correct or incorrect—creating a mismatch between their binary expectations and the nuanced nature of AI-generated responses.

While most users are enthusiastic about the potential of AI and many are already leveraging AI tools to streamline tedious tasks, those who have engaged with these tools frequently encounter a frustrating cycle of trial and error. This process can lead to dissatisfaction, as users grapple with the difficulty of obtaining usable and reliable answers.





Additional Use Cases


Enhancing Customer Retention to Drive Revenue Growth

A redesign of the Analytics and Benchmarking platform dashboard for Human Capital Management (HCM) to enhance user experience, improve data accessibility, and provide actionable insights for HR professionals and business leaders.

view use case
Creating a Seamless Multi-Platform User Experience

A comprehensive redesign of the Executive Manager Insights user experience in Analytics for mobile, ensuring it aligns seamlessly with the Practitioner experience on desktop..

view use case

 

Let's connect! I'd love to share more about these projects and walk you through detailed use cases. Looking forward to our conversation!