Mars A360 Chatbot (YODA)
An Enterprise AI Decision Intelligence Platform transforming HR analytics from fragmented dashboards into conversational business intelligence.
Project Overview
Enterprise HR teams rely on hundreds of dashboards, reports, spreadsheets, and business intelligence tools to understand workforce performance.
Answering what appears to be a simple business question—such as: "Why has attrition increased this quarter?"—often requires navigating multiple dashboards, exporting data into spreadsheets, manually correlating information across systems, creating presentation slides, and writing recommendations from scratch. This process is slow, fragmented, and heavily dependent on the analyst's expertise.
"To address this challenge, Mars envisioned YODA, an AI-powered decision intelligence platform embedded within the Associate 360 (A360) ecosystem, allowing leaders to ask questions in natural language and receive contextual insights."
Business Context
Modern HR organizations generate enormous volumes of workforce data across multiple enterprise systems. Associate 360 (A360) serves as Mars' centralized workforce intelligence platform, bringing together:
Within this ecosystem, YODA acts as the intelligent conversational layer, helping users access insights across multiple dashboards without requiring technical expertise or prior knowledge of where the data resides.
Problem Statement
Business leaders needed faster answers — not more dashboards.
A typical analytical workflow involved opening multiple reporting systems, manually filtering information, exporting datasets into Excel, comparing trends across different business units, interpreting patterns, creating PowerPoint presentations, and writing executive summaries. This fragmented process consumed six to eight hours for questions that should ideally take only minutes to answer.
Primary Objectives
Reduce Time to Insight
Enable users to receive meaningful insights through conversational interactions rather than manual dashboard navigation.
Democratize Analytics
Allow business users without technical expertise to access advanced analytics using natural language.
Improve Decision Quality
Surface correlations, trends, and recommendations that might otherwise remain hidden across disconnected reporting systems.
Increase Trust in AI
Provide explainable responses supported by visualizations, source references, and confidence indicators.
Drive Action
Move beyond descriptive reporting by recommending practical next steps that help leaders respond to workforce challenges.
Unified Intelligence Layer
Position YODA as the central conversational interface across the broader Associate 360 ecosystem.
YODA Conversational Interface
Select a preset question in the bar below, then toggle between Desktop Web and Mobile views.
Mobile App & Analytics Interface
Below is the high-fidelity mobile application layout, designed for on-the-go conversational analytics.
Ask YODA
OnlineMy Design Responsibilities
As the Product Designer, I led the experience design for YODA, focusing on transforming complex analytical workflows into intuitive conversations:
Conversational UX
Dashboard Design
Trust & Transparency
Prompt Engineering UX
Primary User Personas
HR Business Partners
Need quick visibility into workforce trends, attrition, and engagement without manually navigating multiple dashboards.
Talent Acquisition Leaders
Monitor hiring performance, recruitment bottlenecks, candidate pipelines, and workforce planning metrics.
People Analytics Teams
Investigate trends, validate hypotheses, and perform deeper workforce analysis using conversational exploration.
Executive Leadership
Require concise summaries, AI-generated narratives, strategic recommendations, and rapid access to high-level insights.
As-Is Manual vs. YODA Journey Mapping
Compare the workflows of a People Analyst trying to identify and resolve attrition SLA risks.
Understand: "Where are we stuck now?"
Of all the open and recently closed reqs, which ones are critical or at risk, and where are the hotspots?
- Opens the TA Executive Dashboard to see overall pipeline and time-to-fill.
- Runs multiple ATS reports (open reqs, aging, closed-no-hire, SLA breaches).
- Exports to Excel, builds filters/pivots by BU, role, level, recruiter to include further details for ATS to identify relevant hotspots.
Deep dive: "Where exactly is the problem and why?"
For those hotspots, at which stage are requisitions getting stuck or failing, and what is driving it?
- Clicks into individual reqs from TA Executive Dashboard / ATS to check candidate counts/aging.
- Opens candidate timelines, reads notes and disposition codes.
- Builds a manual view in Excel comparing stage-wise conversion and declines.
- Tries to infer reasons (e.g. comp, manager delays).
Decide & Act: "What do we change?"
Which open requisitions this month are likely to become next month's failures, and what actions should we agree?
- Scans open reqs list and manually picks oldest/critical requisitions.
- Writes emails or slides summarizing risky requisitions to push sourcing/interviews.
- Builds PowerPoint presentations by manually copying charts from Excel.
Key Design Decisions
1. Questions, Not Dashboards
Shifted the entry point from dashboard navigation to user intent. Users describe a problem in plain language, removing the need to know backend tables.
2. Analyst, Not an Oracle
Positioned YODA as a partner rather than an oracle. Conclusions are presented alongside what was analyzed, why it matters, and supporting evidence.
3. Conversations as Workflows
Structured chats to mirror analytical investigations (Understand state → Root Cause → Compare Factors → Impact → Recommend Actions → Monitor Outcomes).
4. Progressive Disclosure
Concise summaries first, supporting charts second, detailed evidence third, raw datasets last. Balances executive speed with analyst depth.
Product Architecture
Rather than organizing around features, the platform is structured into four interconnected layers of user intent:
Layer 1: Awareness
Executive summary cards and proactive KPIs surface highlights ("What deserves my attention today?").
Layer 2: Exploration
Natural language input matches intent to relevant workforce datasets and triggers automated SQL retrieval.
Layer 3: Investigation
Progressively uncovers correlations (e.g. Attrition → Unit → Location → Tenure → Recruiter → Compensation).
Layer 4: Decision
Proposes practical interventions, reallocations, and action checklists to resolve bottlenecks.
Designing Explainability & Trust
Trust cannot be added after responses are generated. Explainability is a first-class UX component. YODA supports answers with:
Confidence Indicator
Visually communicates evidence strength, letting users calibrate trust.
Source Attribution
Every response references the specific databases and reports contributing to the insight.
AI-Generated Visuals
Funnel charts and heatmaps reinforce the text narrative for faster comprehension.
Conversational UX Principles
Conversations Replace Navigation
Natural language replaces complex dashboard filtering structures, reducing analytical effort.
Answers Generate Better Questions
YODA suggests logical follow-up prompts, guiding users through deeper layers of investigation.
Minimize Context Switching
Keeps users within a single conversational workspace, seamlessly embedding charts and recommendations.
Technical Considerations
Performance Feedback
Detailed loading states reassure users that complex background calculations are happening.
Security & Privacy
Designed to strictly respect organizational access controls and HR data privacy limits.
Multi-Agent AI
Orchestrates multiple dedicated sub-agents (EDA, SQL query generator, visualization intelligence).
The Final Solution
The resulting platform transforms workforce analytics from fragmented, dashboard-centric reporting into an intelligent conversational decision-support system. By combining conversational AI, enterprise analytics, and explainable reasoning, YODA significantly reduces the effort required to make informed HR business decisions.
Business Impact Mapped
Future Roadmap
Potential future enhancements for the Mars YODA platform:
Predictive Analytics
Forecast future attrition and hiring demand before they become bottlenecks.
Autonomous AI Agents
Proactively monitor organizational metrics and notify partners before queries begin.
Scenario Planning
Evaluate the impact of hiring freezes or compensation adjustments virtually.
Cross-Domain Intelligence
Expand beyond HR into finance, supply chain, procurement, and operations.
Key Learnings & Reflections
This project reinforced that successful conversational experiences depend as much on information architecture, reasoning transparency, and workflow design as they do on natural language processing. Designing for AI means designing the entire decision-making journey, not just the interface where users type their questions.