HR Virtual Assistant v2.0

Context

Leena AI's flagship product was an AI-powered HR virtual assistant deployed across 90+ enterprise clients globally. The v1.0 product handled basic HR queries but had limited understanding of complex, multi-turn conversations and lacked the enterprise-grade features needed by Fortune 500 clients.

The Problem

Enterprise HR departments were drowning in repetitive queries — leave policies, benefits enrollment, payroll questions — consuming 40-60% of HR team bandwidth. While v1.0 addressed basic FAQ-style queries, customers demanded more: multi-turn conversations, policy-aware responses tailored to their organization, and integrations with their HRIS systems. Churn risk was rising as competitors caught up on basic chatbot features.

Discovery & Research

I partnered with 4 Fortune 500 clients to deeply understand their pain points. The key finding: it wasn't just about answering questions — it was about understanding context. An employee asking about 'leave' at 3 PM on a Friday has a different intent than one asking at 9 AM on a Monday. I mapped 200+ intent patterns across client organizations and identified the top 25 features that would cover 80% of unresolved queries.

• ·Partnered with 4 Fortune 500 clients for deep discovery
• ·Mapped 200+ intent patterns across organizations
• ·Identified context-aware responses as the key differentiator
• ·Prioritized 25 features using RICE scoring framework

Solution

I managed a 10-member cross-functional team (NLP engineers, frontend, backend, QA) to rebuild the conversational AI pipeline. Key architectural decisions included moving from rule-based to transformer-based intent classification, adding organization-specific policy embedding, and building a self-service admin portal for HR teams to customize responses without engineering support. I authored PRDs for each feature wave, ran weekly sprint planning, and personally demoed progress to C-suite stakeholders at client organizations.

Results & Impact

The v2.0 launch drove $3M in new annual revenue and expanded the user base to 1.5M across North America, Europe, and Asia. Resolution rates improved from 60% to 85%, and the self-service admin portal reduced implementation time from 6 weeks to 2 weeks.

Key Learnings

Building for enterprises taught me that the product is only 50% of the value — the other 50% is the implementation and ongoing customization experience. The self-service admin portal drove more expansion revenue than any individual chatbot feature.

• ·Enterprise products need implementation excellence, not just features
• ·Self-service customization tools > professional services at scale
• ·Weekly stakeholder demos built trust and prevented scope creep