Grant Platform AI Assistant

The Grant Platform is a web platform that serves as the primary interface through which organizations discover, apply for, and manage funding programs — spanning the full lifecycle from discovery through decision.

The AI Assistant is an embedded chatbot that helps applicants navigate program eligibility, find the right funding stream, get answers to submission questions, and complete applications — without needing to call support or wait for a staff response. It was designed to reduce friction at the highest drop-off points in the funnel without replacing the human advisory layer for complex cases.

As Technical Product Manager, I owned the full delivery lifecycle — from discovery and requirement definition through sprint execution, stakeholder alignment, and go-to-market. I worked directly with engineering, design, and the executive team to define scope, set success metrics, and manage prioritization trade-offs across the delivery timeline.

• Ran discovery across applicant support tickets, drop-off data, and funnel analysis (Mixpanel, Hotjar) to identify the highest-friction points in the application journey — navigation confusion, eligibility uncertainty, and mid-form abandonment
• Defined the AI assistant's scope: answering program eligibility questions, surfacing relevant funding streams, guiding users through submission requirements, and escalating edge cases to human advisors
• Translated OKRs into user stories and acceptance criteria, improving sprint delivery velocity by 30% and compressing release cycles by 25%
• Managed prioritization across engineering and design — balancing AI response quality, platform stability, and delivery timelines within a fixed sprint cadence
• Coordinated go-to-market with communications, customer success, and the executive team — hitting adoption targets within 30 days of launch
• Monitored post-launch performance through engagement metrics and submission funnel data to validate impact and inform follow-on iterations

• AI as an enablement layer: the assistant was most effective when scoped to specific, high-frequency friction points (eligibility questions, navigation) rather than trying to automate the full advisory experience
• Funnel data drives scope decisions: drop-off analysis was the single most credible input for prioritizing which questions the assistant should answer first — more actionable than any stakeholder assumption
• OKR-to-story discipline compounds: translating OKRs into explicit acceptance criteria before sprint planning reduced scope creep and gave engineering a clear definition of done, which directly drove the velocity improvement
• Soft launch with monitoring: shipping to a subset of users first and tracking engagement and submission rate before full release gave us a feedback loop to validate the assistant's responses before broad exposure