Frictionless Customer AfterSale Service (FCAS)

What it does

Our platform acts as an intelligent After-Sale Concierge that centralizes receipts, policies, and warranty data into one seamless dashboard for both consumers and enterprises.

• For Consumers: It automatically tracks coverage and alerts users to claim opportunities (returns, price drops, or expiring warranties).
• For Businesses: It provides an AI-powered agent that instantly looks up specific issues, verifies warranty status across multiple product lines, and triggers repair or return workflows immediately.

How we built it

• Built a full-stack monorepo with Next.js, React, Bun, and a modular package architecture.
• Created a structured receipt + item data model with Postgres/Drizzle to normalize merchants, orders, invoice IDs, and line-level warranty fields.
• Implemented an AI chat router with tool-calling for:
  • customer verification
  • transaction history retrieval
  • team determination and forwarding
  • policy-aware warranty guidance
• Added a private policy knowledge library (RAG-first flow) so answers prioritize curated internal policy sources before falling back to web search.
• Limited fallback web search to a strict cap for cost/control, and surfaced tool outputs directly in the UI with custom components.
• Built operational workflows including ticket creation on successful forward-to-team actions and dashboard pages for ticket analytics and filtering.
• Designed rich frontend experiences: interactive transaction cards, policy management, and voice playback controls for assistant responses.

Challenges we ran into

• Handling complex AI tool-streaming edge cases (reasoning/tool item sequencing errors, partial states, and rendering mismatches).
• Keeping UI stable while rendering many dynamic tool outputs with unique keys and consistent state transitions.
• Preventing brittle matching logic in transaction lookup; partial text matching caused false positives.
• Integrating serverless DB constraints (e.g., Neon HTTP transaction limitations) with workflow-heavy support operations.
• Balancing dark/light theme parity and responsive layouts for dense dashboard views (overflow, action placement, card behavior).
• Ensuring support escalation quality: requiring verified users, requiring item IDs before forwarding, and preserving auditability with ticket records.

Accomplishments that we're proud of

• Delivered an end-to-end After-Sale Copilot: from receipt ingestion to warranty decision to team handoff.
• Implemented RAG-first warranty assistance that uses private policy sources for more reliable, brand-specific answers.
• Built an actionable support experience, not just chat:
  • users can select purchased items directly from history
  • assistant creates structured escalation tickets
  • teams get clearer context and faster triage
• Upgraded the UX significantly with interactive components, streamlined controls, and voice-enabled response playback.
• Created a foundation that works for both B2C self-service and B2B support operations.

What we learned

• Reliability in AI support systems comes more from workflow design and data quality than model prompts alone.
• Tool output UX matters as much as model intelligence; users trust what they can see and click.
• Structured constraints (required item IDs, capped web calls, explicit escalation rules) dramatically improve support safety.
• RAG with curated policy sources is critical for warranty accuracy and reduces hallucinated guidance.
• Small frontend details (stateful controls, responsive overflow handling, visual hierarchy) have outsized impact on perceived product quality.

What's next for Frictionless Customer AfterSale Service (FCAS)

• Expand policy ingestion with stronger PDF parsing, versioning, and source freshness monitoring.
• Add confidence scoring + explanation traces for each warranty decision.
• Introduce SLA-aware routing and priority queues for enterprise support teams.
• Launch proactive automation: claim reminders, return-window nudges, and auto-generated support drafts.
• Add multilingual support and localized policy interpretation for cross-region operations.
• Deepen analytics: claim conversion, time-to-resolution, policy coverage gaps, and support cost savings.
• Integrate with repair partners/logistics APIs for truly one-click repair/return orchestration.

Built With

• nextjs
• openai
• vercel