An AI-native customer support platform for high-volume e-commerce came to Armakuni for genAI customer support assistant for e-commerce: live knowledge retrieval, RAG, semantic intent, continuous learning.
The challenge
A GenAI platform engagement covering language understanding, retrieval, continuous learning, and production architecture, positioning SolvPath to scale to additional merchants and verticals on a single proven stack.
Scale: Real-time query handling at e-commerce peak load. 99.9% uptime sustained across high-traffic shopping events. Knowledge base coverage spanning FAQs, policies, order systems, and loyalty across merchant catalogs.
What we built
Armakuni delivered genAI customer support assistant for e-commerce: live knowledge retrieval, RAG, semantic intent, continuous learning. The engagement ran as fixed-scope SOWs covering NLU and intent layer, RAG and knowledge retrieval pipeline, continuous learning infrastructure, and production runtime hardening.
The outcomes
5 measurable outcomes shipped across the engagement. The ones that moved the business the most:
95% accuracy understanding shopper queries across typos and complex questions. The assistant handles regional phrasing and multi-part questions, resolving the long tail of support volume that previously had to escalate to human agents.
90% of responses contextually relevant and grounded in merchant knowledge. With no hallucination risk, merchants trust the assistant with policy and order questions where a wrong answer carries real cost.
85% semantic understanding accuracy on intent matching. Across paraphrased queries the assistant lifts resolution rates on the highest-volume question types and shrinks the share of queries that need human handoff.
99.9% uptime through peak traffic including Black Friday. SolvPath holds its SLA on the merchant-specific promotional days customers need it most, not just the days the platform is quiet.
Continuous learning loop running in production. Accuracy improves on a measurable cadence rather than as a marketing promise, giving the commercial team data-grounded answers in every merchant renewal conversation.
Built on AWS
The production environment runs on 12 first-party AWS services. Delivered under our AWS AI Services, SaaS Consulting, DevOps Consulting competencies.
What's next
Expanding the assistant into agentic capabilities that take action on returns, exchanges, and order modifications directly, and broadening merchant verticals beyond core e-commerce into adjacent customer-facing categories.
