Home>Work>Confidential SaaS Co.
SaaS & Technology

Enterprise RAG for Support Teams

A production RAG platform giving support agents instant, cited answers from internal documentation — cutting resolution time by more than half within six weeks of launch.

Analytics Dashboard
Real-time Metrics+70% Recall
55%
Faster Resolution Time
33%
Fewer Escalations
92%
Citation Accuracy
ClientConfidential SaaS Co.
Timeline10 weeks
Team Size2 engineers
CapabilitySaaS & Technology

The Challenge

Support agents spent an average of 6 minutes per ticket searching across scattered wikis, PDFs and Slack threads for the right answer — often finding an outdated one. Escalations to engineering were frequent and expensive, and customer satisfaction on first-response accuracy was falling behind target. The client's documentation lived across four disconnected systems, each with its own permissions model and update cadence. Any solution needed to respect access control per-agent while still surfacing a single, unified answer.

The Approach

1

Wiki Scoping & Discovery

Mapped the client's internal wiki structures and identified primary support knowledge gaps.

2

Structure-Aware Chunking

Parsed documents into hierarchical markdown chunks, preserving tables and section dependencies.

3

Hybrid Indexing & Deployment

Configured BM25 + Qdrant dense vector search with cross-encoder rerankers.

Outcomes & Impact

Within six weeks of launch, average resolution time dropped by 55%, and every answer surfaced a source link agents could verify in one click — building trust instead of replacing judgment. Escalations to engineering fell by a third, and the support team reported meaningfully higher confidence handling unfamiliar ticket types. The client has since expanded the system to their internal engineering knowledge base and is evaluating a customer-facing version for their help center.

Key Lessons Learned

Retrieve in context — a flat vector index loses section headers.
Metadata filtering is key for access control permissions.
Observe queries continuously to capture missing documentation gaps.

"The custom search engine built by Botmartz changed how our operations team works. They recover hours every day and container routing errors have dropped to near zero."

MV
Markus Vance
VP of Operations, Global Logistics Leader

Core Technologies

LangChainQdrantFastAPIDocker

Development Team

KG
Karan Goel
AI Engineer
PB
Promit Basu
AI/ML Engineer

Related Cases

SaaS & SupportContext-Aware Agent for SaaSResult: 45% Automated Resolution
Legal TechContract Risk Extraction EngineResult: 90% Time Saved

Want to see similar results?

Let's build a custom AI solution that fits your business specs.

Discuss Your Project