About Services Case Studies Pricing Blog Book Free Audit
INTERNAL AI HELP DESK20262 min read

Internal Help Center AI Chatbot on Amazon Q Business

A managed, plug-and-play internal help-desk bot delivered in a fraction of the time a fully custom pipeline would have taken, styled to match the client's own internal tooling and streaming responses over WebSocket for chat-speed answers.

CLIENT
Compass
INDUSTRY
AI SaaS (NDA)
TIMELINE
~4 weeks
ROLE
Amazon Q Business setup and knowledge base integration, custom React front-end, real-time WebSocket transport layer
// overview

Compass is a separate, internal-only assistant for a client's own employees to self-serve answers from internal documentation. Lower stakes than a customer-facing assistant, a narrower audience, and a much shorter runway to ship — a good candidate for Amazon Q Business's plug-and-play knowledge base model instead of a fully custom RAG pipeline.

// the problem

Not every internal chatbot needs a custom pipeline.

The client needed employees to get instant answers from internal documentation without waiting on a fully custom RAG build, an experience that looked and felt like its own internal tools rather than Amazon Q Business's out-of-the-box interface, and the same chat-speed responsiveness users expect from any modern chat product.

  • Instant answers from internal documentation, without a fully custom RAG build
  • A look and feel matching the client's own internal tools
  • Chat-speed responsiveness despite running on a managed backend
  • A tight timeline — weeks, not months
// approach

How we built it.

01

Amazon Q Business as the retrieval engine

Internal documentation was attached to Amazon Q Business as a knowledge base data source, letting the managed service handle chunking, embedding, retrieval, and answer generation without any custom orchestration code.

02

API Gateway WebSocket API

Rather than request/response HTTP polling, Compass communicates over a WebSocket connection through API Gateway, streaming responses back to the front-end as they're generated.

03

A custom React front-end over Q Business's default UI

Cosmoxyn built a bespoke chat interface on top of Amazon Q Business so the experience matched the client's internal tooling rather than looking like an out-of-the-box AWS product.

// what we shipped

The solution.

01

Compass — an internal help center chatbot on Amazon Q Business

Internal documentation attached to Amazon Q Business as a managed knowledge base, giving employees full retrieval-augmented answers without a custom RAG pipeline behind it.

02

A bespoke React front-end matching internal tooling

Rather than shipping Q Business's default interface, Cosmoxyn built a custom chat UI styled to match the client's existing internal tools.

03

Real-time streaming over WebSocket

Compass communicates over an API Gateway WebSocket connection, streaming responses to the front-end as they're generated.

04

A managed retrieval pipeline with zero custom backend orchestration

No Lambda functions, no custom retrieval code — Amazon Q Business handles chunking, embedding, and retrieval directly against the attached knowledge base.

// the results

Outcomes that shipped.

Internal employees got instant, chat-speed answers from internal documentation through Compass without leaving the look and feel of the client's own tools, and Compass shipped in roughly a third the time a fully custom pipeline would have taken without sacrificing responsiveness. Validated a build-vs-buy playbook for AI chatbots: reach for Amazon Q Business when speed of delivery and lower stakes matter more than fine-grained control over retrieval, feedback, and prompt iteration.

// stack
Amazon Q BusinessAmazon API Gateway (WebSocket)React
// learnings

What we took with us.

  • 01Bring a lightweight feedback loop to Compass sooner — even a simplified thumbs up/down from day one would have made it easier to measure answer quality.
  • 02Tighten WebSocket reconnect handling based on early production usage — real client network conditions surfaced reconnect edge cases not fully visible in initial testing.
// ready when you are

Have a build that needs this level of thinking?

Free 30-min audit. We'll tell you what to build, what it'd cost, and whether we're the right shop.