customer_support · saas · workflow
Architecture of today's LLM applications: five steps and emerging component patterns
Building software with LLMs is fundamentally different from traditional software development, requiring developers to navigate datasets, embeddings, and parameter weights instead of compiling source code, with outputs that are probabilistic rather than predictable.
How it works
Common implementation structure
How this type of workflow is generally built, generalized across documented cases — not tied to any one vendor's stack. Click any stage to read what happens there. Specific products that implement these stages appear in “Tools commonly seen” below.
Stage 1 · User contacts LLM assistant
Dave calls his internet service provider and is directed to an LLM-powered assistant.
Tools used
GitHub Copilotlangchain-ai/langchainMongoDBQdrantPineconeMilvusGPTCacheOpenTelemetry
Outcome
(not stated)
Grounding & classification
Source type: generic use case
19 fields verified against source quotes, 2 dropped as unverifiable.
conversational aiknowledge searchragspeech to textcall recordingknowledge basetools describedworkflow describedsoftwaretelecomgeneric use casecustomer supportit supportautonomous resolutionrag answering