order_processing · travel · workflow
Choco AI scales LLM order extraction with few-shot learning and dynamic context injection
Choco AI had to process restaurant orders arriving in highly varied formats—shorthand texts, PDF email attachments, and voicemail recordings—and information extraction from these orders introduced errors including wrong-column extraction, misinterpreted abbreviations, and speech-to-text mistranscriptions.
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 · Order arrives in varied format
Orders arrive as shorthand text messages, PDF email attachments, or voicemail recordings.
Tools used
Whisperembeddings model
Outcome
Dynamic context injection and semantic few-shot retrieval consistently led to higher accuracy in information extraction, and smaller models achieved comparable performance with reduced cost.
What failed first
Metadata-based retrieval of few-shot examples had limitations with order layout variations, and a generic catch-all prompt improved performance initially but struggled with brand names and uncommon products.
Results
Volumehigher accuracy
Cost replacedreducing costs and increasing agility
Grounding & classification
Source type: technical build writeup
27 fields verified against source quotes, 1 dropped as unverifiable.
data extractiondocument aipersonalizationragspeech to textcall recordingemailbuilder submittedfailure mode describedhuman review describedproduction runtime claimedtools describedworkflow describedhospitalitylogisticsaccuracy improvementcost reductiontechnical build writeupdata entry opsorder processingdocument to recordextract classify route