SayTrue homepage introducing AI customer support
Back to Selected Work
Product

SayTrue: AI customer support with citations and human handoff

SayTrue answers questions across web chat, phone calls and social channels using approved knowledge. Answers cite their sources; low-confidence conversations go to a person.

Ruby On RailsAI/LLM WorkflowsRAGVoice TranscriptionOmnichannel Messaging
Engagement
Own product · AI customer support
Stack
Ruby On Rails
Domain
AI/LLM Workflows · RAG · Voice Transcription · Omnichannel Messaging

Project reference

Answers from approved knowledge

SayTrue is an AI customer service platform for web chat, phone calls, Messenger and Instagram DMs. It answers from approved material, cites the source and hands the conversation to a person when confidence is low.

For regulated teams, an answer about policy needs a traceable source. A customer, supervisor or regulator may need to check it later. A fluent answer alone does not meet that requirement.

What we built

We built the channels around the same approved knowledge base, with citations back to the source documents. Confidence thresholds trigger an explicit fallback and human handoff rather than a guessed answer.

The phone workflows include voice transcription and searchable transcripts, so supervisors can review a conversation without listening through the whole call. Messenger and Instagram DMs use the same source-grounded approach as web chat.

The product is EU-hosted and GDPR-aware. It shipped with an open changelog, a demo and a contact funnel.

Stack

Ruby on Rails, RAG and AI/LLM workflows, with voice transcription, omnichannel messaging and GDPR-aware infrastructure hosted in the EU.

Next project: Uskladi.eu

View all selected work