CometSoftware Group
Consulting | ERP
AI chatbots deployed across three ERP brands. one infrastructure, three tailored experiences.
CometSoftware Group manages multiple ERP products. Volume, Saiga, and Cover3D. each serving different markets with different documentation and user workflows. They needed intelligent support that understood each brand's specifics without tripling their support costs.
Rakam built AI-powered chatbots customized per brand on a shared infrastructure. each bot knows its product inside out, answers questions from documentation (including PPTX presentations), and provides user feedback tracking for continuous improvement.
« CometSoftware needed AI that worked across three distinct product lines without tripling costs. Rakam delivered a shared system where each brand gets its own tailored experience. »
Maxence David-Affre, CEO of CometSoftware Group
Consulting | ERP
3
ERP brands covered
1
Shared infrastructure
PPTX
Document ingestion
Delivered
Production-ready
Business
Business Impact
AI chatbots across three ERP brands provide first-line support for each product's user base, reducing ticket volume and improving response times. Shared infrastructure keeps operational costs low while each bot is tuned to its brand's specific documentation.
Each chatbot ingests brand-specific documentation. including PPTX presentations. and provides accurate, contextual answers. A built-in feedback system enables continuous quality improvement driven by real user signals.
Product
What We Built
Multi-Brand ERP Chatbots
AI chatbots for Volume, Saiga, and Cover3D. each with brand-specific knowledge bases, documentation ingestion (including PPTX), and tuned response styles. Branch-based agent configuration for independent updates per brand.
Document Ingestion Pipeline
Automated pipeline that ingests product documentation in multiple formats. PDFs, PPTX, text. and indexes it for semantic and keyword search. New documents are automatically processed and available to users.
évaluation & Feedback System
Built-in évaluation framework with dataset validation, multi-run comparison, and model cost tracking. User feedback system feeds continuous improvement cycles.
Technical
Technical Architecture
The chatbot system uses pgvector for semantic search across brand-specific documentation, combined with a French GIN index for keyword fallback. Microservice architecture with S3 storage and Nginx reverse proxy ensures each brand's bot can be updated independently. Delivered via GitHub Container Registry.
# Stack
pgvector (semantic search)
French GIN index (keyword fallback)
S3 (document storage)
Nginx (reverse proxy)
GitHub Container Registry (CI/CD)
Kubernetes OVH (orchestration)
# Architecture
3 brands, shared infra
Branch-based agent config
PPTX + PDF ingestion
évaluation + feedback system
Move from AI hype to the real thing
Leave us a message, our team will get back to you within 24 hours.