Need
The copilot already answered questions, routed tickets and used knowledge from multiple business modules. Intervan still depended on Develo for every knowledge update and lacked the operational visibility, controls and channels required for production.
Solution
Copilot for 1,000+ users
A multitenant architecture designed to serve more than 1,000 users while isolating each organization's context, rules and documentation.
Operational backoffice
One workspace for conversations, human intervention, configurable metrics, reports, users, roles and business rules.
Dynamic knowledge base
Intervan can load and edit documents, ingest transcribed video, reindex automatically and evolve RAG knowledge without developer intervention.
Actions through MCP
A governed action catalog connects the copilot to Toba. Explicit confirmation, inherited permissions and full traceability protect every execution.
Architecture
- AWS infrastructure for the copilot and backoffice, designed for resilient horizontal scale
- Multitenant context, rules and knowledge with an initial path for Municipalidad de Neuquén
- Panel for conversations, metrics, configuration, automated reports and exports
- MCP tools over Toba APIs for scoped queries, updates and reports
- Human supervision, real-time notifications, escalation and optional osTicket integration
- WhatsApp Business through Meta APIs, using the same governed RAG layer
Operational outcome
The result is a functional, scalable operating model: Intervan can manage knowledge, inspect service quality and control the copilot from its own backoffice. The product can extend from answering to executing approved workflows, with permissions and audit evidence attached to every action.
Related
Build an AI product your team can operate
Book a meeting to map users, channels, knowledge and governed actions for your rollout.
Book a Meeting