Case Study > Cnnect

Transforming enterprise operations with conversational AI

Transforming regulated B2B tobacco distribution through scalable Odoo automation and compliance- driven workflows.

Wholesale Distribution

Canada

Delivering scalable Odoo solutions for compliance-driven wholesale distribution and B2B commerce operations.

Tabac Arsenal Inc. is a Canada-based wholesale distributor specializing in premium cigars, vapes, and smoking accessories for verified retailers and wholesalers. Operating in a highly regulated B2B market, the company focuses on responsible distribution, compliance, and delivering premium products through a controlled wholesale ecosystem.

As operations expanded, Tabac Arsenal faced challenges related to compliance management, customer onboarding, tax automation, operational visibility, and fragmented workflows. The company required a scalable Odoo-based solution capable of streamlining B2B operations, automating compliance processes, improving reporting accuracy, and delivering a modern customer experience.

Key Challenges Faced

The existing onboarding process lacked structured workflows for business verification, permit validation, and approval-based customer activation, resulting in manual effort and compliance risks.

AI conversations carried extensive CRM context, uploaded files, tool-call results, and historical interactions, causing memory growth, latency issues, and increasing token consumption over prolonged sessions.

Enterprise users required seamless access to third-party systems like HubSpot, Microsoft Teams, and SharePoint directly from the conversational interface without interrupting workflows.

Proposal generation required dynamic CRM data mapping, AI-generated summaries, automated folder creation, and structured document population while preserving enterprise document standards.

When multiple files were uploaded into a single session, retrieval quality and response accuracy degraded due to document ranking conflicts and slow indexing performance.

AI-generated updates to DOCX templates often disrupted fonts, layouts, tables, content controls, and branding consistency, making enterprise documents visually unusable.

The platform needed proactive notifications capable of delivering automated workflow updates, approvals, generated documents, and project provisioning directly into Microsoft Teams.

The enterprise platform required a highly scalable architecture capable of handling concurrent AI conversations, asynchronous workflows, background indexing, and high-volume integrations without degrading performance.

Advanced AI Agent Orchestration

Implemented Azure AI Foundry v2.0 Modern Agents to orchestrate conversational workflows across multiple enterprise systems. Enabled dynamic context injection, intelligent intent routing, and streaming AI responses for seamless conversational experiences. Integrated resilient retry handling and session-aware processing to maintain stability during high-concurrency enterprise operations.

Automated Proposal Generation

Integrated asynchronous webhook processing with HubSpot CRM to automate proposal creation workflows based on deal-stage triggers. Generated structured SharePoint folder hierarchies and dynamically populated DOCX proposal templates using CRM data and AI-generated summaries. Streamlined enterprise proposal operations while reducing manual document preparation efforts.

Mid-Conversation OAuth & MCP Integration

Developed secure OAuth handling directly inside conversational workflows using the Model Context Protocol (MCP). Enabled users to authenticate with third-party platforms like HubSpot without interrupting ongoing AI conversations. Maintained seamless workflow continuity by resuming paused operations immediately after successful authorization.

Proactive Microsoft Teams Automation

Integrated Microsoft Teams and Graph API services to deliver proactive workflow notifications, Adaptive Cards, and enterprise alerts directly within Teams conversations. Enabled automated user resolution, deep-linked workflow actions, and real-time operational updates across enterprise teams. Improved collaboration visibility while reducing dependency on manual communication.

Document Processing & RAG Intelligence

Implemented Azure AI Search-powered Retrieval-Augmented Generation (RAG) workflows for intelligent document indexing and contextual AI retrieval. Enabled users to upload multiple documents, query content conversationally, and receive highly accurate responses across session-scoped files. Improved retrieval speed and accuracy through parallel indexing and hybrid ranking mechanisms.

Design-Preserving DOCX Editing

Engineered XML-level DOCX editing workflows capable of updating document content while preserving original layouts, branding, fonts, tables, and formatting structures. Ensured AI-generated text inherited existing style properties for visually consistent enterprise documents. Added version-controlled recovery mechanisms to safely restore previous document versions when required.

Enterprise CRM Intelligence Layer

Implemented a global CRM lookup cache with asynchronous refresh handling to optimize HubSpot property retrieval and improve UI responsiveness. Reduced API rate-limiting issues by maintaining portal-level cached CRM properties and refreshing them through background synchronization processes. Enabled high-speed enterprise CRM queries across large operational datasets.

Scalable Session & Memory Management

Designed a tiered memory architecture combining in-memory session handling, Azure Table Storage persistence, and Azure AI Search-based long-term retrieval. Implemented sliding-window summarization and lazy session rehydration to optimize token usage and maintain stable AI performance during multi-hour conversations. Reduced latency and stabilized enterprise-scale conversational workloads through intelligent session lifecycle management.

Key Features

Enterprise AI Orchestration

Designed multi-agent conversational workflows capable of coordinating CRM systems, AI agents, document automation, and enterprise integrations in real time.

Distributed Workflow Automation

Built asynchronous webhook-driven automation pipelines capable of orchestrating proposal generation, project provisioning, and enterprise notifications across multiple systems.

Secure Enterprise Authentication

Implemented OAuth and MCP-driven authentication flows to maintain secure access across HubSpot, Microsoft Teams, SharePoint, and enterprise APIs.

Scalable RAG & Document Intelligence

Engineered session-scoped RAG indexing, hybrid retrieval ranking, and intelligent document processing workflows for high-accuracy conversational AI responses.

Design-Safe Enterprise Document Editing

Developed XML-level document editing capabilities preserving enterprise branding, layouts, formatting, and content structures during AI-driven modifications.

High-Concurrency Backend Architecture

Leveraged asynchronous FastAPI endpoints, Azure storage systems, and scalable AI orchestration patterns to support enterprise-scale conversational workloads.

Technology Stack

React 18

TypeScript

Vite

Zustand

TailwindCSS

MSAL

FastAPI

Azure Table Storage

Azure Blob Storage

Azure AI Foundry SDK

Model Context Protocol

Azure AI Search (RAG)

HubSpot CRM

Microsoft Teams

Microsoft Graph API

SharePoint

Ruddr

Azure Active Directory

OAuth 2.0

Building Cnnect required far more than integrating enterprise tools—it demanded creating a unified conversational ecosystem capable of orchestrating AI, CRM intelligence, document automation, and workflow execution in real time. By leveraging Azure AI Foundry, FastAPI, and enterprise-grade integrations, the platform transformed fragmented operations into a scalable, intelligent automation experience for modern enterprise teams.

Tejas Mehta

Software Delivery Manager

Result / Outcome

3X faster proposal automation, 70% reduced administrative effort, and real-time enterprise workflow orchestration

Accelerated Proposal Generation – 3X Faster Document Workflows

Automated proposal generation workflows reduced manual effort by dynamically creating SharePoint structures, populating DOCX templates, and generating AI-driven summaries directly from CRM events.

Reduced Administrative Overhead – 70% Workflow Automation Efficiency

Unified CRM operations, document handling, project onboarding, and enterprise notifications into a conversational interface, minimizing repetitive administrative tasks.

Improved AI Session Performance – Stable Long-Running Conversations

Implemented tiered memory management, session summarization, and Azure-backed persistence to maintain fast response times and predictable token usage during multi-hour AI sessions.

Enhanced Document Intelligence – High-Accuracy Multi-File RAG

Enabled highly accurate conversational retrieval across multiple uploaded documents through session-scoped indexing, hybrid ranking, and parallel retrieval strategies.

Preserved Enterprise Document Branding – 100% Design Integrity

Delivered XML-level document editing workflows that maintained original fonts, layouts, tables, spacing, and branding while allowing AI-generated content updates.

Seamless Enterprise Authentication – Uninterrupted OAuth Workflows

Integrated secure in-chat OAuth handling and MCP execution to enable smooth authentication across HubSpot and enterprise services without disrupting conversations.

Real-Time Enterprise Collaboration – Automated Teams Notifications

Enabled proactive Microsoft Teams communication with Adaptive Cards, workflow alerts, project provisioning updates, and deep-linked enterprise actions.

Scalable Enterprise Architecture – Ready for High-Concurrency Operations

Built a decoupled, asynchronous architecture capable of supporting concurrent AI sessions, enterprise integrations, RAG indexing, and workflow automation at scale.

Ready to Start Your Digital Transformation Journey

More Case Studies

A conversational multi-agent HR platform with dynamic privacy

Client

Enterprise SaaS

United States

Case Study

Transforming Edge AI Commerce Through Tailored ERP Solutions