Proof of Concept (PoC) for Enterprise AI Copilot Accessible via Microsoft Teams

Summary

A US real estate investment company wanted to know whether AI could mine its knowledge base of internal processes and generate intelligent recommendations to improve them. INNERLUXES ran an eight-day proof of concept that compared several AI services, proved the idea feasible, and delivered an optimal tech stack — Microsoft Copilot integrated with Microsoft Teams and Azure AI services — together with a high-level Azure architecture to build on.

About the Client

The Client is a US real estate investment company that focuses on healthcare facilities.

Verifying Whether AI Could Optimize Business Processes

The Client maintains an extensive knowledge base covering its internal processes — such as employee onboarding and investment portfolio management — across diverse file formats including PDF and XLSX. It wanted to put that accumulated knowledge and best practice to work through a custom AI tool that would analyze the process data and generate intelligent recommendations for improvement, and engaged INNERLUXES's AI engineering services to verify whether the idea was feasible.

Discovery and Requirements for an Enterprise AI Copilot

INNERLUXES assembled a team of a business analyst, two .NET developers, and a solution architect, and ran a series of Q&A sessions to understand the Client's expectations. In the Client's vision, the solution would scan process-specific files (onboarding procedure documents, investment portfolio reports) and generate recommendations for improving the related processes, each including step-specific roles and timelines, suggested KPIs, and references to the documents the AI used. The Client's subject matter experts would then review the relevance of the AI output and update the knowledge base with a meeting recap, and the AI would learn from that feedback to refine future recommendations. Given the solution's innovative nature and potential business impact, the Client and INNERLUXES agreed to start with a proof of concept to test the AI's ability to generate relevant recommendations — leaving the human-feedback learning loop for the full-featured version.

Comparing AI Services for Document Analysis and Recommendations

To choose the optimal AI services, INNERLUXES ran a test comparing four options:

• Claude Sonnet (suggested by the Client).
• Azure AI Services (Azure OpenAI + Azure AI Search).
• Microsoft Copilot Studio integrated with Azure OpenAI.
• Microsoft Copilot integrated with Microsoft Teams and Azure OpenAI + Azure AI Search.

As the test knowledge base, the team used four HR PDFs describing procedures with step sequences, references to laws and regulations, and screenshots, all uploaded to Microsoft Blob Storage as the common data source. Using the same prompts across all four options, INNERLUXES compared output relevance and accuracy, the ability to adjust generated content, future model-customization capabilities, and the visual presentation of the output. Microsoft Copilot — both on its own and integrated with Teams — produced the best results, while the other options had notable drawbacks, such as needing to convert Excel files to PDF and giving poor responses to improvement requests. Because the Client already uses Microsoft Teams as its primary business messenger, INNERLUXES recommended Microsoft Copilot integrated with Teams, so employees can send prompts and receive AI responses right inside Teams.

An Azure-Based Architecture Accessed Through Teams

INNERLUXES designed a high-level solution architecture with these components:

• Microsoft Teams as the main user interface.
• Azure Web App, a custom application for data orchestration that automates repetitive data search and analysis and coordinates the other components.
• Azure OpenAI to host and run the AI model.
• Azure AI Search to scan and index the knowledge base and serve data to Azure OpenAI for analysis.
• Azure Blob Storage to hold the knowledge-base files, with dedicated containers per business process so the AI processes only relevant data rather than the whole database.
• Azure Key Vault to store the access keys to the AI services.

A Proof of Concept Delivered in Eight Days

With INNERLUXES's help, the Client proved that its idea to optimize business processes with AI was feasible. After the tests, the Client had an optimal tech stack — Microsoft Copilot integrated with Microsoft Teams and Azure AI services — that gives the most accurate improvement recommendations, lets users adjust the AI output, and fits neatly into established workflows, plus a high-level architecture with dedicated components for data orchestration and security. The proof of concept gives the Client a solid foundation for further development.

Technologies and Tools

Azure OpenAI, Azure AI Search, Microsoft Copilot Studio, Microsoft Teams, Azure Blob Storage, Azure Key Vault, Claude Sonnet, .NET.