Architecting an AI Assistant for Conversational Reporting in Education Platforms

Background

iCEV has supported Career and Technical Education (CTE) programs for more than 40 years. Its online platform provides curriculum across fields such as agricultural science, health science, business, and law enforcement. Through multimedia lessons, customizable courses, and industry-recognized certifications, iCEV helps educators prepare students for real-world careers.

As interest in AI-powered assistants increased across education technology, iCEV began exploring how conversational AI could help educators and administrators interact more naturally with the platform’s reporting tools, data, and support resources.

Challenges

iCEV wanted to evaluate whether an AI assistant could help educators and administrators interact with the platform through natural language questions. Beyond accessing complex reporting data, the goal was to explore how conversational AI could assist users across the broader platform experience—from discovering reports and applying filters dynamically to helping educators find answers within the platform’s support knowledge base.

The assistant prototype also integrated with iCEV’s support documentation, allowing educators to receive immediate guidance drawn directly from existing help resources. In scenarios where additional assistance was needed, the system included a mock integration with the support ticket workflow so users could quickly escalate issues without manually completing lengthy support forms.

Any solution also needed to align with iCEV’s existing technology ecosystem. The initiative required working within their Azure environment and integrating securely with existing identity and access controls so that AI-driven queries respected the same permission boundaries as the platform’s reporting tools.

To explore these questions, iCEV partnered with IntelliTect to design and implement a focused architectural prototype.

Solutions

IntelliTect developed a C# chat agent application that demonstrated how an AI assistant could operate within iCEV’s Azure-based ecosystem. A lightweight command-line interface was used for interaction so the team could focus on architecture and experimentation rather than building a full production user interface.

Key elements of the solution included:

1. Agentic AI architecture
   Semantic Kernel orchestrated interactions between Azure OpenAI, application logic, and discovery tools so the assistant could identify available reports, apply filters dynamically, and route requests to the appropriate platform capabilities.
2. Secure system integration
   Entra ID authentication ensured the assistant respected existing user permissions when accessing reports, data, and platform resources.
3. Knowledge base and support integration
   The assistant connected to iCEV’s library of support articles, allowing it to answer common questions and guide users through platform workflows. When issues required additional assistance, the system included a mock integration with the support ticket process so users could escalate requests directly through the conversational interface.
4. Cloud-based AI experimentation
   The team explored advanced scenarios such as uploading CSV datasets and allowing the AI to generate Python code that executed in a secure sandbox environment to perform joins and analysis.
5. AI-aware testing approach
   Because AI outputs are not deterministic, IntelliTect implemented AI-graded integration tests that validated whether responses met expected behavioral criteria.

This architecture demonstrated practical patterns for secure, permission-aware AI integration within iCEV’s C# and Azure environment.

Working with IntelliTect helped us explore how conversational AI could enhance the way educators interact with our platform’s reporting tools. Their team quickly built a working prototype that clarified what is possible with modern AI and how it can integrate securely with our existing technology stack.

Steven Lubowicz, Chief Technology Officer, iCEV

Outcome

The proof of concept showed that natural language can be an effective interface for tasks such as report discovery, filter selection, and support guidance when the assistant is grounded in platform-specific tools and data. It also helped the team identify where additional constraints, orchestration logic, and evaluation methods would be needed for more complex scenarios.

The project provided iCEV with a working reference architecture for integrating Azure OpenAI into their software solution. By exploring multiple architectural approaches and benchmarking AI behaviors, the initiative produced clear insights that will inform future AI-powered features within the iCEV platform.