AI-Migration-Advisor

AI Migration Advisor is a domain-aligned GenAI prototype that analyzes enterprise data and analytics migration scenarios provided in natural language. It converts a migration scenario into a structured preliminary assessment covering migration complexity, complexity drivers, key migration risks, important assessment areas, questions for the customer, recommended next steps, and information requiring validation. The current prototype uses a locally running open-source LLM and is designed as an AI-assisted assessment tool rather than an autonomous migration decision engine.

The prototype was tested across four scenarios covering valid migrations, cross-platform BI migration, incomplete information, and an intentionally questionable migration path. Test results showed three passed and one failed. The failed test involved an Azure HDInsight to Power BI scenario, where the LLM generated an assessment instead of questioning whether Power BI was an appropriate direct target for HDInsight workloads. A key finding was that prompt engineering can guide model behaviour, but it cannot guarantee domain-grounded technical correctness.

The application is built with a Streamlit UI, Python application controller, structured prompt, Ollama local model runtime, and Qwen language model. Key features include free-text migration input, structured AI assessment, prompt constraints that prevent inventing timelines, costs, SLAs, migration utilities, or unsupported technical information, local LLM execution, and human validation. The prototype uses an ungrounded general-purpose LLM, does not use RAG or an enterprise migration knowledge base, cannot guarantee architectural correctness, requires human validation, and is not intended for production migration decisions.