AI agents · RAG · Master's thesis
ScAi: an AI assistant for sprint reviews
Sprint data is rich but slow to read. ScAi uses specialised AI agents to analyse it, so Scrum teams spend their review time on decisions, not spreadsheets.
The problem
Agile teams generate a lot of Sprint information. Analysing it by hand before a review is slow, and it's easy to miss the patterns that matter: a risk building up, a team dynamic, a process that keeps costing time.
What I built
ScAi is an AI-driven assistant for SAP agile software development teams. It combines retrieval-augmented generation (RAG) and a multi-agent system to turn Sprint Backlog data into actionable insights. Four specialised agents each take one job:
- Sprint Analysis: what happened in the sprint.
- Risk Detection: what could derail the next one.
- Team Dynamics: how the team is working together.
- Process Improvement: what to change next.
Design principle
The assistant supports Scrum teams without replacing human decision-making. It surfaces insights; the team decides.
The prototype connects Agile project data with AI-powered analysis. I'm extending it with MCP-based tool integration and agentic workflows, to see how AI assistants can work with enterprise tools and external data through structured tool calling.
How I'm evaluating it
The thesis uses a Design Science and Case Study approach. It combines building the system with qualitative research involving Agile and SAP practitioners, to judge how useful it is in practice and where AI can empower teams.
Status: in progress. The practitioner evaluation is part of the research, so I'm not quoting results yet.