The Problem
POs and SMs spend significant time on operational tasks: searching for historical decisions, reviewing documentation, preparing refinement sessions, and maintaining progress transparency. While these activities are essential, they often reduce the time available for strategic thinking, product development, and team facilitation.
Traditional workflows assume that humans manually maintain context and track information across tools. As projects grow, this leads to inefficiencies: knowledge becomes fragmented, onboarding slows down, and valuable insights are difficult to retrieve when needed. The question becomes whether AI can reduce this friction without introducing new complexity.
Research Question
Our research focused on a central question:
To what extent can AI Agents meaningfully reduce operational overhead for POs and SMs without compromising decision quality?
We explored whether AI could enhance existing workflows by improving access to information, automating repetitive activities, and supporting Scrum events in real time.
Experiment
We introduced an AI Agent into existing team workflows, designed to assist with semantic search, document retrieval, and automated reporting. The AI Agent analyzed tickets, documentation, and historical project data to provide contextual insights during daily work.
Key capabilities included:
- Semantic search across tickets and documentation
- Automated scheduled reports to improve transparency
- Context-aware suggestions during refinement sessions
- Support in locating past decisions and related requirements
- Recommending tickets for sprints
- Assigning and commenting on tickets
- Writing/updating tickets and documents
The experiment focused on integrating AI into existing processes rather than creating entirely new workflows.
Results
Three key insights stood out:
- Reduced time spent searching for information
- POs and team members reported spending less time looking for historical documents and past decisions related to new requirements. Access to contextual knowledge became faster and more intuitive.
- Prompt quality influenced adoption
- The effectiveness of the AI Agent depended heavily on how clearly prompts were formulated. Unclear prompting sometimes led to skepticism about the tool’s value, highlighting the importance of training and clear usage patterns.
- Structured documentation amplified impact
- Teams with well-structured documentation benefited the most. However, initial configuration and fine-tuning were necessary to achieve meaningful results.
What This Means for Our Work
Introducing AI Agents into Scrum workflows changes collaboration dynamics in several ways:
- Improved transparency: Automated reports provide consistent insights into work progress and completion.
- More focus on product delivery: By reducing administrative overhead, teams can allocate more time to building increments rather than maintaining documentation.
- More effective refinement sessions: Semantic search enables faster preparation and clearer context during discussions.
Importantly, the experiment reinforces that AI Agents enhance (not replace) human roles. Contextual judgment, accountability, and decision-making remain core responsibilities of Scrum roles.
At the same time, this challenges existing assumptions. AI is often seen as lacking contextual awareness, yet when properly configured, it can meaningfully support decision-making processes. However, inaccuracies still occur, especially when prompts are unclear or input data contains noise.
Rethinking Operational Work in Scrum Teams
This experiment shows that AI Agents can significantly reduce operational overhead and improve workflow transparency when integrated thoughtfully into Scrum environments. Success depends on structured documentation, clear prompting, and human oversight. Rather than simply inserting AI into existing processes, the real opportunity lies in rethinking workflows around AI-supported collaboration. Freeing teams to focus on delivering value while maintaining strong human decision-making at the core.
