Split & Conquer AI: Two ARTs for more flow and focus

The Digital Business Unit (DBU) faced a noticeable problem in the 10th Program Increment (PI): A single Agile Release Train (ART) was oversized with around 140 employees. This led to the following challenges:

• High operational complexity

• Bottlenecks due to cross-team dependencies

• Delayed decision making

• Pressure on the flow due to different team and architecture logics

• WSJF prioritization came up short

• High time expenditure for substantive debates, clarification of technical questions, coordination agreement and change of context

• Support for the participatory design of the two new ARTs

• Facilitation of a workshop for the development of a split plan and elaboration of the role and team assignment

• Establishment of peer review instead of top-down assignment of responsibility

• Help with storytelling about the new organizational structure

• WSJF decisions faster and data-based; significantly fewer escalation loops.

• Rollover rate has dropped noticeably.

• Planning and delivery are more closely interlinked.

• Clear responsibilities per domain and team; fewer votes across many teams.

• Critical dependencies have been reduced: from 47 to 15 – Flow is no longer slowed down by couplings.

• More efficient PI planning: Total duration and coordination effort have decreased noticeably (e.g. 8 h → 5 h).

• Key performance indicators improved: approx. +12% velocity, approx. −30% lead time after the first joint PI.

• Sustainable flow: Value stream-oriented cutting prevents overload and keeps structures lean.

• Scalable blueprint: The dual-ART model serves as a template for future growth and similar reorganization steps.

• Data-driven decisions: AI-supported analyses as standard – greater transparency and acceptance of change.

• Reduction of resistance overall and emergence of new, accelerated acceptance in change processes