Paper Number
ECIS2026-2216
Paper Type
CRP
Abstract
Generative, agentic AI promises to accelerate venture learning, yet we lack concrete designs for embedding them into entrepreneurial experimentation. This design science study proposes a multi-agent artefact that operationalises the Build–Measure–Learn (B-M-L) cycle as a closed-loop control system. Drawing on the Dynamic Capabilities View, we derive fifteen meta-requirements and thirty-three design principles (consolidated into seven goal-directed groups) for sensing, seizing, reconfiguring, orchestration, and governance. We instantiate them in a Node.js package instrumenting a production-grade SaaS codebase. Controlled simulations compare agentic and manual B-M-L cycles on feature ideas. The Multi Agent System reduces time-to-validated-learning by roughly an order of magnitude while preserving statistical rigour, traceability, and nuanced Persevere/Iterate decisions. Logs render capabilities observable at the feature level, turning “agentic AI” into a disciplined experimentation infrastructure rather than a generic assistant. We discuss implications for IS design and future field evaluations.
Recommended Citation
Jelinek, Elias and Rothe, Hannes, "Multi Agent Systems In The Lean Startup Cycle: Operationalising Dynamic Capabilities" (2026). ECIS 2026 Proceedings. 6.
https://aisel.aisnet.org/ecis2026/entmodel/entmodel/6
Multi Agent Systems In The Lean Startup Cycle: Operationalising Dynamic Capabilities
Generative, agentic AI promises to accelerate venture learning, yet we lack concrete designs for embedding them into entrepreneurial experimentation. This design science study proposes a multi-agent artefact that operationalises the Build–Measure–Learn (B-M-L) cycle as a closed-loop control system. Drawing on the Dynamic Capabilities View, we derive fifteen meta-requirements and thirty-three design principles (consolidated into seven goal-directed groups) for sensing, seizing, reconfiguring, orchestration, and governance. We instantiate them in a Node.js package instrumenting a production-grade SaaS codebase. Controlled simulations compare agentic and manual B-M-L cycles on feature ideas. The Multi Agent System reduces time-to-validated-learning by roughly an order of magnitude while preserving statistical rigour, traceability, and nuanced Persevere/Iterate decisions. Logs render capabilities observable at the feature level, turning “agentic AI” into a disciplined experimentation infrastructure rather than a generic assistant. We discuss implications for IS design and future field evaluations.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.