AI Agents for Value Stream Mapping

It changes what they're for. The current-state mapping — which used to consume most of the workshop — is generated by the agent in advance and shows up in the room as a starting point. Workshops become focused on judgment, prioritization, and change management, which is what they should have been about in the first place.

That's actually where agents do best. Multi-system value streams — order management to billing to fulfillment to support — are exactly the ones where humans struggle to draw the map manually because the data lives in five different tools. Agents stitch the case across systems using shared keys, time correlation, or causal inference when keys are missing.

Yes — most of our VSM engagements are in service operations. Loan origination, claims processing, hiring pipelines, customer onboarding, and clinical pathways all map cleanly. The agent uses time-in-step instead of inventory boxes, but the core logic — lead time, value-add ratio, waste classification — translates directly.

Useful, with bounds clearly stated. The simulator models the historical event log forward under the proposed intervention. It captures first-order effects (removing a step) and second-order effects (reducing a wait) but cannot predict third-order effects like behavior change or capacity rebalancing. Output always includes a confidence interval and a list of assumptions for the human to challenge.

Standard VSM iconography — process boxes, inventory triangles, push/pull arrows, supplier/customer endpoints, data boxes, timeline at the bottom — rendered from current data. Hovering a step shows the cycle time distribution; clicking opens the variant analysis. Exportable to PNG/SVG for the kaizen room wall when needed.