Dragos Dumitriu was the engineering manager of Microsoft's XIT (Extended IT) Sustaining Engineering team who, in October 2004, requested Anderson's help designing a pull system for his team's work. This request initiated the microsoft-xit-kanban-2004 experiment — the first virtual kanban system for knowledge work — and set in motion the intellectual journey that led to the Kanban Method.
The October 2004 request
Dumitriu's team faced a chronic problem common to software sustaining engineering: unpredictable work intake, an offshore development team in India whose capacity was difficult to manage, and a backlog of change requests that grew faster than the team could process them. Dumitriu recognized that the team needed a pull system — a mechanism that would limit the work flowing to the offshore team to what the team could actually handle — and brought Anderson in to design one. The framing as a "pull system" rather than a "project plan" or "process improvement" was significant: it meant Anderson approached the problem through a flow management lens from the outset.
The first implementation
Working with Dumitriu, Anderson designed a system using Microsoft Product Studio (the internal work management tool) to create a virtual WIP-limited pull system. The design was inspired by Goldratt's Drum-Buffer-Rope mechanism from Theory of Constraints — the offshore team's capacity was the drum, a buffer of approved work items sat upstream, and a WIP limit (the rope) constrained how many items entered active development simultaneously. Physical kanban cards were not used; the workflow was managed entirely through a software tool. This virtual implementation was a key innovation: it demonstrated that the kanban concept could be applied to knowledge work without the physical infrastructure of Toyota's manufacturing system.
Significance in the intellectual biography
Dumitriu's role is primarily that of the catalyst who created the practical conditions for Anderson's first kanban experiment. Without the specific request and the specific problem of managing an offshore sustaining engineering team, it is unclear whether Anderson would have implemented a virtual pull system at this time. The Microsoft experiment also gave Anderson empirical data — the team's performance improved measurably — that grounded the theoretical claims he would later develop with don-reinertsen's queueing theory framework. Dumitriu appears in the kanban-book as the initiating figure of the story Anderson tells about the method's origin.