Shared Repertoire

In the AI Story

The Xerox technicians' repertoire included not just repair techniques but stories about unusual failures, implicit standards for what a 'tricky machine' felt like, and the collective judgment about which diagnostic shortcuts to trust. The repertoire lived not in any individual but in the practice — maintained through breakfast conversations, calls from the field, the constant circulation and recombination of shared experience.

The software development team's repertoire includes coding standards that started as explicit rules and became implicit habits, architectural patterns no one documented but everyone follows, debugging techniques passed along through production-incident stories, and a collective sense of 'good code' that no style guide fully captures. The repertoire is what makes a senior engineer's judgment more than the sum of her formal knowledge.

The repertoire is actively maintained through participation. Each code review is a negotiation of standards. Each design debate is a refinement of shared sensibilities. Each war story deposited in the community's collective memory extends the repertoire. When the interactions that maintain the repertoire are redirected to AI tools, the repertoire stops being actively maintained — and what is not maintained gradually ossifies into inherited assumption that no one remembers how to defend.

In the Trivandrum scene described in this volume's foreword, a senior engineer's correction of a junior's AI-assisted decision — 'the one you picked will break under load; we tried it eighteen months ago on payments and it took down production for six hours on a Saturday night' — is exactly a repertoire transmission event. The knowledge lived in one person's memory, connected to a specific production incident, embedded in a community's shared understanding of what that Saturday night cost. No AI training set contained it. No documentation captured it. The eight-second interaction in a stand-up meeting transmitted it.

Origin

Wenger introduced the concept in Communities of Practice (1998) as the third constitutive element of community alongside shared domain and mutual engagement. It drew on traditions in the sociology of science (Collins, Latour) that had documented how scientific communities maintain tacit knowledge through ongoing practice rather than through codification.

The framework has proven especially consequential in organizational knowledge management, where attempts to capture repertoire in documents have repeatedly failed. Organizations discovered that what could be documented was only the reified surface; the deeper repertoire resisted extraction because it was constituted through use, not through storage.

Key Ideas

Includes both tangible and tacit resources. Procedures and sensibilities, vocabulary and judgment, stories and standards.

Maintained through use. Repertoire persists because the community actively negotiates and renegotiates it, not because it has been recorded.

Exceeds individual knowledge. The repertoire is distributed across the community; no single member holds all of it.

Transmitted through participation. New members acquire the repertoire through sustained engagement, not through manuals.

Ossifies when unmaintained. When the participatory interactions that sustain it cease, repertoire degrades into inherited assumption.

Debates & Critiques

A persistent question is whether AI systems can be trained on a community's repertoire and then serve as repertoire reservoirs for the community's ongoing work. Some organizations have experimented with training custom models on internal documentation, code, and communications. The results suggest that while surface patterns can be captured, the repertoire's deeper dimensions — its implicit standards, its connection to specific shared experiences, its role in identity formation — resist this kind of extraction.