When Coordination Is Avoidable: A Monotonicity Analysis of Organizational Tasks
Source: https://www.semanticscholar.org/paper/09497b3f05849d8ba740ef226bead8a00ce4335c ↗
Full text: open-access via OpenAlex ↗
Ju takes a question that organization theorists have circled for decades — when does coordination actually change the outcome rather than merely consuming resources? — and gives it a formal answer borrowed from distributed systems.
The key move is mapping Thompson's 1967 taxonomy of task interdependence onto monotonicity, the distributed-systems criterion for whether new information can invalidate prior conclusions; if a task is monotonic, no agent needs to wait for another before acting.
The empirical sweep across 65 APQC workflows and 13,000+ O*NET occupational tasks is unusually grounded for a theoretical contribution, and the finding that 24–57% of coordination spending may be structurally unnecessary reframes the perennial complaint about meeting overhead as a solvable design problem.
For product directors running multi-disciplinary teams or managing AI pipelines, the monotonicity criterion offers a precise diagnostic: decompose work into units, test for non-monotonicity, and treat coordination as a tax that should be levied only where logically unavoidable.
The paper is strongest as a bridge theorem — connecting organizational theory, formal methods, and applied AI governance — rather than as a standalone empirical study, which gives it durability beyond its 2026 data.
Central argument
Ju argues that a significant portion of organizational coordination is structurally unnecessary for correctness — not merely inefficient, but categorically avoidable. By mapping Thompson's classic interdependence taxonomy (pooled, sequential, reciprocal) onto the CALM theorem from distributed systems, he proves that tasks with monotonic specifications — where new information never invalidates prior conclusions — can execute correctly without coordination. Applying this framework to 65 APQC enterprise workflows and 13,417 O*NET occupational tasks, he finds that 74% and 42% respectively are monotonic, suggesting that uniform coordination regimes waste 24–57% of coordination spending on tasks that don't require it for correctness.
Critique
The framework's practical bite depends heavily on how tasks are decomposed and scoped before classification — a pooled task at one level of abstraction may embed reciprocal dependencies at a finer grain, making the monotonicity verdict sensitive to analyst choices that the paper delegates partly to a calibrated LLM annotator. More fundamentally, the Coordination Tax is explicitly bounded to correctness under formal assumptions (deterministic evaluation, reliable delivery, eventual consistency) that rarely hold cleanly in human organizations, where coordination also manages ambiguity, political alignment, and shared sense-making that the model deliberately excludes. The gap between 'unnecessary for formal correctness' and 'safe to eliminate in practice' is therefore larger than the quantitative estimates imply.
Why it matters for product
For a CPO, this work offers a principled basis for challenging coordination defaults baked into team topology and delivery processes — specifically, it suggests that many cross-functional syncs, review gates, and alignment rituals surrounding pooled or non-retractive sequential work (content production, independent feature development, parallel discovery streams) impose real costs without correctness benefits. The distinction between additive feedback (single-loop refinement, monotonic) and retractive feedback (double-loop revision, coordination-required) is directly actionable when designing review and iteration cycles: stages where teams genuinely revise prior outputs based on downstream input warrant structured coordination, while refinement stages do not. As AI agents take over more workflow steps and coordination costs become token-countable, this framework also gives product leaders a concrete lens for auditing which parts of their multi-agent or human-AI pipelines are paying a coordination tax they didn't consciously choose.