AI-mediated organizational networks

AI-mediated organizational networks tracks how AI changes the human-human interaction structure of work: who collaborates, who becomes a knowledge source, which ties become cheaper, and where new coordination burdens appear. The point is not only that AI helps individuals perform tasks; it is that AI can alter the social graph through which organizations move knowledge, translate expertise, and coordinate work.

2007-serenko-organizational-size-knowledge-flow gives this topic a pre-AI baseline: organizational units tend to lose internal knowledge flow as they grow because formal structure becomes more complex, interpersonal trust weakens, communication becomes less frequent, and employees doubt that shared knowledge will reach the people who need it. That makes AI-mediated networks theoretically important: if grounded GenAI lowers search, translation, routing, and connective-efficacy costs, it could weaken a classic size penalty rather than merely speeding up individual knowledge work.

2026-buechsenschuss-genai-organizational-networks gives the wiki its first direct field-experimental evidence for this network-rewiring claim. In a European technology services firm, 42 teams were randomized to receive a grounded GenAI assistant or continue without it; after three months, employees with access to the tool had larger increases in collaboration-network and knowledge-sharing-network centrality, as well as higher satisfaction with knowledge access and more projects executed. That shifts the adoption question from “did workers use AI?” to “did AI change the pattern of human dependence and exchange around the work?”

The source’s mechanism is useful for work design: grounded GenAI can act as a translator, lowering the cost of collaboration across expertise boundaries, and as a knowledge catalyst, making AI-augmented employees more valuable sources for others. That means AI can increase human-human ties rather than simply replacing them. It also means organizational value may appear as changed brokerage, centrality, and knowledge flow before it appears as a clean job-level automation or substitution outcome.

Role differences matter. 2026-buechsenschuss-genai-organizational-networks found specialists gained more in knowledge-network in-degree than generalists, while generalists gained more in satisfaction and project output. In I-O terms, the same AI resource may change different roles through different pathways: for specialists, by making deep expertise more accessible and sought after; for generalists, by improving coordination and throughput across a broader project portfolio.

2026-shi-when-bots-join-the-team extends the network-rewiring question to open-source projects, where rule-based bots are visible participants in a shared interaction record rather than private individual tools. Across 2,991 projects and 12-month windows around first bot adoption, repeated human engagement rose from 0.224 to 0.400 and bot-directed social memory rose from 0.016 to 0.062, while conflict cascades fell; these are precisely timed associations, not causal estimates, because the study has no untreated comparison group. Its joint models further associate repeated engagement, bot-to-human addressing, and role differentiation with fewer conflict cascades, suggesting that bots may change organizational coordination partly by stabilizing recurrent human ties and making interaction history addressable through the bot.

2025-hautamaki-fully-leveraging-ai-b2b-sales adds qualitative evidence on the managerial work behind AI-mediated knowledge flow. In B2B sales organizations, managers used social capital to make data a common language, frame AI as an analytical collaborator, surface use cases in team routines, and connect sales with IT, data, HR, finance, and subject-matter experts. That is not network measurement evidence, but it explains a plausible management mechanism by which AI changes who needs to collaborate with whom.

Connections

• Connects to ai-mediated-teamwork because team benefits and network benefits are adjacent but not identical: teams are bounded units, while organizational networks capture emergent ties across roles, groups, and expertise boundaries.
• Connects to work-redesign because AI can reorganize collaboration and knowledge-flow architecture without formally changing job descriptions first, and because unit size and cross-unit exchange are design variables when organizations try to preserve knowledge flow at scale.
• Connects to human-ai-collaboration because the human-AI interface can change the value and frequency of subsequent human-human collaboration.
• Connects to ai-mediated-work-experience because becoming more central can bring satisfaction, empowerment, overload, coordination fatigue, or pressure to be constantly available as a knowledge source.
• Connects to ai-workforce-impact-measurement because productivity-only impact measures can miss changes in network centrality, knowledge access, and interaction structure.
• Connects to ai-adoption and ai-readiness because cross-functional knowledge transfer is both an adoption mechanism and a readiness capability when AI use depends on domain data, IT infrastructure, and shared use-case learning.

Contradictions & open questions

• 2026-buechsenschuss-genai-organizational-networks supports network expansion after grounded GenAI introduction, but its network outcomes are self-reported and observed in one firm over three months; digital-trace and longer-term evidence are needed before treating network rewiring as a general adoption effect.
• 2007-serenko-organizational-size-knowledge-flow is a conceptual pre-AI theory rather than a direct test, so the useful question is not whether the 150-person threshold still holds literally; it is whether AI changes the mechanisms that made size costly: trust, communication frequency, connective efficacy, routing, and cross-unit reach.
• Increased centrality is not automatically good. The same mechanism that improves knowledge flow can concentrate requests on key people, create coordination fatigue, or make specialists more interruptible.
• The source’s role comparison is theoretically useful but not causal because employees were not randomly assigned to specialist versus generalist roles; role-based redesign claims should stay tentative until tested across organizations and task structures.
• 2025-hautamaki-fully-leveraging-ai-b2b-sales gives manager-reported mechanisms for cross-functional AI knowledge transfer, but not observed network data. It should be read as a qualitative mechanism source, not as evidence that AI actually increased network centrality, tie strength, or collaboration volume.
• 2026-shi-when-bots-join-the-team supplies digital-trace evidence from thousands of open-source projects, but its before/after event-study design lacks an untreated comparison group and studies predictable rule-based bots in established repositories. Whether the associations with repeated engagement and bot-directed social memory persist as durable human-human network structure, generalize to other workplaces, or hold for more capable agents remains open.