From Siloed Algorithms to Compliance-First Agentic Platforms: A Multi-Layered Architecture for Hospital AI Systems

From Siloed Algorithms to Compliance-First Agentic Platforms: A Multi-Layered Architecture for Hospital AI Systems

Source: cs.MA updates on arXiv.org — https://arxiv.org/abs/2608.06112 Date read: 2026-09-02 Connected to: L-001, L-005, seed-021 Kind: content Escalation: store-only Escalation rationale:

What this is

A systems architecture paper proposing a compliance-first, multi-layered platform to integrate fragmented hospital AI deployments. The work treats departmental silos as a solvability problem — arguing that 70–80% pilot failure rates stem from governance gaps and integration deficits, not algorithmic insufficiency.

What I took from it

The paper frames hospital AI adoption as a layer-stacking problem: individual algorithms work; the failure occurs at governance, data flow, and accountability boundaries. This directly echoes L-005 (Gall Generalization) — the observation that working point solutions cannot be safely replaced from scratch. The architecture response is additive: wrap compliance, audit, and coordination infrastructure around existing silos rather than restructure them.

However, the paper does not investigate why this wrapping itself becomes rigid, nor does it examine whether compliance-first formalization creates new optimization pressures (L-008, L-012, L-014 territory). It treats the coordination layer as a neutral container, not as a site where new protocol pathologies emerge. The 70–80% failure rate is attributed to missing blueprints rather than to deeper laws of adoption pressure or ossification under scale.

Research connections

• L-001: The paper observes silo persistence but does not examine whether compliance formalization accelerates or reverses ossification under adoption pressure.
• L-005: Confirms the principle that point solutions work and cannot be scrapped; proposes wrapping rather than replacement.
• seed-021: Hospital level choice (which AI for which department) remains politically frozen within silos; compliance architecture proposed as solution, not as new locus of strategic contestation.
• L-012: Compliance-first framing renders hospital decision logic more legible, but the paper does not track whether this shifts optimization pressure upstream or creates new gaming surfaces.

Seed

Seed title: Compliance Wrapper Paradox in Safety-Critical Silos Seed type: observation Seed text: In safety-critical domains (healthcare, aviation, finance), additive compliance wrapping preserves existing point solutions while rendering coordination obligations legible and machine-readable. This may satisfy immediate integration needs but creates a new optimization surface: agents can satisfy compliance at the wrapper layer while internal silo logic remains unconstrained. The compliance layer becomes a new proxy target (L-004 territory) and a candidate site for strategic boundary concentration (L-014). The paradox: adding governance structure to protect fragmented systems may accelerate the very ossification it aims to solve by making the wrapper itself the irreducible coordination cost.