Paul Bryant

Paul Bryant is a Principal Multicloud Architect at Dell Technologies and the author of Digital Thought Disruption. He is an eight-time VMware vExpert and Microsoft Azure-certified architect focused on VMware Cloud Foundation, Azure Local, NVIDIA accelerated infrastructure, enterprise AI, hybrid cloud, networking, automation, governance, and resilient platform operations. His work includes technical whitepapers and DOI-registered publications on Zenodo, industry interviews and speaking engagements, and articles cited by peer-reviewed research, industry reports, and as non-patent literature across seven issued U.S. patents. He translates complex infrastructure and AI requirements into practical enterprise architectures and operating models.

VCF 9.1.1 assurance separates VKS agent workloads and shared models from policy, scoped NSX execution, independent verification, and protected evidence.

Independent Agent Assurance on VMware Cloud Foundation 9.1.1

Published September 19, 2026 by Paul Bryant

Apply independent agent assurance to VMware Cloud Foundation 9.1.1. Separate agent workloads from authorization, scoped execution, verification, and evidence.

AI proposals and human policy meet at an execution gate, followed by a scoped executor, target system, and independent verification.

The Architecture That Keeps AI From Authorizing Itself

Published September 19, 2026 by Paul Bryant

Design an AI agent authorization architecture that separates proposals, policy, execution, and evidence, including indirect paths that can bypass approval.

Intent, authority, execution, and independent target-state observation feed protected action evidence with verified or unresolved outcomes.

The Agent Action Evidence Contract: What Every AI Action Must Record

Published September 19, 2026 by Paul Bryant

Define a proposed agent action evidence contract that connects identity, approval, execution, independent observation, and recovery while preserving unresolved outcomes.

The Assurance Independence Model tests six dimensions against a defined action and failure scenario, then applies gates for hold or bounded release.

The Assurance Independence Model: Six Boundaries for Agentic AI

Published September 19, 2026 by Paul Bryant

Apply the proposed Assurance Independence Model to six trust boundaries. Assess shared failures, require evidence, and use mandatory gates before expanding agent authority.

An LLM judge supplies findings to an execution gate governed by identity and policy, with target-system outcomes independently verified.

LLM as a Judge: Evaluation Is Not Authorization

Published September 19, 2026 by Paul Bryant

Use LLM-as-a-Judge for scoped evaluation without confusing a favorable score with proof or permission. Keep authorization and outcome verification independent.

AI review shares context and assumptions; independent assurance adds owned policy, external evidence and human stop authority, with an enforcement gate controlling enterprise systems.

Who Audits the AI Auditor? Independent AI Assurance

Published September 19, 2026 by Paul Bryant

Explore six dimensions of independent AI assurance, with controls that separate agent judgment, authorization, execution and evidence, plus a proposed benchmark.

Preferred AI and a qualified alternative meet the same identity, evidence, state, and business-authority requirements, preserving the choice to continue, replace, or retire the platform.

The Enterprise Architect’s Guide to Surviving the AI Power War

Published September 19, 2026 by Paul Bryant

Build an enterprise AI exit strategy with tested model substitution, retained authority, recoverable state, contract rights, and realistic transition costs.

Five possible AI futures, integrated platforms, agents, personal AI, open models, and utility services, connect a 2026 evidence base to enterprise decisions for 2029.

Who Leads AI in 2029? Five Scenarios, Not One Prediction

Published September 19, 2026 by Paul Bryant

Explore five scenarios for AI leadership in 2029, with evidence triggers, counterarguments and a practical method for testing enterprise platform commitments.

Different AI models depend on cloud and inference services, compute, networks, facilities, manufacturing, power, and cooling. The diagram cautions that shared demand does not guarantee supplier profit.

The Companies That Win No Matter Which AI Model Wins

Published September 19, 2026 by Paul Bryant

See how NVIDIA, TSMC, cloud platforms and infrastructure suppliers benefit across AI models, and what their advantages mean for enterprise buying decisions.

Customization, deployment control, and new tasks pass through evidence gates before becoming new enterprise AI choices.

AI Dark Horses: Who Could Change the Competitive Balance?

Published September 19, 2026 by Paul Bryant

Assess AI dark horses including Thinking Machines, SSI, Mistral, World Labs and robotics labs, with evidence gates for enterprise pilots and supplier choices.

Qwen, DeepSeek, Kimi, GLM, and Seed are shown above available compute and an alternative Ascend software stack leading to qualified enterprise service. Model parity does not prove stack independence.

China’s AI Counteroffensive: Model Parity Under Silicon Constraints

Published September 19, 2026 by Paul Bryant

Evaluate Chinese AI models for enterprise use. Compare Qwen, DeepSeek, Kimi, GLM, Seed and Huawei across capability, silicon constraints and deployment risk.

Multiple models feed an agent control plane for identity, memory, tools, and routing. Approval, limits, and stop controls form an authority boundary before access to business systems.

The Agent Control Plane Is the Real Prize in the AI War

Published September 19, 2026 by Paul Bryant

Build an agent control plane strategy that separates provider capabilities from business authority, with clear ownership, approval, recovery and exit tests.

Browsers, work apps, messages, and devices lead to a familiar AI assistant. Permitted context supports useful outcomes, while familiar access does not grant authority.

Distribution May Defeat Intelligence: Google, Meta, Microsoft and Apple’s Hidden Advantage

Published September 19, 2026 by Paul Bryant

Why Google, Meta, Microsoft and Apple can turn devices and everyday apps into an AI distribution advantage, and how enterprises should assess the tradeoffs.

Stargate, Colossus, and Google TPUs depend on power, cooling, networks, and workload validation to provide usable service capacity. Announced scale is not the same as delivered work.

The Compute Arms Race: Stargate, Colossus, TPUs and the Gigawatt Battlefield

Published September 19, 2026 by Paul Bryant

Compare Stargate, Colossus and Google TPUs. Separate AI compute promises from usable capacity with evidence on power, networks, workload fit and recovery.

Changing model leaders and durable controls for identity, memory, tools, and authority connect to an agent control plane supported by applications, cloud, silicon, networks, and energy.

The AI Power Stack: Who Is Winning the AI Race?

Published September 19, 2026 by Paul Bryant

Explore the AI power stack, compare leading ecosystems, and see why compute, distribution, agent control, and exit options matter beyond model benchmarks.

A zero-trust network fabric distributes policy enforcement across a private-cloud city, containing red-marked threats while blue governed workload zones remain connected.

NSX as the Zero Trust Control Fabric: From Central Policy to Distributed Containment

Published September 19, 2026 by Paul Bryant

Translate workload context and security intent into NSX distributed enforcement. Plan microsegmentation, policy ownership, telemetry, exception handling, and validation around realistic containment requirements.

Approved evidence supports AI proposals, external controls authorize scoped actions, and separately permitted knowledge updates require reviewed corrections and observed outcomes.

AI Feedback Loops: Building Systems That Stay Correctable

Published September 19, 2026 by Paul Bryant

Keep AI systems correctable as feedback accumulates. Govern knowledge promotion and withdrawal, protect authorization boundaries, reconcile unknown execution outcomes, and test the conditions that should stop the workflow.

Model signals and current telemetry or source records meet at claim-level validation, producing either a supported answer or further investigation while action authorization remains separate.

AI Uncertainty: Why Confidence Scores Are Not Enough

Published September 19, 2026 by Paul Bryant

Give token entropy, semantic uncertainty, calibration, and abstention distinct jobs. Evaluate how uncertainty changes the next workflow step instead of relying on a single confidence score.

A confident answer undergoes claim-by-claim examination of sources, inventory, tests, and conditions, producing supported, unresolved, or contradicted evidence records before authority is considered.

AI Output Verification: Check Claims Before Acting

Published September 19, 2026 by Paul Bryant

Evaluate infrastructure claims against scoped evidence. Use a two-cluster example to separate supportability, capacity, isolation, and resilience, including the difference between normal and failure-state capacity.