<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Axionalytics</title>
    <link>https://www.axionalytics.com/blog.html</link>
    <atom:link href="https://www.axionalytics.com/rss.xml" rel="self" type="application/rss+xml"/>
    <description>Architecture writing on enterprise agentic AI: deployment topology, execution isolation, human approval gates, data provenance, and analytics governance.</description>
    <language>en</language>
    <lastBuildDate>Fri, 14 Aug 2026 00:00:00 +0000</lastBuildDate>
    <generator>_build/scripts/make-feeds.py</generator>
    <item>
      <title>Deploying an Agent Against Databricks Unity Catalog</title>
      <link>https://www.axionalytics.com/databricks-unity-catalog-agentic-deployment.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/databricks-unity-catalog-agentic-deployment.html</guid>
      <description>Schema injection collapses somewhere past a few hundred tables. A narrow probing tool set, retrieval-driven discovery, a pointer pattern for bulk results — and the three failure modes that only appear once a real catalog is behind it.</description>
      <category>Playbook</category>
      <category>Enterprise Agentic AI</category>
      <pubDate>Wed, 29 Jul 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Connecting an Agentic System to Azure DevOps</title>
      <link>https://www.axionalytics.com/azure-devops-ai-test-case-deployment.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/azure-devops-ai-test-case-deployment.html</guid>
      <description>Tools split by consequence rather than by endpoint, a credential the sandbox never sees, and the two operations — pipeline triggers and stale approvals — that break a naive integration.</description>
      <category>Playbook</category>
      <category>Agentic Test Engineering</category>
      <pubDate>Wed, 29 Jul 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Snowflake, BigQuery, Azure SQL: One Pipeline, Five Sets of Quirks</title>
      <link>https://www.axionalytics.com/cloud-warehouse-governed-bi-deployment.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/cloud-warehouse-governed-bi-deployment.html</guid>
      <description>Role-scoped visibility breaks more dashboards than any other single cause. What the warehouses share, what they do not, and why the profiling step decides your deployment topology.</description>
      <category>Playbook</category>
      <category>Agentic Business Intelligence</category>
      <pubDate>Wed, 29 Jul 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>LangChain and LangGraph in an Enterprise Deployment</title>
      <link>https://www.axionalytics.com/langgraph-langchain-enterprise-deployment.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/langgraph-langchain-enterprise-deployment.html</guid>
      <description>An orchestration library is not an alternative to a governed system — it is a component inside one. Twelve decisions the library leaves to you, and when owning them is the right call.</description>
      <category>Comparison</category>
      <category>Enterprise Agentic AI</category>
      <pubDate>Tue, 28 Jul 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Bedrock Agents and Vertex AI Agent Builder: What a Managed Runtime Decides for You</title>
      <link>https://www.axionalytics.com/bedrock-vertex-managed-agent-runtimes.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/bedrock-vertex-managed-agent-runtimes.html</guid>
      <description>Four architectural decisions you inherit with a managed agent runtime, the one written commitment that rules it out, and why the switching cost is not where people look for it.</description>
      <category>Comparison</category>
      <category>Zero-Trust AI Governance</category>
      <pubDate>Mon, 27 Jul 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Microsoft 365 Copilot and the Work It Does Not Do</title>
      <link>https://www.axionalytics.com/microsoft-copilot-vs-governed-agents.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/microsoft-copilot-vs-governed-agents.html</guid>
      <description>An assistant and a governed execution substrate are not competing for the same budget line. The boundary sits exactly where an action stops being reversible — and inheriting a permission set is not the same as intersecting one.</description>
      <category>Comparison</category>
      <category>Enterprise Agentic AI</category>
      <pubDate>Fri, 14 Aug 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Building It In-House: What the Estimate Leaves Out</title>
      <link>https://www.axionalytics.com/build-vs-buy-enterprise-agentic-ai.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/build-vs-buy-enterprise-agentic-ai.html</guid>
      <description>Many organisations should build this, and the ones who should not are identifiable by one property of the work. The demo is the first fifth; the substrate is the rest, and it cannot be retrofitted onto a prototype that already works.</description>
      <category>Comparison</category>
      <category>Enterprise Agentic AI</category>
      <pubDate>Fri, 14 Aug 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Warehouse-Native Agents and Where the Boundary Bites</title>
      <link>https://www.axionalytics.com/warehouse-native-agents-vs-governed-platform.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/warehouse-native-agents-vs-governed-platform.html</guid>
      <description>They inherit governance rather than asserting it, which is why they win for analytical work inside one platform. They also stop where the warehouse stops, and most enterprise workflows do not.</description>
      <category>Comparison</category>
      <category>Agentic Business Intelligence</category>
      <pubDate>Fri, 14 Aug 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>CrewAI and AutoGen Against Enterprise Constraints</title>
      <link>https://www.axionalytics.com/crewai-autogen-multi-agent-frameworks.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/crewai-autogen-multi-agent-frameworks.html</guid>
      <description>Coordination frameworks solve delegation well. Delegation is also how an injected instruction escalates privilege, and agent count is a design cost rather than a capability.</description>
      <category>Comparison</category>
      <category>Enterprise Agentic AI</category>
      <pubDate>Sun, 26 Jul 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Teardown: The Two-Plane Execution Architecture</title>
      <link>https://www.axionalytics.com/two-plane-execution-architecture.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/two-plane-execution-architecture.html</guid>
      <description>A control plane that holds no customer data, an execution plane inside your VPC, and a hypervisor boundary between agent-generated code and everything else. What each layer defends against, and what the shape costs.</description>
      <category>Teardown</category>
      <category>Zero-Trust AI Governance</category>
      <pubDate>Tue, 28 Jul 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Teardown: The Pre-Execution Write Gate</title>
      <link>https://www.axionalytics.com/write-gate-architecture.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/write-gate-architecture.html</guid>
      <description>Policy the model cannot read, a tool roster compiled per turn, and a rendered blast-radius diff pinned to a version token. Three checks in the execution path, and the throughput ceiling they buy.</description>
      <category>Teardown</category>
      <category>Enterprise Agentic AI</category>
      <pubDate>Mon, 27 Jul 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Teardown: Deterministic Citation Provenance</title>
      <link>https://www.axionalytics.com/citation-provenance-architecture.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/citation-provenance-architecture.html</guid>
      <description>The model writes the answer; the backend decides what it may point at. A grounding critic, marker resolution against the execution record, and why rendering a citation is a security boundary.</description>
      <category>Teardown</category>
      <category>Enterprise Agentic AI</category>
      <pubDate>Sun, 26 Jul 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Why Enterprise AI Pilots Die in Security Review</title>
      <link>https://www.axionalytics.com/why-ai-pilots-fail-security-review.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/why-ai-pilots-fail-security-review.html</guid>
      <description>The pilot worked. Eighteen months later it still is not in production. Five architectural questions decide that outcome, and all five are settled before the first line of code.</description>
      <category>Security</category>
      <category>Zero-Trust AI Governance</category>
      <pubDate>Mon, 20 Jul 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Prompt Injection Is a Containment Problem, Not a Prompting Problem</title>
      <link>https://www.axionalytics.com/prompt-injection-tool-abuse-defense.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/prompt-injection-tool-abuse-defense.html</guid>
      <description>The defence and the attack share a channel, so no instruction closes the class. Four named attacks, and the architectural layers that make a successful one harmless.</description>
      <category>Security</category>
      <category>Zero-Trust AI Governance</category>
      <pubDate>Fri, 17 Jul 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Outbound Broke Because Volume Got Cheap</title>
      <link>https://www.axionalytics.com/outbound-research-not-volume.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/outbound-research-not-volume.html</guid>
      <description>Research and volume trade against each other, and every team eventually picks the losing side. The way out is to automate the research and leave the send alone.</description>
      <category>Revenue</category>
      <category>Agentic Revenue Development</category>
      <pubDate>Fri, 24 Jul 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Your Sending Domain Is an Asset. Bounces Are How You Spend It.</title>
      <link>https://www.axionalytics.com/email-deliverability-firewall.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/email-deliverability-firewall.html</guid>
      <description>Guessed addresses cost more than the campaign that produced them. Three escalating verification tiers, and the design decision that matters most when none of them can answer.</description>
      <category>Revenue</category>
      <category>Agentic Revenue Development</category>
      <pubDate>Fri, 10 Jul 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>BYOC, SaaS, or On-Premises: Choosing a Deployment Topology for Enterprise AI</title>
      <link>https://www.axionalytics.com/byoc-vs-saas-enterprise-ai.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/byoc-vs-saas-enterprise-ai.html</guid>
      <description>Deployment topology is not an infrastructure detail you settle later. It determines which data the system may touch, and it is the hardest decision to reverse.</description>
      <category>Architecture</category>
      <category>Zero-Trust AI Governance</category>
      <pubDate>Tue, 14 Jul 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Human-in-the-Loop That Actually Holds: Write Gates, Blast Radius, and Replay Protection</title>
      <link>https://www.axionalytics.com/human-in-the-loop-ai-architecture.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/human-in-the-loop-ai-architecture.html</guid>
      <description>Most human-in-the-loop is a confirmation dialog. Three properties separate a real approval gate from a checkbox that manufactures consent.</description>
      <category>Architecture</category>
      <category>Enterprise Agentic AI</category>
      <pubDate>Tue, 07 Jul 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>How an Agent Queries a 50,000-Table Warehouse Without Reading the Schema</title>
      <link>https://www.axionalytics.com/agentic-data-exploration-at-scale.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/agentic-data-exploration-at-scale.html</guid>
      <description>Every demo works against twelve tables. Real catalogs have tens of thousands, and the standard approach collapses on cost, latency, and accuracy at the same time.</description>
      <category>Engineering</category>
      <category>Enterprise Agentic AI</category>
      <pubDate>Tue, 30 Jun 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Requirements Traceability Without the Spreadsheet</title>
      <link>https://www.axionalytics.com/requirements-traceability-automation.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/requirements-traceability-automation.html</guid>
      <description>The traceability matrix is accurate the day it is written and wrong within a sprint. Generating the link at authoring time is what makes it survive.</description>
      <category>Verification</category>
      <category>Agentic Test Engineering</category>
      <pubDate>Tue, 23 Jun 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Line Coverage Is Not Verification Coverage</title>
      <link>https://www.axionalytics.com/coverage-gap-analysis.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/coverage-gap-analysis.html</guid>
      <description>Most organisations can report line coverage to two decimals and cannot say which implemented behaviour nobody has written a test for. Those are different numbers.</description>
      <category>Verification</category>
      <category>Agentic Test Engineering</category>
      <pubDate>Fri, 26 Jun 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Paying Down Legacy Test Debt Without an Engineer-Year</title>
      <link>https://www.axionalytics.com/legacy-test-suite-modernization.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/legacy-test-suite-modernization.html</guid>
      <description>The assertions in a decade-old suite are usually fine. What is missing is everything around them — and retrofitting that is tractable in a way rewriting is not.</description>
      <category>Verification</category>
      <category>Agentic Test Engineering</category>
      <pubDate>Fri, 19 Jun 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Your BI Backlog Is a Transcription Bottleneck, Not a Capacity Problem</title>
      <link>https://www.axionalytics.com/bi-backlog-bottleneck.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/bi-backlog-bottleneck.html</guid>
      <description>Adding analysts to a forty-item queue buys a quarter of relief. The queue is long because the same dashboard is being hand-built for the eleventh time.</description>
      <category>Analytics</category>
      <category>Agentic Business Intelligence</category>
      <pubDate>Tue, 16 Jun 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Excel Sprawl vs a Governed Semantic Model</title>
      <link>https://www.axionalytics.com/excel-vs-power-bi.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/excel-vs-power-bi.html</guid>
      <description>The question is not which tool is better. It is which numbers the board sees when four departments each maintain their own version of revenue.</description>
      <category>Analytics</category>
      <category>Agentic Business Intelligence</category>
      <pubDate>Tue, 09 Jun 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Building the Business Case for Enterprise AI</title>
      <link>https://www.axionalytics.com/data-analytics-roi.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/data-analytics-roi.html</guid>
      <description>Finance rejects AI business cases for predictable reasons. Understated loaded cost, unfalsifiable benefits, and an automation rate nobody believes.</description>
      <category>Strategy</category>
      <category>Agentic Business Intelligence</category>
      <pubDate>Tue, 02 Jun 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Why Enterprise Dashboards Get Abandoned</title>
      <link>https://www.axionalytics.com/business-dashboard-guide.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/business-dashboard-guide.html</guid>
      <description>A dashboard nobody opens is not a design failure. It is a specification failure — it was built from the data that existed rather than the decision it serves.</description>
      <category>Analytics</category>
      <category>Agentic Business Intelligence</category>
      <pubDate>Tue, 26 May 2026 00:00:00 +0000</pubDate>
    </item>
    <item>
      <title>Build, Buy, or Assemble: Sourcing Enterprise AI Capability</title>
      <link>https://www.axionalytics.com/ai-tools-for-business.html</link>
      <guid isPermaLink="true">https://www.axionalytics.com/ai-tools-for-business.html</guid>
      <description>The build-versus-buy framing hides the option most enterprises actually need, and the three constraints that decide it have nothing to do with engineering capacity.</description>
      <category>Strategy</category>
      <category>Enterprise Agentic AI</category>
      <pubDate>Tue, 19 May 2026 00:00:00 +0000</pubDate>
    </item>
  </channel>
</rss>
