<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
  <title>Neckles IO Articles</title>
  <link>https://neckles.io/articles/</link>
  <description>Observations from real engagements: CRM migration, n8n automation, AI agents under governance, and owned infrastructure, written for operators.</description>
  <language>en-ca</language>
  <lastBuildDate>Fri, 12 Jun 2026 12:00:00 +0000</lastBuildDate>
  <atom:link href="https://neckles.io/feed.xml" rel="self" type="application/rss+xml" />
  <item>
    <title>Your AI Agent Should Never See a Password: Credential Isolation for Agentic Operations</title>
    <link>https://neckles.io/articles/credential-isolation-for-ai-agents/</link>
    <guid isPermaLink="true">https://neckles.io/articles/credential-isolation-for-ai-agents/</guid>
    <pubDate>Fri, 12 Jun 2026 12:00:00 +0000</pubDate>
    <description>Agents are being handed API keys in prompts and environment variables, where one injected instruction or leaked transcript exposes everything. The fix is architectural: a per-client gateway that injects credentials server-side, so the agent can use a system it can never log into.</description>
  </item>
  <item>
    <title>AI Enrichment That Does Not Hallucinate: Typed Failure Contracts and Deterministic Scoring</title>
    <link>https://neckles.io/articles/ai-enrichment-typed-failure-contracts/</link>
    <guid isPermaLink="true">https://neckles.io/articles/ai-enrichment-typed-failure-contracts/</guid>
    <pubDate>Fri, 12 Jun 2026 12:00:00 +0000</pubDate>
    <description>Letting a model invent a website is cheaper than checking one, so at scale it will. The fix is structural: typed failure contracts, claims verified against their sources, and confidence computed by code the model never touches.</description>
  </item>
  <item>
    <title>Leaving Webflow Without Losing Your Rankings: An SEO-Parity Replatform</title>
    <link>https://neckles.io/articles/webflow-to-astro-seo-parity/</link>
    <guid isPermaLink="true">https://neckles.io/articles/webflow-to-astro-seo-parity/</guid>
    <pubDate>Wed, 10 Jun 2026 12:00:00 +0000</pubDate>
    <description>Rankings live at URLs, and a careless replatform throws away years of search equity in a week. The SEO-parity discipline (inventory, URL conventions, metadata carry-over, and a verified cutover) is how a site leaves Webflow without losing what it earned.</description>
  </item>
  <item>
    <title>Version Control for n8n: Running 100+ Production Workflows from a Repository</title>
    <link>https://neckles.io/articles/version-control-for-n8n-workflows/</link>
    <guid isPermaLink="true">https://neckles.io/articles/version-control-for-n8n-workflows/</guid>
    <pubDate>Wed, 10 Jun 2026 12:00:00 +0000</pubDate>
    <description>Browser-edited automation has no history, no review, and no answer to &#x27;what changed and when&#x27;. GitOps for n8n (workflows in a repository, deployed one-way through CI) is how automation stops depending on one person&#x27;s memory.</description>
  </item>
  <item>
    <title>Should You Self-Host Your CRM? An Honest Decision Guide</title>
    <link>https://neckles.io/articles/should-you-self-host-your-crm/</link>
    <guid isPermaLink="true">https://neckles.io/articles/should-you-self-host-your-crm/</guid>
    <pubDate>Wed, 10 Jun 2026 12:00:00 +0000</pubDate>
    <description>Self-hosting a CRM is right for some businesses and wrong for many. An honest look at what you take on, what you gain, and the questions to answer before committing.</description>
  </item>
  <item>
    <title>Moving a Product Catalogue from QuickBooks to Shopify Without Wrecking the Store</title>
    <link>https://neckles.io/articles/quickbooks-to-shopify-migration/</link>
    <guid isPermaLink="true">https://neckles.io/articles/quickbooks-to-shopify-migration/</guid>
    <pubDate>Wed, 10 Jun 2026 12:00:00 +0000</pubDate>
    <description>The two obvious paths out of QuickBooks (manual retyping and a one-shot CSV import) both fail in predictable ways. A catalogue migration should be built like infrastructure: batched, checkpointed, and safe to run twice.</description>
  </item>
  <item>
    <title>Migrating from Pipedrive to a Self-Hosted Twenty CRM: A Four-Phase Playbook</title>
    <link>https://neckles.io/articles/pipedrive-to-twenty-crm-migration/</link>
    <guid isPermaLink="true">https://neckles.io/articles/pipedrive-to-twenty-crm-migration/</guid>
    <pubDate>Wed, 10 Jun 2026 12:00:00 +0000</pubDate>
    <description>The four-phase playbook we use to move businesses from Pipedrive to a self-hosted Twenty CRM: audit, implementation, migration, and training, with the failure modes of skipping each phase.</description>
  </item>
  <item>
    <title>The Ticket Volume Trap</title>
    <link>https://neckles.io/articles/ticket-routing-vs-ticket-volume/</link>
    <guid isPermaLink="true">https://neckles.io/articles/ticket-routing-vs-ticket-volume/</guid>
    <pubDate>Sat, 30 May 2026 12:00:00 +0000</pubDate>
    <description>When support volume goes up, the instinct is to hire. The companies that break the volume-to-headcount relationship look at routing, not staffing.</description>
  </item>
  <item>
    <title>Field Teams Don&#x27;t Have a Technology Problem</title>
    <link>https://neckles.io/articles/field-teams-data-access/</link>
    <guid isPermaLink="true">https://neckles.io/articles/field-teams-data-access/</guid>
    <pubDate>Sat, 30 May 2026 12:00:00 +0000</pubDate>
    <description>A field operation with over a thousand people working from yesterday&#x27;s data is not a technology failure. The data exists. It just lives in a system back at the office.</description>
  </item>
  <item>
    <title>Why Most Dealer Networks Run on Spreadsheets and Group Texts</title>
    <link>https://neckles.io/articles/dealer-network-lead-distribution/</link>
    <guid isPermaLink="true">https://neckles.io/articles/dealer-network-lead-distribution/</guid>
    <pubDate>Sat, 30 May 2026 12:00:00 +0000</pubDate>
    <description>Enterprise dealer management platforms are built for franchisors with hundreds of locations. For a network of five to twenty dealers, they are the wrong answer. Here is what actually works.</description>
  </item>
  <item>
    <title>Your CRM Data Is Dirty Because of How Leads Enter It</title>
    <link>https://neckles.io/articles/crm-data-quality-intake/</link>
    <guid isPermaLink="true">https://neckles.io/articles/crm-data-quality-intake/</guid>
    <pubDate>Sat, 30 May 2026 12:00:00 +0000</pubDate>
    <description>Clean data is a downstream problem. By the time someone is running a deduplication script or auditing contact records, the damage is already done at intake.</description>
  </item>
  <item>
    <title>The Moment Your CRM Starts Costing You Deals</title>
    <link>https://neckles.io/articles/crm-costing-you-deals/</link>
    <guid isPermaLink="true">https://neckles.io/articles/crm-costing-you-deals/</guid>
    <pubDate>Sat, 30 May 2026 12:00:00 +0000</pubDate>
    <description>The warning signs are rarely dramatic. A quote takes two days instead of two hours. A contract sits waiting for someone to generate it. That is a workflow problem, not a sales problem.</description>
  </item>
</channel>
</rss>
