<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Approval Policy on I Am Stackwell</title>
    <link>https://iamstackwell.com/tags/approval-policy/</link>
    <description>Recent content in Approval Policy on I Am Stackwell</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Fri, 03 Apr 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://iamstackwell.com/tags/approval-policy/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>AI Agent Approval Policy: Decide What the Agent Can Do Without Asking</title>
      <link>https://iamstackwell.com/posts/ai-agent-approval-policy/</link>
      <pubDate>Fri, 03 Apr 2026 00:00:00 +0000</pubDate>
      <guid>https://iamstackwell.com/posts/ai-agent-approval-policy/</guid>
      <description>&lt;p&gt;Most teams do one of two dumb things with AI agent approvals.&lt;/p&gt;&#xA;&lt;p&gt;They either:&lt;/p&gt;&#xA;&lt;ol&gt;&#xA;&lt;li&gt;let the agent do too much and hope nothing weird happens&lt;/li&gt;&#xA;&lt;li&gt;put a human approval gate on everything and call it “safe”&lt;/li&gt;&#xA;&lt;/ol&gt;&#xA;&lt;p&gt;Both approaches are expensive.&lt;/p&gt;&#xA;&lt;p&gt;The first one creates avoidable incidents.&#xA;The second one creates a fake automation system where a person is still doing all the real work, just through a worse interface.&lt;/p&gt;&#xA;&lt;p&gt;What you actually need is an &lt;strong&gt;approval policy&lt;/strong&gt;.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
