diff --git a/articles/ai-rewriting-the-job-description.md b/articles/ai-rewriting-the-job-description.md
index ac70fd63401..d4fdd21e32f 100644
--- a/articles/ai-rewriting-the-job-description.md
+++ b/articles/ai-rewriting-the-job-description.md
@@ -1,8 +1,19 @@
 # AI isn't just replacing jobs, it's rewriting the job description
 
-When General Motors laid off more than 500 IT workers in May and immediately posted 80 open roles for AI-native developers and data engineers, TechCrunch called it a "skills swap." That framing is a little too neat. But it points at something real, and it's not just happening at GM.
+*AI isn't erasing IT work so much as redrawing who's positioned to do it. The dividing line runs straight through IT teams, and the data shows where.*
 
-The question worth asking isn't whether AI is disrupting IT employment. It is. The question is how, and whether the answer is what you think it is.
+When General Motors laid off more than 500 IT workers in May and immediately posted 80 open roles for AI-native developers and data engineers, TechCrunch called it a "skills swap." That framing is a little too neat, but it points at something real, and it's not just happening at GM. The question worth asking isn't whether AI is disrupting IT employment. It is. The question is how, and whether the answer is what you think it is.
+
+## Key takeaways
+
+- **Experience, not job function, decides who benefits.** The workers AI displaces and the workers it amplifies often share the same title. What separates them is the accumulated judgment that turns AI into a collaborator instead of a substitute.
+- **The jobs didn't disappear, the required skills moved.** Roles demanding AI skills now carry a large wage premium, and it's climbing fast. The openings exist; the qualifications are changing faster than most people can keep pace with.
+- **AI rewards structured, code-first config.** AI tooling closes the loop in seconds against a legible GitOps repo and stalls against a GUI. It doesn't make ClickOps faster; it makes code-first workflows dramatically faster.
+- **The productivity divide between teams is already measurable.** Client platform engineers on code-first workflows are pulling ahead of those still clicking through consoles, and the gap is widening now, not later.
+- **The disruption is real and lands unevenly.** Transition costs fall on individuals while the productivity gains accrue to organizations, and developers' own trust in AI accuracy is falling, not rising.
+- **Judgment is the skill that holds its value.** Knowing which policies matter, understanding your org's risk tolerance, and catching output that's technically correct but operationally wrong stay human problems.
+
+<a purpose="cta-button" href="https://fleetdm.com/articles/why-ai-powered-device-management-requires-gitops">See why code-first config wins</a>
 
 ## The pattern in the data
 
@@ -48,6 +59,8 @@ Invest in judgment over syntax. The skills that hold their value are the ones AI
 
 If you're waiting to engage until you're sure this is real: the gap between teams moving now and teams moving later is already measurable. The job description is being rewritten. The question is whether you're involved in writing it.
 
+*Ready to build the code-first surface AI needs? [See Fleet's GitOps workflow](https://fleetdm.com/docs/configuration/yaml-files) or [get a demo](https://fleetdm.com/contact).*
+
 <meta name="articleTitle" value="AI isn't just replacing jobs, it's rewriting the job description">
 <meta name="authorFullName" value="Kitzy">
 <meta name="authorGitHubUsername" value="kitzy">