Notes and posts
I published a new article on LinkedIn about choosing AI automation targets from incident data instead of agent demos: rank recurring work by total hours rather than ticket counts, mine resolution notes for fixes that close codes miss, and test each self-heal candidate for volume, a repeatable fix, a small...
Discovery, observability, and your cloud provider will never agree on what depends on what, and a project to reconcile them into the CMDB rarely finishes. Worse, automation doesn’t notice the conflict. It picks a source and runs. In my latest LinkedIn article, I argue for a different approach: write a...
I shared some thoughts on LinkedIn about adapting the ServiceNow SDLC for AI agents: converting metadata to TypeScript for better type-checking, using developer sandboxes to prevent concurrent edit conflicts, treating agent output as reviewable code in Git, writing detailed stories as prompts for agents, and automating test creation. Read the...
I published a new article on LinkedIn about why successful CSDM implementations focus on modeling only the elements that directly impact workflows, rather than trying to populate the entire framework at once. Read it here: Two Million CIs, and It Still Can’t Route a Ticket.
You can find me on LinkedIn at linkedin.com/in/amadosierra.
