Check evaluation, reliability, observability, security, cost, human oversight and ownership before an AI application goes live, and see what is still open.
What it helps you do
A prototype that works on friendly examples is very different from a system that handles every input, every day. This checklist covers the areas that most often cause problems after launch.
How to use it
Review it with engineering, product and security together. Tick an item only when it is in place and tested.
Checklist
0 of 20 checked · 20 items still openWhich gaps block launch
Open items in evaluation, security or ownership should normally block a launch. Observability and cost items can sometimes follow shortly after, if the risk is understood and accepted.
Revisit after launch
Run through the checklist again after the first month in production and after any major change to models or data.
If you have a prototype and want help preparing it for production, our team can review it against this checklist with you.
AI Consulting Services- OWASP Top 10 for LLM ApplicationsOWASP GenAI Security Project · Industry standardgenai.owasp.org
- Semantic conventions for generative AI systemsOpenTelemetry · Specificationopentelemetry.io
- MLOps: Continuous delivery and automation pipelines in machine learningGoogle Cloud · Architecture guidecloud.google.com