Every few months a new AI certified developer badge shows up. Some are useful signals. Most test whether someone read the documentation, not whether they can build something that holds up after launch. Here is what we would ask before trusting one, including our own.
Ask what happens after launch, not just at build time
A certification that only tests build-time skill misses the part that actually matters: who is accountable when the system is wrong in production. Ask directly whether the certifying standard requires any post-launch responsibility, or whether it is a one-time exam with no connection to what gets shipped.
Ask how they define done
A developer who defines done as matches the ticket is optimizing for a different outcome than one who defines done as moved the metric we agreed on. The second answer takes longer to get to in conversation and it is the one worth hiring for.
Ask about failure handling, specifically
Do not ask whether they have worked with Claude. Ask what happens in their systems when the model returns a confident wrong answer, how they catch retrieval drift before a user does, and what their evaluation harness actually checks. Vague answers here are the clearest signal a certification was about API familiarity, not production engineering.
What CCA-F requires that most badges do not
Our own certification, CCA-F, Claude Certified Associate: Forward, exists because we kept seeing this gap. It covers six engineering disciplines, ties every engagement to a measurable business outcome before any code gets written, and keeps the certified engineer accountable for what they shipped for a defined period after launch. That is the bar we think any certification claiming production-readiness should have to clear.
