the ai layer that was never running

i sat with a client this week and walked module by module through a marketing tool that had been shipped to them. keyword clusters that did not match the source data. content that came out identical across different inputs. a rank tracker throwing errors.

the root cause was almost funny. the ai api key had never been configured. the entire intelligence layer of an ai product had never run, and nobody noticed, because the templates underneath produced output that looked plausible.

we added the key live on the call. the results barely changed, which exposed the second failure: hard coded logic was overriding the ai anyway. two lessons for the price of one. never ship without confirming the core dependency is actually live, and plausible output is the most dangerous kind of broken.

which of your systems looks like it is working because nobody has checked?