You paid an agency to build your system. They used AI heavily. You're about to go live with something that will handle real users and sensitive data. Do you actually know what you have?
I keep being the person who comes in after the fact. Sometimes before go-live - someone with the instinct to know they don't know, calling because something feels off but they can't name what. Sometimes after - someone who's live, something happened, and now they need to understand what they're sitting on and how bad it is.
Both situations are fixable. They're not the same kind of fixable.
What I find when I go in is pretty consistent - not obvious failures, but decisions that went wrong somewhere in the spec or the execution. Things that passed internal review, sometimes looked genuinely good, but break when you trace how they actually behave rather than how they were supposed to behave.
The code had often passed internal review. Sometimes it looked genuinely good. The problem is that internal reviewers spend weeks internalising the same assumptions as the people who built it. They know what the code is supposed to do, so they see what they expect to see.
There's nothing wrong with AI-assisted development - I use it every day. But AI writes to the spec it's given. It has no opinions about what the spec should have included, and no production instinct - no accumulated experience of watching these specific patterns fail. Those gaps are exactly what independent review is for.
Independent is the important word - not the team that built it, not the person who approved the spec, not someone who's spent months sharing the same assumptions as everyone else. Someone with no stake in the outcome, looking for what they weren't told to look for.
Before you go live - not after, before - is when this is useful. Two to four weeks depending on scope. You'll know what you actually have. Either you'll leave with the confidence to ship, which is worth something real, or you'll find out now. Finding out now is better than finding out at 2am when something's wrong and you're working out who to call and what to tell them.
If that's where you are, the AI & Agent Production Audit is what it's for.