What's been hard about building AI features/apps?
There's a plethora of hype out there, it can be easy to drink the kool-aid. But for those of us who are actually working on "AI features" or apps, what's been the hardest part of getting your idea or requirement to production?
Same as every other feature: political (understanding requirements, getting consensus, assigning priority, setting expectations), technical (configuration, connection, authn/z) and operational (monitoring, observability, evolution).