Last year, our Image Recognition pipeline was partially built (cleaning, clustering, invariance - essentially ready to feed into 'AI') when we had to stall that part of the project, whack a classifier in front, and pivot to tackling some unexpected problems elsewhere. As such we weren't comfortable calling it Image Recognition, so called it Image Analysis or 'IA'. I loved the fact that IA was AI backwards, almost "It isn't AI, it is IA!".

Now I am waiting to use the sentence "Our IA pipeline makes use of several types of AI" without contriving the situation for it. Its the little things that keep you slightly insane.