This week, I looked at several homegrown AI systems built from scratch. Some of them work well. Some of them don’t. Some don’t work at all.
For each case, I asked what was special about their situation, such that they couldn’t use a standard tool or library.
I can believe there are cases with a good answer to this.
Unfortunately, none of these was such a case.
I think that the average waste was the better part of a million dollars, and the average needless delay to deployment was over a year.
No, there was never a Chief AI Officer (CAIO), fractional or otherwise, in the house.
Yes, there should have been.
Deloitte Consulting said less than one in five AI adopters felt confident they had the necessary skills.
I am guessing that all of these failures were members of that one in five who thought they were among the ones with the appropriate skills. They did have the skills. But only skills in building, not knowing how to build, or, more importantly, whether to build.
