When organizations are dipping their toes in the water it is natural to think of each experiment as an isolated one-off project. As those experiments succeed and we start to focus on “generating value from data” as a core differentiator then we can no longer just build “proofs of concept”, but instead think of maintaining production systems where we are implementing continuous improvement in the face of a constantly changing environment.
It is “remarkably easy to incur massive ongoing maintenance costs … when applying machine learning”. Minimizing costs and maximizing agility over the long term is best done by asserting your “intention” from the beginning rather than re-working after the issues too painful to be ignored.