Flawed Expectations from AI !

Many fall prey to the expectation that AI gives us a magic wand of sort and which can help solve all our unstructured business problems where RPA doesn't fit in ! So if we can't apply RPA, let's apply AI !!

This is where most of the AI initiatives fail.

AI is not a magic wand and its important one has the right expectations while driving AI

a) Firstly driving AI is not quick and easy.

Most of the fancy tool-kits in the market boast of ready to use, pre-built ML algorithms. While they can be very much well leveraged, that might be just useful to drive PoCs and not full scale implementations. So, let's not assume driving AI is going to be quick and easy. To get the real business value, algorithms need to be fine tuned, data needs to be gathered etc.

b) Data availability

This, although most of us know theoretically, we go with the assumption data would be easily available and any sort of data can drive the AI program. Firstly getting data in itself is difficult with complex systems sitting behind the processes and secondly any data provided need to be diligently cleaned and curated to be even ready to use for training and testing the ML algorithms.

Infact any project on AI needs to be started with data. Consider the elements of digitizing the data, effort which goes into cleaning the same and then curate.

c) Maintain the Audit Trail

We might have seen algorithms in machine-learning pilots that come to right answer every time, and yet the projects were shut down because there was no clear way to determine how the algorithm did so. One has to be conscious of this, and remember the audit trail is everything in certain highly-regulated industries.