AI Needs a lot of Work

I came across this interesting article on AI (https://sloanreview.mit.edu/article/the-fatal-flaw-of-ai-implementation/) which sought of echos what we have been discussing with customers for implementing AI.

Key implementation aspects

Data Quality Specialists: Quality of data is crucial. Bad data means bad results. Nothing can be achieved with bad data. So people are getting a lot excited about AI but are forgetting or are not aware of the quality of data they have been collecting all along

Data Scientists: Resources would need very good statistical skills to apply the right algorithms to make the best of the captured data and actually come up with trends which matter. This will require significant understanding of data context. That is where domain comes into play

Domain Experts: We had once applied different statistical models and had come up with trends. Interpreting the trends required deep understanding of the functional domain. Domain is crucial both from a vertical and functional perspective.

Analysts: Domain experts should work with analysts who can interpret the trends and relate it to trends from other domains and draw meaningful conclusions and recommend actions. Analysts keep a broader and longer term perspective when interpreting trends

Action Owners: Actions owners should understand the recommended actions and its impact on day to day business. Impact of actions on a long term basis should also be thought through

Automation Gurus: The actions which are repetitive in nature should be embedded into the regular business process to maximize the benefits

Please share your experiences