I remember 30 years ago, I got a summer job via my Computer Science Operations Research professor, and the new thing was Artificial Intelligence. I was actually working on a system that would predict the occurrence of low forming clouds as these were the best predictors of imminent rain.

We were trying to simulate human (smart, educated, successful meteorologists, mind you) rules-of-thumb in a sort of decision tree. I was using the LISP programming language, if anyone remembers that. I digress.

Today, I would be more likely to be offered such a summer job by my Stats professor, in the Mathematics department. While this is true, there are still core practices that have nothing to do with Mathematics OR Computer Science that organizations still struggle to put in place, creating obstacles to adopting and leveraging AI. Here are the top five, according to many surveys.

Lack of a solid strategy Organizations have not yet put a solid strategy in place. From understanding their data, identifying the key entities, the complex integration required and prioritizing that work, all the way to creating a linked digital transformation strategy (we have to create the data in many cases, not to mention the need to have all the right touch points with customers and employees so we can automate as much of the AI we create and really effect the customer experience), this is an impediment.

Lack of foundational technology Without a solid foundation of technology, it would be like hiring the best F1 driver available but not getting the car right. All the tools must make data easy to find and easy to get to, easy to play with, wrangle, clean up and then apply to various algorithms. And this must be repeatable.

Lack of end-to-end mindset The most digitized organizations that have greater adoption of AI will tell you that the biggest payoffs come from eliminating the concept of silos, such as looking at the full customer journey and all its touch points, not just focusing on sales or web browsing and the likes. If on my last 8-hour flight with my favorite airline my entertainment system was not functional, maybe I will think twice about re-booking that airline. Maybe I am not a complainer, so if the airline doesn't know this happened to me, they are missing something. This will affect models - your data scientist will definitely tell you that.

Lack of skills Talent if of course important. But organizations sometimes look for this super-human who can do it all. It is possible to break down the work and add processes allowing talent becomes more accessible. Because for AI to work, we are not just talking about needing PhD's in Math that can write Python code and come up with a Data Strategy. It is more than that.

Lack of leadership This one seems to come back in all my blog posts. But nothing gets done in larger (or any?) organizations without a solid, results-oriented manager "owning and committing to" something. In the case of AI, it has to come from the business and preferably from an executive who has a good sense of all the implications. When the CEO is convinced, he typically assigns a Chief Data Officer or a Chief Analytics Officer (or both!).

Conclusion Whatever your endeavor, the are pitfalls that you can avoid. A little bit of research can go a long way.

©2020 by Modern Data Analytics