ETI Software executives Pete Pifer, CEO and Tom Taylor, Chief Strategy Officer, recently attended the AI Summit for Business in New York. The mission? To learn about the latest events in the Artificial Intelligence world as well as to understand better how to position ETI to improve services for its customers.
Many of the speakers highlighted that while AI is important, the first step is to understand the business problem to be solved. Too many times, AI Engineers and Data Scientists jump in with their tool of choice without first understanding the problem at hand. Once you have jumped that hurdle, consider these factors:
Does a great set of data exist? If there is no data, there will be no AI solution.
Is the data clean? Most likely it is not completely clean data. Several speakers got laughter from the audience when ask if there is old, legacy data in their data mix. There is almost always old data and this data will most probably not be in the best format. Data cleansing will be an important step to maximize the value of the data.
How should the data and the problem be approached? Different people will approach the data differently. Data Scientists will take a disinclined approach to get as much out of the data as possible. Statisticians will take a more rigorous approach to try to obtain specific binary or variable results. Analysts will take more of a blue-sky approach and ask questions like; What can the data tell us? What are the unknown, unknowns? Therefore, always know who is approaching the problem and what their bias perspective is.
There is no question that AI is growing quickly and having a big impact on businesses that are doing it well. If you are not doing some sort of AI work, get started. The early leaders will grow their business significantly, the laggards will lose out.
Think big but start small. Small pilots that impact only a part of the business are the best place to “fail fast”. Once there is a pilot project that is working well, it will be easier to spread to other parts of the company and begin to build an enterprise-wide Data Culture.
People will always be a part of the AI environment. People make the data, people use that data/AI results and mostly importantly people handle the problems that need creative solutions. Don’t forget about the human element.
How much human interaction will be needed? A good definition is below:
Partial AI – there is a human in the loop
Supervisory AI – there is a human on the loop
Full Autonomy – the human is out of the loop