Start with Your Most Pressing Problems, before You Start an AI Project


A large part of my job involves talking to business leaders and CEOs. However, my responsibilities also include training our internal sales folks to help them be more adept at having informed conversations with their customers. In addition to webinars, where I walk them through the various scenarios involved in artificial intelligence, I also produce interactive learning courses.

Recently, I was asked to produce a course on how to introduce the concept of AI to a novice, since the initial conversations our sales people have with their customers are often with individuals who know very little about the technology. This may seem to be a trivial task for someone who deals with the complexities and the cutting-edge of AI on a day-to-day basis. However, when I started working to simplify the material, I faced a challenge I had not expected. In the hopes that this may also resonate with you, I would like to tell you briefly about the course we created and, ultimately, about how we presented the material in a way that would make sense to the average person who is not well-acquainted with AI.

My objective was to introduce AI in the context of a company that was already operating today. Many conversations I have where I’m introducing AI assume my audience has some experience with the preceding technologies that have led to AI. And I felt this was the key ― I had to build a bridge from what people knew and understood, what they had implemented in a typical IT infrastructure, to where the latest AI technologies eventually came into play and helped make their existing infrastructure even more effective.

The way I like to think of AI technologies is that they are just another toolkit to help you solve your problems more effectively. When I present it this way, it makes it easier for people to view it as a different and better problem-solving technique. Rather than trying to find a way of applying AI, in my mind, a better approach is to list all the problems you are currently trying to solve and then see which one is best solved using AI. And, to make things real, you can’t ignore the customer journey, since let’s-face-it problem-solving is a serial process.

You start at one place in time and try out various solutions, some of which work very effectively for a while until you realize their shortcomings, then you must find a new solution that draws on the previous one. Let’s assume we have an agricultural company and that, in its quest to become more efficient and more profitable, its operators are looking at technology to help them. Initially, some of the efficiency boosters they use come from better agricultural techniques, utilizing fertilizers and farm automation equipment that allows much larger scale of cultivation. But, ultimately, they run into some limiting factors where newer and different problem-solving techniques are needed.

For example, the agriculturalists realize that having simple moisture and pH sensors gives them better visibility into production. And believe it or not, as simplistic as this might seem, this is the beginning of the analytics process. In the past, farmers may have relied on manually collecting this data and combining it with some form of intuition to come up with conclusions to help them better their crop yield. Think of this as step one of the entire customer journey. The farmers knew that having more sensors would help them make better decisions.

However, manually collecting all the data took too long. So, when wireless sensor technology finally became cost-effective enough, it made sense to install the sensors that led to the automation of the initial process of data collection of soil parameters. So, this information could now be inputted into a simple spreadsheet that could help in increasing the yield. This is step two of the customer journey. The farmers knew they could do better ― that there were other parameters they could monitor that could even further increase their crop yield, which led to more data being collected more frequently.

Before they knew it, the volumes of data that were being collected and that need to be processed quickly outgrew the spreadsheet methodology they had been using thus far, and this led to the adoption of data analytics technologies. In feeding in all of this information and analyzing it, the realized there were a few other things that could help predict better crop yield that they hadn’t been considering. However, it would require them to bring in additional data sources, like weather forecasts and market data on their crop pricing, to come to the most effective decision. This is step three of the customer journey.

I hope this gives you an idea of how problem-solving requirements change and grow over time. At each step along the way, there was a need for more data and better analytical techniques, which led to better outcomes and, ultimately, this is a continuous cycle of improvement that never stops in the competitive environment of today.

In the past, this was about as good as the technology got. The farm went from using purely manual techniques to some degree of automated data collecting and analyzing. However, most of the analysis was still done using human operators. One of the major problems we face today is a deluge of data that makes it humanly impossible for us to analyze the data effectively. So, this is where some of the newer technologies come into play, where they can parse many terabytes of data and find patterns without being explicitly told what to look for. We still need the human operator to look at the patterns that are being found and to identify which one of these patterns is worth investigating further, but this is a considerably easier process.

As a next step, the farmers realize they can optimize their crop yield by using drones to deliver the insecticides. However, this requires human operators to control the drones. But then they quickly realize there are software applications that allow them to map out their entire farm and have the drone operation become completely automated.

This example is based on an actual agricultural company. Although I haven’t explicitly mentioned it anywhere in my description above, my question to you is: Can you tell me where AI came into the picture?

The point I’m trying to make is that the agriculturalists were just looking to solve their business problems using the best techniques available to them at the time. Initially, it was just collecting data and using spreadsheets, progressing to more complicated data analytics techniques, and ultimately to a machine learning technique. AI wasn’t used just for the sake of using AI but, instead, it was the only way to solve a problem.

So, the next time you have someone who says they would like to start a project with AI, ask them what problems they are looking to solve, and then see which one of those problems would benefit the most from the machine learning techniques we know today.

I hope you find this simple example to be instructive. As fundamental as it is, as I was creating this learning module, I realized how often we overlook the basic steps of what makes a project successful. What I often tell people is that you’ve got to start somewhere, so start with your most pressing business problems and find ways of alleviating them and of making your business more profitable, using whatever techniques it might take ― be it data analytics, machine learning or deep learning. I’d love to hear your comments, so please feel free to drop me a note.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *