Share this post:
Jumping into the AI World blindly and attempting to solve real-world problems could cause more problems than is trying to solve. Throwing AI Tools at data indiscriminately could cause much confusion, and unless we’re very lucky, the difficulties encountered and/or results might cause us to abandon AI entirely. In this blog, I suggest some approaches, identify some issues and help to define where and how to start with AI, particularly in the IP Space.
First of all, we must recognize that a computer is just a tool. Software, including AI, is similar to the instructions for the tool that most people tend to ignore and toss away with the packaging. I suggest that we read the instructions and because you cannot actually break anything with the tool, try it. If you think it, try it…learn from it…try it again.
IDENTIFY PAIN POINTS
Where is the pain? How strong is it? Can the pain be ignored or do we need to focus on it first? Take some time to identify the pain points in the process you are trying to improve. Rank the pain points. Ranking can be performed in several ways and can be used to prioritize the steps to take. Is it that there is too much data? Is the data too complicated? Is your current process ineffective?
KNOW THE DATA
One needs to look at what data is available. Is it in a format that AI tools can handle and if not, can it be transformed to that format if needed? Can you identify the fields that are required or should you import them all? Might there be connections among data sets that we want AI to look for and provide insights that we might not have seen? How will the data sets interact?
Take some time to consider these issues.
PLATFORM DESIGN AND IMPLEMENTATION
When looking at the ranked pain points and potential use of AI, look for common tools and develop a platform that is flexible and addresses as many of the pain points as possible. One example is web crawling. Web crawling can be used for several sources of data: financial, technical, social and so on, but the actual tool that performs the web crawling might cover all of the sources — so if designed correctly, the platform module that performs the web crawling could be utilized in various ways for multiple pain points.
Match the pain points to the AI tools that are available. Again, look for commonalities in the tool features and leverage them. Think outside the box. Can/should a document be broken into chapters? Are images important? If there are no AI tools available for a particular pain point, get creative!
WHAT DO YOU WANT AI TO DO FOR YOU?
Now that you have identified your pain points, data sets and AI tools, what do you want AI to do for you? This is sometimes a difficult question. Many users want an “Easy Button” — click and get the answer. Although for some data and in some cases an “Easy Button” might be possible, AI tools should be used to provide helpful insights. Our goal is Augmented Intelligence, not an “Easy Button”.
Leverage subject matter experts and users to understand their desires, but keep the pain points in mind. Providing additional insights that do not help users just provides more data, which is unwanted. AI should provide correct and useful insights to augment users’ decision processes.
Once you identify what you want AI to do, it is a matter of learning the nuances and limitations of AI. Some of these are training the tool, limitations on data types and size, overall AI tool and data storage costs, and others.
In the next blog, I will discuss one of these: training, learning and testing AI results.