4 Priorities and Action Items for Leveraging Artificial Intelligence and Machine Learning in 2020

Artificial intelligence and machine learning (AI/ML) are always a mixture between different learning models and a tree structure of hardcoded rules. Let’s accept this fact, at least for now, and then let’s leverage what is possible today and increase productivity and our ability to continuously bring our product offerings closer to what customers want. Without specifically endorsing the Google approach toward AI/ML, I recommend watching the below video for a very recent overview of what foundational tools are at your disposal today.

Priority #1: Solving Small Problems Is Big

While we are all hunting for a turn key AI that will complete entire tasks based on our training, we need to accept that the key strength of AI in 2020 is the ability to solve specific problems better and cheaper than with traditional rule-based coding. If you don’t believe me, take a look at Christopher Manning’s concise and brilliant explanation of this limitation.

Action item: Make AI part of your software development process. Enable your development and product teams to recognize opportunities for NLP, text analysis, clustering, object recognition, image categorization, and so on. Create an arsenal of more and more fine tuned AI artifacts as part of your DevOps pipeline.

My conversation with Adam Carmi, CTO of Applitools shows you how the clever combination of traditional coding practices and AI/ML models can significantly enhance a product.

Priority #2: Combining Domain Knowledge and AI Models Is Critical

Unfortunately, we cannot simply flood our convolutional neural network (CNN) with data from all imaginable sources and expect high quality predictions. Remember, that the same AI that drives McDonald’s drive through ordering will actually not work for Burger King. Therefore, there is always a human in the mix to select what data sources make sense for what use case and define how these data streams need to be filtered and transformed for optimal results.

The new Dynatrace Kubernetes monitoring dashboard is a great example for the combination of using human expertise to find the right data sources and identify feedback loops, and the use of AI for automatically processing these data streams and identifying actionable metrics. Check out my interview with Bruce Gain at TheNewStack.com.

Action Item: Create a virtual center for AI/ML excellence. While this might sound intimidating, you can pull this off without too much disruption. Remember, that you do not need to churn out data scientists, but you simply have to create subject matter experts who are able to understand the basics of what is needed to qualify input data, select the appropriate learning model, and validate the results. Initially, you need to limit this to a basic set of use cases to show your organization some early successes.

Priority #3: Make AI Accessible to Everyone

The more of your staff develops a basic understanding for what AI/ML can do and where the limits are, the more you will be able to identify high value use cases that can get you ahead in the market place by freeing up staff time or making better decisions.

There is no better tool than Exploratory to start analyzing data from almost anywhere. This tool is fantastic for business staff of almost any kind to start drilling into already existing data and test some hypothesis. Exploratory then lets you naturally progress in your ability to understand and actually test opportunities for data driven decision making. It is actually much easier to use than Excel. Check it out at: www.exploratory.ai

Action Item: Offer some basic Tableau training for everyone or buy a bulk license of Exploratory and hook those up to your CRM, CMS, and other data sources (internal and external). Staff needs to get a feel for data-driven decision making and experimentation.

Priority #4: Use Natural Language Processing (NLP) Now

There is a ton of information out there, but it is too time consuming for us humans to digest. While NLP is far away from becoming a universal tool we can have “real” conversations with, we can definitely use it for tasks that will make us humans much more productive, simply by helping us filter our daily reality.

Action Item: Think about sources of knowledge that are generally too unstructured or large to be of much value. But on the other hand, you know that somewhere within this massive haystack of data, there are the answers for significantly improving how you serve your customers. You need to start small, but eventually, you can take this very far to create more and more specific and actual instructions and guidelines for humans.