7 Tips for Value-Driven AI
How your business can improve the skills of its talent to take greater advantage of AI.
- By Sabina Stanescu
- November 30, 2022
There’s no doubt that artificial intelligence (AI) is changing the way business is done today. AI will ultimately transform every business in every industry. However, despite their desire to use data science when making decisions, many organizations can't find enough qualified data scientists to develop and run their data science initiatives.
Nonetheless, with online training and readily available tools, any software engineer -- or even a business user with a math background -- can become a data scientist. Even if data science wasn’t part of a user’s undergraduate or graduate studies, it’s possible to make the transition to AI and bring the power of machine learning to your enterprise.
Here are seven tips from my experience as a data scientist about how to make that happen.
Tip #1: Brush up on the basics
Before starting an AI journey, it’s best to conduct a basic math review including linear algebra, calculus, probability, and statistics. The best course to complete first is one on introductory Python; from there, you can move on to machine learning and advanced math classes. One of my client companies has a program to teach their programmers Python. After becoming familiar with the basics, many aspiring data scientists get their feet wet by participating in a Kaggle competition.
Tip #2: Know your data
The most sophisticated algorithm can’t reach any conclusions without high quality data. Companies can’t use analytics and machine learning to increase revenues or reduce costs if they don’t have enough information to build an accurate picture of their operations and don’t monitor how efficiency improved as a result of their AI initiative. For one of my most successful data science projects, we gathered data for an entire year before beginning the project.
Tip #3: Pick an attainable goal
The goals for your first AI project should be communicated in a single sentence. One of my first projects was to make the most popular hotels appear higher on the search list for a travel site. The project was a success because the goal could be easily described, executed, and measured. Most important, optimizing the search results dramatically increased the number of nights booked, which had an immediate positive impact on the bottom line.
Tip #4: Keep models up to date
Models must be monitored for accuracy; pipelines need to be maintained to ensure a steady stream of quality data. Every machine learning project needs to address the long-term requirements by having a framework to retrain, test, roll back, and start over if there are problems.
Tip #5: Be prepared to customize
It is always a good idea to see if someone has already developed, trained, and tested a model you can use without having to reinvent the wheel. However, the specific requirements of an application will likely require tweaking to accommodate the specific types of data your business generates. When one of my construction safety projects needed to verify that factory workers were wearing helmets and gloves, the open source object detection models weren’t useful (they could only detect a narrow set of objects for specific environments), so we customized them to suit our needs.
Tip #6: Build a sustainable model
Consider the amount of additional work that a model can create and whether your team has the bandwidth to take on that work. For example, for one project we performed a simple calculation to pick the model that would have the greatest positive financial impact and would produce a manageable number of transactions to review.
Tip #7: Be prepared to scale up
The success of one of our projects created a strong demand for other departments that wanted machine learning added to their IT agenda. To leverage the power of AI throughout your enterprise, create a program for training developers in AI basics, reinforce the power of data to users, and evaluate AI opportunities with an eye on immediate positive results to improve the bottom line.
A Final Word
AI has become the key for creating differentiating operational efficiencies and customer experiences, and is becoming a must-have for companies to stay competitive. With proper training and access to good data, the ability to generate machine learning models will very soon become a required skill for every developer and software engineer. The seven tips presented here will get you off on the right foot.
About the Author
Sabina Stanescu is an AI innovation strategist at cnvrg.io. In her career, Stanescu has been a Data Science leader and practitioner and managed AI, ML and MLOps cloud platforms. You can reach the author via email or LinkedIn.