Assembling the AI Dream Team
By Stephen Thomas
A few years ago when analytics and artificial intelligence were starting to attract mainstream interest, many companies thought they could just hire a few data scientists and be good to go. Other companies found a better way than hiring only data scientists. They learned that many supporting pieces must be in place — beyond data science — to have a successful analytics project.
Here’s an example of the range of roles and skills that should make up an analytics or AI team.
First are technical architects who design the data flow, determine how your computer systems are connected, and choose database products. They are the big planners.
Data engineers develop the data pipeline and implement the flow of data from one database to another. They usually have computer science backgrounds and can model data, write SQL queries, and construct ETL (Extract, Transform, and Load) processes.
The data analyst is like a data detective, someone who can combine business acumen with data analysis skills. Data analysts are adept at doing whatever is necessary to quickly answer very specific business questions.
Data scientists research and build machine learning models. They spend a lot of time organizing, cleaning, and massaging data and then running various machine learning models, assessing their performance, and selecting the one that works best for the particular business problem.
A machine learning engineer then takes the model developed by the data scientist and implements it at scale so that it is useful to companies and end users. The machine learning engineer ensures the model is robust and efficient enough to run for a million people a day without failure.
Let’s not forget the product engineer. This is the person who actually builds the product that the end user will interact with.
And finally there is the analytics/AI translator, a manager who oversees the entire team. The ideal analytics/AI manager knows quite a bit about analytics and AI and understands what each of these people do on a day-to-day basis. These managers also have the business skills to interact with upper management to translate business strategy into a data science solution, and vice versa.
On top of that, some companies may also have client solutions specialists who interact with the client to understand their needs and wants. They work directly with the manager, who translates client information into a technical roadmap and strategy.
It Takes an AI Village
The point here is that you cannot just hire two or three data scientists to make your company ready for AI. You need to build a team with overlapping skill sets if you want to make an analytics-AI strategy actually work.
What skill sets do you need on your analytics team? Certainly analytical skills, everything from analytical decision making, data modeling and SQL, descriptive analytics with visualization and storytelling (for reports and dashboards), and predictive analytics using machine learning. You also need all the domain-specific analytics tricks in functional areas such as marketing, finance, and operations. The data scientists may require more specific AI skills in the various types of machine learning.
From what we have observed and heard from our industry partners, all analytics-AI roles are in great shortage right now. Almost every organization is trying to build out its analytics capability by either training existing employees or hiring recent graduates.
But one role, in particular, is attracting a lot attention — the “translator.” By translator, I mean the analytics manager who understands both the business and the technical sides. Without these people, your technical teams and senior managers will talk past one another, with the result that the organization will not really capitalize on all of the benefits of analytics and AI. Translators are crucial to success.
In many ways, the analytics-AI team described here applies to large firms. But if you’re from a smaller organization, there are still steps you can take to develop and implement an AI-powered project. There are plenty of powerful cloud-based tools available— Amazon Web Services, IBM Cloud, Google Cloud Platform, Microsoft Azure — that make it easy for small companies or even individuals to benefit from machine learning. All it requires is plugging in a good data set and a business query, and the cloud-based platform will generate some predictions and recommendations.
With just one or two knowledgeable people within your firm to manage this process, you can start small and explore how analytics and AI can streamline a business decision process. From there, you can assemble your team and scale your efforts at a manageable pace.
Stephen Thomas is an adjunct faculty member at Smith School of Business, Queen’s University. He is the Director of the Smith Master of Management Analytics program, and the inaugural Director of the Smith Master of Management in Artificial Intelligence program.