Why Aren’t More Organizations Realizing the Potential of Machine Learning?
As organizations increasingly invest in artificial intelligence (AI) initiatives, most are not achieving their goals. Gartner estimates that through 2023, only 50% of organizations will take their AI projects past proof of concept.
We asked MIT Professor Youssef Marzouk what factors might be holding companies back from faster and broader machine learning adoption and impact. Marzouk is co-director of the MIT Center for Computational Engineering and the director of MIT’s Aerospace Computational Design Laboratory. He is also one of the lead faculty for the MIT xPRO online program Machine Learning, Modeling, and Simulation: Engineering Problem-Solving in the Age of AI.
Along with cost and organizational inertia, Marzouk believes that another major hurdle is limited AI knowledge and skill. According to a 2021 Genpact and MIT Sloan CIO Symposium survey of more than 500 CIOs, 49% say they don’t have sufficient talent inside of their companies and are relying on external providers to help with hiring employees who have experience with AI and cloud systems. While hiring to fill the AI capabilities gap is important, Professor Marzouk sees a distinct advantage in also building AI knowledge and skills among your current staff.
“AI is not fairy dust... to realize its potential, we have to help people understand what machine learning is and what it isn’t, and how they can best apply it to their business. It’s time to extend this knowledge beyond the data scientists and de-mystify machine learning for the whole organization.”
-- Professor Youssef Marzouk, Director of MIT’s Aerospace Computational Design Laboratory, Lead Instructor of MIT xPRO's online program, Machine Learning, Modeling, and Simulation: Engineering Problem-Solving in the Age of AI
“AI is not fairy dust,” says Marzouk. “We’ve already seen that the thoughtful application of machine learning can lead to positive business outcomes. But to realize its potential, we have to help people understand what machine learning is and what it isn’t, and how they can best apply it to their business. It’s time to extend this knowledge beyond the data scientists and de-mystify machine learning for the whole organization.”
Here are three ways that Professor Marzouk recommends for organizations that want to better prepare your workforce for scaling and optimizing AI and machine learning initiatives.
Get Your Organization Fluent with AI and Machine Learning Vocabulary
Ideas for how to leverage AI tools can come from any function or level of the organization, if they understand what’s possible. Understanding of these tools shouldn’t live only in technical groups. Instead, everyone should have a baseline understanding of the possibilities and potential of AI.
The average non-technical professional might not be able talk in detail with a data scientist about the parameters of a machine learning pilot. Having professionals across the organizations learn a common language will open up the meaningful dialogue and collaboration required for successful, holistic, and sustainable outcomes.
Develop AI Capability in Your Domain Experts and Bring Data Scientists Closer to the Business
Applying AI for impact requires a level of sophistication around how to apply the tools for positive business impact. This is only possible when you marry industry and domain knowledge with AI understanding. Marzouk suggests that real progress occurs when people understand a bit of both.
Providing on-the-job training for AI skills to business or domain experts can result in stronger business cases for action and can also help teams flag issues early on so projects are more likely to succeed.
According to Marzouk, “When it comes to machine learning, you want your team to have an appreciation and a healthy skepticism for the technology. You can’t blindly trust every result. The conclusions reached need to be interpreted by someone that has domain expertise.”
Domain experts can raise questions such as:
- How trustworthy are the results of our analysis?
- Do the findings align with the technical knowledge and data we already have?
- What is it going to cost to get results?
Build Out Your Machine Learning Toolbox with Industry Best Practices
When initiating an AI or machine learning project, starting small can make an impact. Data collection, curation, communication and interpretation are key steps for organizations looking to move their AI projects past proof of concept.
“It’s hard to move the ship and challenge the status quo, so companies leading with AI are starting with small exercises and pilots,” says Marzouk. “Once small projects work, they are translated to other applications and move into broad production.”
Typically, getting the data set ready takes more time than the actual application of machine learning. For that reason, developing and maintaining best practices for cleaning the data and feeding the models is as important as the model itself. Moreover, teams who adopt AI and machine learning tools will also need crucial non-technical skills to interpret and communicate the results, to allow for quick cycle refinement and action.
“Developing more informed users and customers for AI tools across the organization is what creates a positive “push and pull” for this technology -- the back and forth dialogue between the business and the AI leaders required to make the best use of these tools,” says Marzouk.
Whether you’re an individual contributor or an executive, a domain expert or a technical professional, understanding the key concepts and applications of AI and machine learning through online workforce training is the first step toward pushing your projects past ideation and toward meaningful impact.
If you or your organization are ready to get started or refine your next AI and machine learning initiative, join Professor Marzouk for MIT xPRO’s upcoming online program, Machine Learning, Modeling, and Simulation: Engineering Problem-Solving in the Age of AI.