McKinsey offers an inside look at the state of AI and data science critical success factors. MLOps and cloud technologies emerge as essential factors to successful Data Science.
The 2021 McKinsey survey on the state of AI (Artificial Intelligence) uncovers MLOps (Machine Learning Operations) and cloud technologies as important success factors to Data Science as detailed by Michael Chui, Partner at McKinsey.
MLOps and cloud technologies stand out for the first time
The annual McKinsey State of AI survey considers the opinion of approximately 1,800 companies from around the world and across multiple industries. This is the first time in the last four years that MLOps and cloud technologies have surfaced as critical success factors for data science.
This makes perfect sense. While many enterprises have been trying to implement data science in the last few years, only a few have successfully made it to production. The survey data unanimously shows that the successful ones have come up with common success factors based on a couple of critical technologies.
Staying on budget
One of the indicators of successful AI, according to McKinsey, is staying on budget. Enterprises that have successfully built and productized AI reported that their projects stayed within their planned budgets, while the other organizations reported running over budget.
MLOps drives faster success
The successful companies that stayed within budget reported that they adopted MLOps platforms to train, test, validate, score, and deploy their machine learning models. They also reported that by using an MLOps platform they reached success a lot faster than if they had approached AI as a “craft”.
“MLOps …, is many times faster than when AI is approached as a craft.” Michael Chui, Partner at McKinsey.
Manually building your machine learning models from scratch does not lead to a high success rate. A lot of time is initially spent finding the right engineer resources and technological tools. Getting up to speed on all these tools, integrating them together, and testing them, leads to budget and time overruns. In short: your data science project turns into a technological project before your eyes.
Then comes operationalization – going from pilot to industrialization. Unsuccessful companies struggle with this step. Even if they do make it past this phase, it takes them months to productize their ML models on the cloud.
MLOps platforms that automate cloud operationalization are crucial to the success of data science. They provide the flexibility to administer cloud resources elasticity, and to manage resiliency and security controls while providing good predictability.
“The cloud also provides flexibility to ramp compute up and down as needed, which is especially useful for retraining models when necessary.” Michael Chui, Partner at McKinsey.
But, migrating to the cloud without an automated platform is difficult. Companies need a platform like the ForePaaS unified end-to-end platform to automate the cloud infrastructure without being lock-in to a single cloud provider and to mass-produce their AI projects at scale. It helps companies create and operationalize repeatable Machine Learning and Analytics projects on any cloud at any scale the easy way – without adding pressure on your teams, with no technological complexities, without sacrificing enterprise requirements and within your budget. Monitoring the quality of your data and the performance of your models as you go through several iterations, is also an important feature to look for in a comprehensive platform.
Read the full article:
For more articles on cloud infrastructure, data, analytics, machine learning, and data science, follow Paul Sinaï on Towards Data Science.
Get started with ForePaaS for FREE!
Discover how to make your journey towards successful ML/ Analytics – Painless
The image used in this post is a royalty free image from Unsplash.