We think so. As part of our data science, machine learning, and AI trends and observations that we’ve compiled to help organizations recalibrate for the new year, we’re predicting that, of the machine learning projects that organizations would like to make it to production, more than 50% of them actually will. Here, we’ll unpack some of our reasoning behind this anticipated trend.
What's Fueling This Trend?
Here are some of the key reasons we’re observing an increase of projects making it to production (and believe this will continue into 2022 and beyond):
1. Organizations are becoming more agile in their approaches to driving business value from their AI projects.
According to the O'Reilly book (written with Dataiku experts!) "Introducing MLOps: How to Scale Machine Learning in the Enterprise," business leaders "view the rapid deployment of new systems into production as key to maximizing business value. But this is only true if deployment can be done smoothly and at low risk." Therefore, in order to truly add value, teams need to make sure they develop strong alignment on and governance of their MLOps processes, assessing the risks, determining their own set of fairness values, and implementing the necessary process to manage them.
2. The ongoing rise of robust MLOps practices indicates that the data science and ML industry is continuing to grow in maturity, as it demonstrates more and more models are being deployed in production every day. It also shows that teams are taking ownership of making sure they have a clearly defined plan for standardizing and managing the entire ML lifecycle.
3. More organizations have realized that they can not wait to have all of their data ducks in a row before attempting to scale analytics and AI. They believe they need to conquer traditional or BI analytics first (including data catalogs, data lineage, master data management, etc. before planning for AI).
Contrary to popular belief, data quality is actually best improved by using it to create value (i.e., operationalizing a project that is going to drive tangible business value). If teams want to build a lasting data culture, they need people to confront data, use it, understand its flaws, and then be encouraged to experiment with it and become data champions.
4. Finally, and more importantly, deployment to production isn't just a technical exercise, it's an organizational one.
Companies are not only catching on to the value of production, but they're investing in tools that help make the process frictionless. Dataiku does just that by:
- Providing a platform that makes it easy for data science and IT teams to collaborate on building user-friendly, real-time and batch scoring systems
- Offering production-related features such as scheduling, monitoring, and scenarios, so that teams can build production-ready workflows from the first step
- Enabling the ability to track the status of your production scenarios
- Providing the infrastructure for organizations to govern AI projects at scale, including production lifecycle management (monitoring, retraining, and testing)
Check out this article to see how Dataiku 10 gives data scientists and IT operators additional tools and more flexibility to deploy, monitor, and manage ML models at scale.