2021 Will Be the Year of Enterprise Machine Learning
To say that 2020 was a year of uncertainty and disruption would be an understatement.
Every organization had to adapt to the realities of the pandemic. Business leaders adjusted their strategic roadmaps to refocus on their most pressing priorities and scale back or eliminate less critical plans. In the midst of considerable change, we at Algorithmia saw something fascinating unfold: Many enterprises not only moved full speed ahead with their AI/ML initiatives, but they actually doubled down on those efforts.
That is one of the overarching takeaways from our new 2021 enterprise trends in machine learning report: We saw really dramatic investment in AI/ML last year in the face of pandemic-related uncertainty.
Now in its third year, our independently conducted report includes survey input from a great mix of more than 400 leaders and practitioners with a stake in their organization’s AL/ML strategy. This year’s report produced 10 key trends that business and technology leaders should be aware of with 2021 underway. You’ll find rich data and analysis for each of these trends in the report.
In this article, I want to highlight the most critical themes we found — and what they mean for 2021 and beyond.
Priority Shifts: AI/ML Budgets Are Skyrocketing
As I mentioned earlier, most enterprises didn’t slow down their AI/ML work in 2020. Instead, they dramatically accelerated their investments.
AI/ML-related spending is growing, with 83% of survey respondents indicating year-on-year increases in their budgets. That’s reflected in hiring, too, with the average number of data scientists employed in these enterprises growing 76% YoY. The report includes plenty of other numbers that tell a similar story: Enterprises made major investments in AI/ML last year.
We believe this is a direct response to the ongoing uncertainty and economic impacts of the pandemic. More and more organizations are realizing AI/ML is key to transforming their operations in the midst of ever-changing conditions, no matter the root cause.
This is evident in the use cases that bubbled up as top priorities. Enterprises are laser-focused on AI/ML initiatives that drive bottom-line cost reductions and top-line revenue growth. That’s why automation and customer experience were the clear leaders in AI/ML use cases last year. Automating business processes with machine learning can make them more efficient and reduce operational costs. Compelling customer experiences — from acquisition through loyalty and retention — form the foundation of growing, sustainable revenue streams.
Significant Challenges Remain
This is all very exciting, as it points to a growing number of enterprises not only adopting AI/ML but achieving tangible results. Yet many organizations still face real challenges in their implementations, even as investments and staffing soar.
Enterprises still run into hurdles in some of the fundamental phases of the ML lifecycle. This limits their full potential when it comes to deploying, managing, and integrating AI/ML technologies.
Organizations still struggle with basic integration and compatibility issues at the outset of ML implementations as they try to get their first models into production, for example. How do I get all of these technologies talking to each other? How do I begin integrating them into my existing business processes and tech stack? And how do I enable my IT organization to rapidly adopt this emerging technology and bring it to market?
We’re at an inflection point: There is a second major wave of AI/ML adoption underway, and enterprises that are just getting started can still get in on it.
Longer term, enterprises stumble when it comes to governance, which surfaced as one of the top challenges last year. How do I actually manage all of this in an efficient, compliant manner? How do I align and integrate our AI/ML portfolio with existing business processes? And how do I maximize the business value of our deployed models over time?
Governance is not simply a matter of regulatory compliance and audit risk, either. It’s the bedrock that helps protect the company’s bottom line and brand. Organizations with strong AI/ML governance better understand all of the variables that might affect model results. This, along with high visibility and granular control, allows these companies to quickly identify and mitigate potential issues such as model drift or degrading performance that can negatively affect outcomes — and potentially jeopardize customer trust and the overall health of the business.
Especially in larger organizations, a lack of organizational alignment can be a major cause of these challenges, especially around governance. That’s because AI/ML requires true interdisciplinary collaboration. AI/ML maturity is virtually impossible to achieve without organizational alignment.
Technical Debt Is Piling up
This massive investment in new technologies, methodologies, and people — in the midst of significant ongoing challenges that still need to be solved — is also causing technical debt to grow.
Organizations are actually spending more time and resources getting models into production, even as AI/ML gets a bigger share of organizational focus and budget. Models still take too long to get into production: Just 11% of enterprises can put a model into production within a week. And model deployment still sucks up far too much of data scientists’ time: At 38% of organizations, data scientists spend more than half of their time in this operational area.
Both of these trends are moving in the wrong direction, too. We saw models taking longer to get into production compared with our previous report’s data, while data scientists are spending more of their time on model deployment.
We think this points to an underlying issue that needs to be addressed: Too many organizations are taking highly manual approaches to scale their AI/ML initiatives. They’re throwing more people and money at core needs like model deployment. They would be better served maximizing the return on their growing investments by using MLOps to improve operational efficiency.
What This Means for 2021 and Beyond
We expect the significant increases in AI/ML investments last year to fuel a palpable urgency for these initiatives to begin producing results in 2021. That means continuing to invest in and scale up initiatives while also tackling those barriers to success head-on.
It also means there will be a growing gap between the AI/ML “haves” and “have-nots:” Enterprises that are proactive will build a sizable — and eventually insurmountable — lead over those that remain on the sidelines. Organizations need to act now to remain competitive.
We’re at an inflection point: There is a second major wave of AI/ML adoption underway, and enterprises that are just getting started can still get in on it. A robust MLOps platform can even help them speed up their initiative. In fact, that was another key finding in our latest report: Enterprises are experiencing improved outcomes when they use a third-party MLOps solution.
All organizations — even those with dozens of models already in production — need to invest in operational efficiency and scale in 2021. This is the key to getting the greatest return on your ML investment and generating high-value results for the business. As a result, this will be one of the major areas of strategic focus this year — and why 2021 will be the biggest year yet for enterprise ML.
Feature image par jplenio de Pixabay.