AI Projects Get to Down to Business in 2020

Artificial intelligence looked very different in 2019 versus today. Technology was often relegated to “labs” within the business instead of firmly within business units, and AI/machine learning projects were typically experimental in scope, sprawling in budget, and inconclusive in ROI, if implemented at all.

But issues around the pandemic pushed AI adoption to the forefront across business units. It’s become essential for understanding the rapidly changing patterns and data sets, from the macroeconomic to the individual consumer, that characterize 2020.

But there’s still work to be done to ensure AI is implemented in a smart, successful way—and that starts by shifting away from fun, flashy projects and turning focus to the unglamorous but foundational work of augmenting existing projects and practices.

AI Pre-COVID: All Style, No Substance

AI in the enterprise was showing steady growth: the 2019 McKinsey Global Survey shows a nearly 25% year-over-year increase in the use of AI in standard business processes.

Unfortunately, not all AI implementations are created equal. Some enterprises began their AI journey by running large numbers of AI pilot projects with no specific expectations for tangible productivity on the first project go-around. Some deployments are based on bad models and poor infrastructure. This poor execution shows why proper AI planning in the enterprise is essential in order for projects to have impact. In fact, according to ESI ThoughtLab, 13% of AI projects are failing to break even. Those kinds of margins just don’t fly in 2020. So how can a company ensure its using AI effectively and getting true substance, not just style?

Work with What You’ve Got

Working to improve existing projects through the use of AI is one of the most effective ways to implement new technology while efficiently bringing big business impact. For example, most businesses we work with see organizational change when applying AI to augment human tasks, so that people can focus on more interesting parts of their work, rather than trying to invent new processes to be able to automate them.

One such example comes from a manufacturing company using automation to harness large, heterogeneous datasets and develop robust predictive maintenance solutions. Automation enables the team to plan for maintenance issues by taking data from multiple and varied sources, combining it, and applying machine learning techniques to anticipate equipment failure before it happens. The manufacturing team is able to gain flexibility to use code and a point-and-click visual interface to speed up the work process. The team can then test their machine learning models in production, deploy dashboards across several labs, and roll out future manufacturing improvements.

In general, analytics are ripe for AI implementation: applying AI or ML to the large swaths of the data in the pipeline has the potential for impact at scale in both the long- and short-term. Other processes that apply to industries across the board and can be quickly improved through AI are planning & forecasting, marketing attribution, and supply chain optimization

Democratize Your Data

The good news for leaders advocating for AI in the enterprise: ROI doesn't have to be so low, or so slow. In order to achieve results through AI, it must be used across teams and departments, not siloed to technical experts. With the right tools and company culture, organizations can leverage data from the bottom up, helping to democratize data use across teams and roles and thereby radically increasing its impact.

To encourage a culture that embraces AI, a data-first mindset needs to be pervasive throughout an organization, and this must start at the top with C-level executives demonstrating buy-in and showing all teammates and departments how AI can benefit them in their work. Teams empowered to work with data can augment data management, efficiently solve complex problems, and dramatically impact their bottom line. This means that AI has the potential to transform a company and its revenue in both the long- and short-term.

Optimize for Impact

As 2021 approaches, enterprises are strategizing on how best to move beyond the pandemic, and AI-aided projects are ramping up as a result. Before the pandemic, the ESI ThoughtLab reported 1,200 global companies across industries with an average revenue of $12.9 billion were upping their AI investments with the goal of optimizing impact.

The pandemic appears to have added urgency, according to the ESI white paper: AI investments have boosted by an average of 4.6% over the last year, to $38 million or 0.75% of revenue for each company. Through 2023, that could nearly double with an increase to 8.3% per year.

Though unglamourous, these investments are likely to focus on the core business functions that AI can improve, not the experimental implementations we were seeing pre-2020. 

Make It Human

Despite the longstanding fear that automation is coming for our jobs, we don’t think of AI as a human replacement. The goal of AI is not to eliminate the human touch, but rather to augment it and provide data-supported, human-centric solutions to big problems. AI doesn’t need to be the stuff of science fiction or stray far from human intellect and instinct. Rather, it should be reliable, flexible and the ultimate complement to the human workforce from conception to completion. When AI is implemented in a human-centric way, teams of people are able to better connect -- especially as those workforces become more remote –– while freeing up time to focus on other important, more interesting work. 

Kurt Muehmel is chief customer officer, Dataiku.