Enterprise AI adoption on fire, but foundation lacking for sustainable AI

McKinsey surveyed thousands of executives on their AI projects to find out what was working and what wasn't. Here's the lowdown.

It's easy for CIOs to get caught up in the hype around AI. However, successful early adopters are focusing on applying AI technology in places where it can enhance traditional analytics, said Michael Chui, partner at the McKinsey Global Institute, at the O'Reilly Artificial Intelligence Conference in London.

That said, most CIOs are not doing much to create the infrastructure required for sustainable AI.

"In some ways, we have seen an acceleration of [AI] adoption," Chui said. The number of enterprises adopting AI as part of core business processes jumped to 50% -- up from 20% a year ago. "Yet, there is a bit of a soft underbelly in which some of the foundations necessary to capture value in a sustainable way going forward are not there," he said.

One of the big challenges facing CIOs is the disconnect between how businesses can viably create value from AI and how it is presented in science fiction and exaggerated in the mainstream press. Technologies like deep learning can be deceptive in that executives hear the term and imagine that AI can learn new things, which can be immediately applied to improve the business.

"It's not like that," Chui said.

To cut through this disconnect, McKinsey interviewed thousands of executives. The management consulting firm identified over 400 potential use cases for AI to get a better understanding of the potential value of different approaches, particularly around deep learning and simple analytics.

Where AI adoption is succeeding

Michael Chui, partner at the McKinsey Global InstituteMichael Chui

There is a lot of potential value from cutting-edge use cases. But, today, most of the value comes from improving the kinds of analytics enterprises are already doing, according to unpublished research from McKinsey.

"If you want to know where AI has the potential to create value in your enterprise, follow the money," Chui said.

For example, if sales and marketing efforts benefit from using analytics to target customers, then AI is a good use case. Or, if the value of analytics for a company comes from increasing operational effectiveness through predictive maintenance, then that is probably where AI should be applied to drive additional value for the business.

The good news about AI adoption: Some corporate executives appear to understand the principle of deriving value from AI by using it to build on their companies' current analytics strengths.

For example, AI adoption to enhance service operations is led by the telecom sector, followed closely by travel, transportation and logistics -- sectors where service operations are critical to success. The adoption of AI to enhance product development is led by the high-tech sector. The retail industry leads in applying AI to both marketing and sales, as well as to supply chain optimization. The automotive industry leads in applying AI to manufacturing. The financial services industry leads in applying AI to risk management.

Data infrastructure lacking

However, McKinsey found mixed results when it came to the adoption of practices for capturing the value of AI in a sustainable way. One big challenge is in constructing the data infrastructure that makes it easy to test out new AI use cases and scale the successful ones. McKinsey recently published an interview with computer scientist Andrew Ng, who said companies that are serious about AI are playing multiyear chess games in order to acquire the data they need to train their AI models.

But most companies are not that serious yet. Only about a third use data effectively to support the goals of AI adoption, according to McKinsey research. Less than a fifth of organizations have a clear strategy for accessing and acquiring the data they need for doing AI work. And less than one-tenth have all relevant data accessible to AI systems across the organization.

"We do see potential and adoptions," Chui said. "But when you want to talk about moving to AI at scale, data practices have a lot left to be desired."

AI leadership practices required

If you want to know where AI has the potential to create value in your enterprise, follow the money.
Michael Chuipartner at the McKinsey Global Institute

Another big component of sustainable AI adoption revolves around having the right leadership practices in place. Only about a quarter of enterprises surveyed by McKinsey demonstrate true ownership of AI initiatives. Less than a fifth have mapped out where all the potential AI opportunities lie.

In the long run, successful AI adoption will require companies to continually refresh their AI opportunity maps as new technologies emerge and the competitive landscape shifts. This involves running an effective, continual process for developing portfolios of the most valuable AI opportunities. 

Making AI actionable is a people thing

Another key part of adopting AI across the enterprise concerns people. For AI adoption to succeed, CIOs and their companies need to figure out how to attract or grow AI talent and make AI useful for frontline workers. Only about a quarter of organizations have access to internal and external talent with the right skill sets to support AI work. Only about a quarter of organizations where AI has been adopted have found a way to integrate it into day-to-day operations.

One big problem is only about 16% of employees trust AI-generated insights. Only about 6% of organizations embed AI into the formal decision-making and execution process of frontline workers.

"All of us know it is fun to solve tech problems," Chui said. "The real problem is changing how people operate."

In the long run, there is a lot of foundational work that needs to be done in order to capture new value from AI at scale and in a sustainable way.

"Unless you build that foundation, you cannot keep building forever," Chui said. "You eventually need to adopt some of these practices in order to build AI at scale."

Dig Deeper on Digital transformation

Cloud Computing
Mobile Computing
Data Center
and ESG