A few months ago, Babson College professor Tom Davenport convened a virtual meeting of chief data and analytic officers (CDAO) from a variety of industries to discuss how to best achieve a return on investments (ROI) in AI. “We wanted to learn to what degree these senior leaders shared our perspective that AI faces an important economic return issue, and what, if anything, their companies were doing to address it.” Many participants said that a decent ROI remains a critical issue for AI projects.
Davenport has long been analyzing how companies should build up their AI capabilities to achieve their business objectives. For example, in a 2018 Harvard Business Review article he co-authored, Davenport advised companies to build their AI capabilities through the lens of business opportunities, rather than technology. As has been generally the case with new technologies, highly ambitious, multi-year moon shot projects are less likely to be successful than low-hanging-fruit projects. Business process automation is one of the least expensive and easiest capability to implement, since companies have long been engaged with enhancing and automating their business processes.
The following year, Davenport talked about the state of AI in the enterprise at the annual conference of MIT’s Initiative on the Digital Economy His talk was based on recent surveys by Deloitte of executives in US-based companies who were involved in AI projects. The surveys found that 20-30% of enterprises were early adopters, having implemented at least one AI prototype or production application; most of the projects were in pilots with relatively few already in production; simple projects prevailed over more ambitious and complex ones; and implementation, integration, data issues and talent topped the list of challenges faced by these early adopters.
Transformative technologies are prone to hype cycles, - remember the dot-com bubble. All the excitement and publicity accompanying their achievements often leads to inflated expectations, followed by disillusionment if the technology fails to deliver. But AI may well be in a class by itself, as the notion of machines achieving or surpassing human levels of intelligence, - let alone science fiction books and films, - have long led to feelings of both wonder and fear.
“Let’s face it - we’ve all benefitted from the hype around AI. But now we have to deliver,” said one of the CDAOs at the recent virtual meeting. Business stakeholders may expect outsized impact from AI given these hype levels. Organizations often spend too much time and emphasis on AI tools, technologies and models, and not enough time on the measurable, incremental value of AI projects, said another. As a result, project teams may not have sufficient funding to be effective. The benefits from AI often take quite a bit longer than anticipated, given their dependence on access to large volumes of data, process and systems integration, and a supportive organizational culture.
There was widespread agreement that “return on AI” is a serious issue. As one CDAO said, “We’re describing AI as a progression from data to insights to outcomes,” and without outcomes AI simply doesn’t matter. Some of these ROI issues would be faced by any emerging technology, but some are specific to AI because expectations are so high. As is the case with most technologies, it’s easier to measure the value of AI projects involving operational improvements and cost savings. But it’s much harder to quantify the more complex benefits of AI, such as better insights and decision making.
In 2018, PwC released a report on the long-term economic value of AI. The report’s overriding finding was that AI is the biggest commercial opportunity for companies, industries and nations over the next few decades. PwC estimated that AI advances would increase global GDP by up to 14% between 2018 and 2030, the equivalent of an additional $15.7 trillion contribution to the world’s economy.
Around 40% of the expected GDP growth was projected to come from productivity gains, especially in the near term, including the automation of routine tasks, and the development of sophisticated tools to augment human capabilities. Companies that are slow to adopt these productivity improvements will find themselves at a serious competitive disadvantage.
Over time, increased consumer demand for AI-enhanced offerings will overtake productivity gains and result in an additional $9 trillion of GDP growth by 2030. Moreover, the report predicted that network effects, - that is, the more data and better insights companies are able to gather, the more appealing the products and services they’ll be able to develop, - will further increase consumer demand. AI front-runners will gain an enormous competitive advantage through their ability to leverage this rich supply of customer data to shape product developments and business models, making it harder for slower moving competitors to catch up.
Another 2018 report on the economic impact of AI, this one by McKinsey, predicted that AI has the potential to incrementally add around $13 trillion by 2030 to current global economic output, - an annual average contribution to productivity growth of about 1.2% between 2018 and 2030. McKinsey’s growth estimates are fairly similar to PwC’s, which is noteworthy given that they’re each based on different data sets and on different modeling and analysis methods. McKinsey’s models showed that AI marketplace adoption will likely follow a typical S curve pattern, - that is, a slow start in the early stages, followed by a steep acceleration as the technology matures and firms learn how to best deploy it.
But, despite their expanding experience with AI initiatives, companies still face significant obstacles in their development and implementation. In the 2018 HBR article referenced above, Davenport recommends a four-step framework that will help companies achieve AI’s considerable economic value, - whether the projects are business-process enhancements or moon shoots. These include:
Understand the Technologies. Before embarking on an AI initiative, it’s important to understand the strength and limitations of the available technologies. This requires employees with the proper skills, including data analysis algorithms. Offer AI training to augment the skills of the workforce and help them transition to new jobs.
Create a Portfolio of Projects. Companies should develop a prioritized portfolio of projects based on their needs and capabilities. This includes determining which areas of the business would benefit most from AI projects, which use cases would generate the most value, and whether the available AI tools and skills are up to the task.
Launch Pilots. Given the experimental nature of most AI applications, companies should create pilot projects with a limited scope before rolling them out across the entire enterprise. Do think big, - figure out how AI can transform business processes, business models and overall strategy over time, - but start out with less ambitious pilots.
Scale Up. Scaling up an AI application will require integration with existing systems and processes, as well as close collaboration between technology experts and the owners of the business process being automated. Integrating AI into the rest of the business is often the greatest challenge in AI initiatives.
“It’s clear that the broad and prominent visibility of AI in the world today has created both an opportunity and a problem for organizations hoping to achieve return on the technology,” noted Davenport in conclusion. It’s important to keep expectations under control, to monitor and publicize success and actual results, and to minimize fears of job loss or major changes in skill requirements.”
Comments