May 18, 2018 | Written by: Michael Wong
Categorized: Industry Insights
While 81% of Fortune 500 CEOs have recently cited artificial intelligence and machine learning as either very important or extremely important to their company’s future, up from just 54% in 2016*; many well-intended AI pilots continue to struggle with scaling and securing C-Suite business outcomes. In this Q&A, HBS Professor Prithwiraj Choudhury explains two crucial steps required to pragmatically scale AI pilots into economic value.
Q. Based upon your research, why are so many AI efforts failing to scale?
A. While well intended, just providing new AI tools is not going to deliver on productivity growth. First, a number of executives are not factoring in their existing operating models and concern about substitution by employees. Minimizing considerations like unionized workforces are a risk that C-Suite officers take at their own peril. Second, impacted employees often don’t have the complementary skills, such as a prior background in computer science and engineering, that can help accelerate the adoption curve of new AI tools. Consider some of the research that my team and I did with technology being developed by the United States Patent and Trademark Office (USPTO). The USPTO has developed new process technology based on a machine learning algorithm to help with prior art search and reduce the time necessary to review patent applications. I ran a randomized control trial where a group of Harvard students were asked to examine a patent with one half of them being randomly assigned the machine learning technology and the other half being assigned older Boolean technology. One key research finding was that those with Computer Science experience did better with the machine learning tool, however most end-users of AI based technology at Fortune 500 companies are not computer science majors. So, it’s up to the leadership team to determine the right re-skilling/training for their target end users who are supposed to help leverage the new AI tools to drive business outcomes.
Q. So building upon your point around the potential complementarities between human capital and machine learning, what does that mean for the pilots’ executive sponsors, who often are commercial leaders, as well as their CIO counterparts?
A. In effectively using machine learning based process technology, both sets of executives need to focus on two key aspects of human capital within their workforce, the prior domain expertise of workers and their familiarity with using advanced machine learning interfaces. Continuing advancement in new technical offerings can help enhance capabilities to drive better predictions. But for companies to secure productivity gains from these AI based prediction technologies, one should consider:
- Tacit knowledge that people have and which link with how they frame the predictive search and ultimately productivity
- User interfaces which can simplify the adoption of tools which end users (Accountants, lawyers, et al) are expected to leverage in today’s evolving digital labor work force
Q. So, what are the two pragmatic next steps that you would recommend for Fortune 500 executives as they consider their strategic AI investments?
A. First, the way I see AI in the workplace is not that different from the past, such as ERP. There is a large body of research around successful technology adoption and the secret sauce is a company’s leadership team willing to invest and re-skill its workforce. Workers without a computer science degree will have to feel comfortable manipulating the user interface of these tools. Maybe the AI development teams need to experiment with voice-based user interfaces rather than graphical user interfaces for ease of use.
Second, CEOs need to share their vision around the business value of using AI in the workplace for not only the clients and shareholders but also the employees. They need to take the workforce along this journey and explain the value of these new predictive tools. That way experienced employees will be more comfortable sharing their tacit knowledge with the machine learning development teams, without which the search strategies using the new prediction technologies will not be effective.
Prithwiraj (Raj) Choudhury is an Assistant Professor in the Technology and Operations Management Unit at the Harvard Business School. He was an Assistant Professor at Wharton prior to joining HBS. He studies knowledge worker productivity and innovation, with a focus on studying how global R&D, talent flows and AI will shape the future of work.
His research has been published or is forthcoming in Organization Science, Strategic Management Journal, Review of Financial Studies, Harvard Business Review, Journal of International Business Studies, and has been cited in the Wall Street Journal and Forbes among other outlets. He was awarded the Haynes Prize by the Academy of International Business and won the 2017 Best Paper Award in Strategic Human Capital awarded by the Strategic Management Society. He earned his Doctorate from the Harvard Business School and has degrees from the Indian Institute of Technology and Indian Institute of Management. Prior to academia, he worked at McKinsey & Company, Microsoft and IBM.
* Murray, Allan, Fortune 500 CEOs on Trump, the Economy, and Artificial Intelligence, June 8, 2017: http://fortune.com/2017/06/08/fortune-500-companies-ceo-survey/