Is your organization ready for AI? Tips and more from the latest business strategy research

Stern plaza at class change

On Friday April 3, 2020, Professor Robert Seamans hosted the second annual "AI in Strategic Management" workshop, producing tips and ideas for managers and businesses from the latest business strategy research. The following LinkedIn article includes summaries of the presentations given during the workshop. 

On Friday April 3, 2020, the Management & Organizations department at the NYU Stern School of Business hosted its 2nd annual “AI in Strategic Management” workshop. What follows is a brief summary of the papers presented in the workshop by Michael Impink (NYU), Nan Jia (USC), Harsh Ketkar (Michigan), Mike Teodorescu (BC), Neil Thompson (MIT) and special guest Stephanie Beers (Microsoft). A number of others were present at the workshop, including co-organizers Prithwiraj Choudhury (HBS) and Rob Seamans (NYU).

The goal of the workshop was to convene a group of mostly junior business school professors to discuss early stage research on how organizations are using and responding to advances in artificial intelligence (AI). The workshop—originally planned to be held in-person at NYU—was held via video conference. While the format change was due to new policies designed to minimize exposure to the coronavirus, it felt fitting to have a workshop on technology be mediated by technology.

The first speaker was Harsh Ketkar (Ross School of Business, University of Michigan, How Automating Knowledge Integration Shapes Knowledge Scope). His paper explores the tradeoffs of utilizing algorithms to aid knowledge-intensive managerial tasks such as producing consultant reports for clients. For this study, Ketkar analyzes software projects on GitHub where project owners automated tasks that were previously conducted manually. He finds that automation can enhance coordination efficiency but ultimately drives organizations to provide and integrate narrower, less systemic outputs. This paper, one of the first to provide empirical evidence of the effects of cognitive automation on organizations, offers a theoretical framework to guide future research on the impact automation has on organizational design, coordination and knowledge production.

Harsh was followed by Mike Teodorescu (Carroll School of Management, Boston College and MIT D-Lab, A Framework for Fairer Machine Learning in Organizations). Mike’s paper proposes an organizing framework for selecting and implementing fair algorithms in organizations. A timely paper in light of the growth of automation within organizations, the proposed framework intends to mitigate risks from errors in automation by applying methodical remedies (i.e., “fairness criteria”) to data and automated tests to check the outcomes. The authors explore the efficacy of using machine learning algorithms in various management contexts, and why more research is needed to ensure that the benefits of machine learning outweigh the costs for organizations.

The third speaker was Neil Thompson (MIT Computer Science and MIT Initiative on the Digital Economy, The Computational Limits of Deep Learning). Neil’s paper investigates the potential limitations of Deep Learning due to the computational demands required for its enhanced performance. His article analyzes the relationship between computing power and Deep Learning models in image recognition, object detection, named entity recognition, machine translation and question answering. According to Neil and his co-authors, performance in these areas relies heavily on the exponential increases in the amount of computing power used. With further increases in computing power becoming technically or economically prohibitive, the authors predict that the machine learning community will either require substantial modifications to how researchers improve Deep Learning or a shift to more computationally-efficient machine leaning techniques.

Stephanie Beers (Microsoft) provided a keynote presentation during lunch. Stephanie discussed the ways in which Microsoft works with governments and policy makers to address the changing nature of work, including education and skills, and the responsible use of AI. High technology firms are uniquely positioned to provide feedback into governments and academia (@SueGlueck) on how they expect the technologies that they are developing to impact the labor market. Stephanie finished her presentation by highlighting opportunities for more work between academic research and businesses and policy-makers.
Stephanie was followed by Nan Jia (University of Southern CaliforniaCan Artificial Intelligence (AI) Substitute or Complement Managers? Divergent Outcomes for Transformational and Transactional Managers in a Field Experiment). Nan argued that transformational leadership-style managers (those with greater social skills who lead by motivating with a shared vision) see greater benefits from AI assistance than transactional leadership-style managers (those with fewer social skills who promote compliance through rewards and punishments). Through a field experiment with a fin-tech company, Nan and her co-authors conclude that employees coached by AI-assisted managers (both transformational and transactional) achieve superior collections on delinquent personal loans than those coached solely by the manager. However, transformational managers see the greater benefit of AI assistance. This research brings to light the value of integrating AI with human interaction so organizations can achieve optimal performance.
Rounding out the day, Michael Impink (New York University Stern School of Business, The Value of Proprietary Data in Developing AI) presented a paper exploring the impact of access to proprietary data sources on technology development within a firm and the relationship between proprietary data and competitive advantage. More specifically, this paper provides a framework for how this data affects a firm’s strategy, technology choices and overall success. The authors find that access to proprietary data benefits firms by allowing them to develop more complex technologies and AI products internally. This differentiation enables may enable firms to more effectively compete. This research has implications for larger firms, such as Alphabet and Microsoft who invest heavily in “human-level” AI, as it suggests accessing this data will lead to increased product value and firm performance. Also, these findings could influence policy, such as the EU’s GDPR and antitrust policy, because it supports the claim that access to proprietary data affects competition.