With the endless promotion of artificial intelligence by analysts, media, and vendors, one can be forgiven for assuming AI is proliferating across and running enterprises far and wide. However, the reality is beyond simply automating narrow applications — such as credit scoring, upselling recommendations, chatbots, or managing machine performance — AI still has a limited range, and barely begun to achieve its full potential as a true augment to human intelligence and talent.
That’s the takeaway from recent panel discussion hosted by New York University Center for the Future of Management and LMU institute for Strategy, Technology and Organization, joined by Daron Acemoglu, professor at MIT; Jacques Bughin, professor at the Solvay School of Economics and Management; and Raffaella Sadun, professor at Harvard Business School.
“I am not that excited about the narrow applications of AI,” says Sadun. “I’m not excited about the dumb approach of automating one process and them claiming that you are on a different tier of technology.”
At the moment, “AI is being pushed too much into automating narrow tasks, where there are not a lot of productivity gains and a lot of displacement,” Acemoglu agrees. For the most part, “AI is being used at a narrow process level as a cost-cutting measure. That does not really work very well at scale, and is really not the best way of using AI. But a lot of managers are doing that. Many firms investing in AI, bit doing it the wrong way.”
AI will start reaching its potential when it aligns broader human and machine intelligence, says Acemoglu. “Machines need to become a tool for the selection and presentation of the right type of information, for leveraging the skill, judgement, creativity, and social learning and situational learning of humans,” he says. “Very few organizations are doing that. There is a failure of these organizations, but more importantly, it’s a failure in the field of AI,” says Acemoglu. “Managers are following what AI leaders are telling them. And AI leaders are not discussing this type of use of AI. They are hyping up cost-cutting process-level AI use. In that sense, we are in an AI trap.”
The embrace of a broader and more meaningful AI needs to start at the top, where there is still much resistance, Sadun says. “There are too many leaders and executives who think about AI as a technology, not as about a lever for organizational change,” she explains. “And they have no idea how the technology is actually used by their employees. People who are really using AI effectively in their firm spent time on the shop floor, seeing how the technology is used by people. They get a sense of the software and the hardware change that needs to be there for people to actually leverage AI.”
Embracing AI needs to be an organizational transformation. “It requires different skills — not just coding skills,” Sadun says. “It’s much more about the soft skills. The ability to coordinate knowledge that comes from different pockets of the organization. I’m surprised that even some organizations are getting it right. It’s actually a really tough issue.”
The companies that are using AI to its full potential “are thinking about incorporating new input,” says Sadun. They are thinking about training, making sure that the technology is actually usable, and can leverage continuous self-improvement. Fundamentally, it’s about the technologies that think about a human and a machine as a team.”