Thomas Edison was granted a patent for the light bulb in 1880. But even though the potential of harnessing electricity was immediately apparent, it was not until the 1920s that 50 per cent of factories were electrified. Companies had to rethink how they operated, and it took decades to reimagine industrial production for a new era of electrification. But once organizations figured it out, the economy was profoundly transformed.

“With artificial intelligence (AI), we are just at the beginning of this process,” says Avi Goldfarb, a professor of marketing and the Rotman chair in artificial intelligence and healthcare. “It kind of feels like we are in the 1880s.  We are in ‘the between times’ where we can see the technology’s potential, but have not quite figured out how to make it work.”

AI has had the tech sector abuzz since 2012, when a team led by University of Toronto computer scientist and cognitive psychologist Geoff Hinton used deep learning neural networks to make enormous improvements in AI’s ability to recognize images. It brought the world’s biggest tech companies calling, and earned Hinton the informal title of “godfather of AI.”  But it was not until OpenAI released ChatGPT in November 2022 that the hype really hit a fever pitch.

In mere seconds, the generative AI chatbot can produce reems of text. It can help users conduct research and even write jokes. And though ChatGPT’s accuracy and quality is sometimes lacking, it opened the world’s eyes to the game-changing potential of generative AI. Companies have clamoured to adopt it, and for some, it has driven growth. But that has not necessarily shown up on the bottom line.

“If anything, AI been an expense. Right now, most companies are just doing what they always did, but trying to do it a little bit better,” says Goldfarb. “And because implementing AI requires investments in data and employee training, it is not worth it. You need a huge payoff to make the upfront investments worth it. And we are still figuring out what that huge payoff looks like.”

Goldfarb — who has published several books exploring the implications of AI, including 2022’s Power and Prediction: The Disruptive Economics of Artificial Intelligence — says at its core, the technology is a prediction machine. And as its predictions have gotten better and cheaper to make, generative AI has been used for an ever-increasing array of applications. It doesn’t only produce text and images, it can be applied to medical diagnostics and credit evaluation too. But the first step toward successfully integrating AI into business operations is understanding its strengths and weaknesses.

“You don't need to understand the math under the hood. Most people don't understand how internet protocols work, but businesses still thrive by using the internet,” Goldfarb says. “Similarly, you don't need to know the detailed engineering behind AI – but you do need to know its capabilities. One of its weaknesses is judgment. AI can make predictions, but humans will be needed to determine which predictions to make, and what to do once you have them.”

Goldfarb says organizations should look for learning opportunities that teach both employees and management about what generative AI is, so that they can understand which applications could have the most impact in their operations and which ones might be feasible to implement in the near term. AI training should help prepare participants to develop an AI strategy by carefully considering which frictions in an organization the technology might help remedy. Sometimes, these inefficiencies can hide in plain sight. Goldfarb cites the example of an airport concourse.

“Organizations need to recognize the things they do that are not actually about serving stakeholders, but compensating for failures,” he says. “The best airports in the world are multibillion-dollar structures, with great restaurants and designer boutiques. But all of it exists because you have to stop and wait. It is there to compensate for the failure to achieve efficient transportation. But with better prediction, you could overcome that. Organizations need to imagine how to overcome an industry’s inherent frictions, and what parts of a business are ripe for disruption because companies are failing to serve customers.”

And in the years to come, managers will also need to understand how humans and AI can work together to achieve their organization’s aims. The technology can help predict whether somebody will pay back a loan or will make an insurance claim in the future. But it’s important to remember that its outputs are only predictions.

“Generative AI is not like the Jetsons, where machines can do everything we can do and listen to us. And it is also not like the Terminator, where machines can do everything we can do and do not listen to us. AI is a computational statistics prediction machine,” says Goldfarb. “Predictions can help people make better decisions, but predictions are not the same as decisions. Decisions involve judgment. A human being decides which predictions an AI should make, and what action to take with those predictions. That is the fundamental human role in using this technology.”

A new course offered by Rotman Executive Programs will help leaders navigate the transition. Generative AI and Organizational Transformation will demonstrate what AI can do – and what it can’t. The three-day course will build understanding of the implications of generative AI, and help leaders develop a strategy to leverage the once-in-a-generation opportunities that are only beginning to present themselves. Generative AI and Organizational Transformation course will be held October 30 to November 1, 2024.