Lack of AI implementation may have cost enterprises $4.26T, Signal AI finds

AI’s potential impact on the U.S. economy could reach into the trillions of dollars, according to a report published this week.

Signal AI, which offers a decision augmentation platform infused with AI, interviewed 1,000 C-suite executives in the U.S. for the study. The report found 85% of respondents estimate upwards of $4.26 trillion in revenue is being lost because organizations lack access to AI technologies to make better decisions faster.

According to the Signal AI survey, 96% of business leaders said they believe AI decision augmentation will transform decision-making, with 92% agreeing companies should leverage AI to augment their decision-making processes.

More than three-quarters of respondents (79%) also noted that their organizations are already using AI technologies to help make decisions.

In general, 96% of business leaders said they believe they can leverage AI to improve their business decision-making processes, with 80% noting they already feel they have too much data to weigh when making decisions. On average, 63% of respondents said they spend upwards of 40 hours a week on decisions.

Reputations and expectations

More than two-thirds of respondents (69%) ranked data higher than instinct in terms of influence on business decisions, even though many execs have been skeptical of the quality of data being employed within analytics and business intelligence (BI) applications.

Arguably the most surprising survey result is that just over 85% ranked reputation as a bigger priority than profit margins, Signal AI CEO David Benigson said. There’s a growing appreciation for the impact reputation has on both profitability and revenues, he noted.

But some business leaders may have unrealistic AI expectations, Benigson reported. “Just like with other technologies, they are overestimating the impact of AI in the short term and underestimating it in the longer term,” he said.

Estimating the potential revenue impact of AI is an inexact science. But a lot of complex business processes are occurring in near real time that are impossible for humans to optimize with AI augmentation. The challenge is building AI models that accurately reflect those business processes. Many of the data science teams that have been hired to build AI models lack a deep understanding of the process they are being tasked with automating. Many AI models, as a consequence, never get deployed in a production environment.

Nevertheless, the volume of AI models being deployed continues to increase. The next big challenge for organizations will be the maintenance of all those AI models, many of which are subject to drift as new data sources become available. This means an AI model may not be as efficient as it once was because it needs to be retrained or replaced altogether.

Regardless of the path forward, AI models will increasingly become just another type of artifact to be incorporated into the application development process. The challenge will be aligning the efforts of application developers with the data science teams that build AI models to ensure neither is waiting for the other to finish a project before an application can be deployed.

In the meantime, business leaders may want to temper their AI expectations. Implementing an AI model is roughly akin to hiring a junior member of a team that needs some time to learn how processes work. Unlike a human, however, that AI model never takes a day off, quits, or forgets what it learns unless it is retrained. The only real issue is that when an AI model does make a mistake it may be at a level of scale that is difficult for the business to recover from unless the proper guardrails are in place.