[Adobe Firefly]
“Some of the more advanced customers that we see are thinking about how AI can change their industry,” McMillan said. He draws an analogy between how high-frequency trading changed the financial services industry in the mid-2000s. “AI now has the power to transform industries,” he said. “A good example is one of our customers, a [consumer packaged goods] company. They see a day where the human interaction in place today between a buyer and a seller in their ecosystem is powered by two bots negotiating over the price of commodities.”
But achieving such a radical business evolution, McMillan noted, hinges on a critical factor often overlooked amidst the AI hype: high-quality data. “A lot of organizations feel as if they’ve fallen behind in terms of the implementation of [AI] solutions,” he said. “But the key point here is that AI is really only as good as the data it’s trained on. You’ve got to be able to trust that data.”
“If you have an average AI model operating with really good data, it can still be a fantastic outcome,” McMillan said. “But if you have a fantastic AI model that’s trained on really terrible data, then you can’t trust that outcome.”
Industry research echoes this emphasis. The Wavestone 2024 Data and AI Leadership Executive Survey, for instance, found that while 87.9% of data executives reported that data and analytics investments are a top organizational priority, only 15.9% believed the industry has adequately addressed data and AI ethics.
Laying the groundwork for AI trust
Beyond building volumes of data that is as clean as reasonably possible, building trustworthy AI necessitates a commitment to ethics and sustainability. “Technology doesn’t have a moral compass,” McMillan aptly noted. Today, data breaches and algorithmic bias are ongoing concerns.
This trust extends to the use of third-party AI tools. Rouz Tabaddor, vice president/chief intellectual property & privacy officer at First American Financial Corp., warned against blindly trusting external AI technologies. “If you rely on a third party, when agencies come to audit you, your answer can’t be, ‘Well, we got this answer from a third party and we don’t have the model,'” Tabaddor explained. Agencies will scrutinize the testing and output evaluation processes.
The stakes are high. If you get caught with a sloppy approach to AI, “you’ll lose trust of the client and the consumer. Once you lose that, that could be [greater] than any other kind of penalty.”
Information security and intellectual property considerations also play fundamental roles in building trust. Jason Urso, CTO of Honeywell Connected Enterprise, highlighted potential safety issues that could result from cyberattackers targeting industrial infrastructure.
Tabaddor added that AI systems trained on large volumes of data raise significant IP concerns: “There’s access to a lot of your key intellectual property.”
Legal frameworks are still evolving to address these challenges. Different regions are taking varied approaches, with the European Union adopting a more conservative stance compared to Japan’s more lenient approach. In the U.S., debates continue regarding the extent to which AI-assisted works are eligible for intellectual property protection. “There’s much debate at the federal courts right now on that,” Tabaddor said.
AI as a knowledge amplifier
While AI’s potential to upend industries seems persists, even incremental efficiency gains can be game-changers, particularly in industrial sectors grappling with talent shortages. AI is emerging as a powerful tool for knowledge transfer and accelerated learning.
Consider the wealth of knowledge held by seasoned engineers in a complex environment like a chemical plant. Years of experience have equipped them to diagnose equipment issues based on the subtlest of cues — a high-pitched whine from a pump, an unusual vibration — things that are difficult to codify into standard procedures.
Urso proposes tapping AI to bridge this knowledge gap. Imagine a system that allows a novice engineer to describe what they are observing: “This pump is making a high-pitched whine and vibrating excessively… What might that be?” An advanced machine learning system, trained on historical data and expert knowledge, could respond, “Based on our enterprise history, similar symptoms have indicated cavitation in the past,” and even suggest troubleshooting steps.
“Now you’re augmenting human with technology to help the human perform better,” Urso explained.
From job displacer to digital coworker
As AI continues to evolve, its role in the workplace is becoming increasingly nuanced. Rather than replacing human workers, AI is poised to augment and enhance human capabilities across various industries.
McMillan of Teradata envisions AI agents taking on routine tasks, freeing up employees to focus on higher-value work. “Being able to automate [mundane tasks] with an AI agent to augment how they do their role, enables [employees] to step up and deliver a higher level of value to the organization,” he said.
Urso of Honeywell shared a similar sentiment. “It’s less about displacing work,” he explained. “It’s more about addressing work that can’t be performed, number one, and number two, allowing people to perform expertly allows us to do things that are extraordinary and wouldn’t have been possible before.”
