AI’s potential is enormous. It can transform the nature of work, making tasks more efficient and jobs more fulfilling. It can help executives collaborate better, foresee problems, predict outcomes, and make better business decisions. But if it’s not used responsibly, AI may introduce errors and bias or misuse proprietary information.
To succeed with AI, companies must use and develop products responsibly, keeping humans at the center of all decisions. In this article, Workday Co-President Sayan Chakraborty explains how organizations and regulators can work together to forge a path allowing business leaders to reap AI’s many potential benefits while incorporating guardrails to avoid negative economic or other impacts.
Managing AI’s limitations
AI’s extended reach through chat applications and browser tools has thrust its rapidly evolving capabilities into the limelight. Generative AI has impressed many with its ability to quickly answer questions, explain complex processes, create images, or supply code to help developers build new products.
Companies are eager to extend their use of AI and gen AI, but they also worry about the risks. In a 2023 Workday AI IQ report on AI in the enterprise, 39% of senior leaders listed potential bias as a top concern for AI deployments. Other top-of-mind risks include data privacy and security, concerns on accountability, the inability to measure ROI, and decision-making errors.
“People are afraid of artificial intelligence for some very understandable reasons,” Chakraborty says. “We should be concerned and thoughtful about how we adopt this technology.”
AI is different from ordinary computing because it is designed to mimic the way the human brain works. Our brains don’t just memorize facts—they make fact-based assumptions about new information. Similarly, AI uses machine learning algorithms to absorb massive amounts of data, then make inferences and predictions.
“Not only does it use data to predict things like statistics would—it’s also modifying itself in response to that data to get better at predicting things,” Chakraborty explains.
Humans are an essential part of the process, training algorithms to distinguish a can of tuna fish from a can of cat food, or a pothole from a shadow on the road.
Eventually, algorithms can learn to make such distinctions themselves. They’re not perfect—they can only be as good as the data they’re trained on, which may contain errors or bias. But over time, as they analyze and receive training on larger data sets, their accuracy improves.
In the AI IQ survey, only 29% of leaders said they are confident that AI is being applied ethically to businesses today, but 52% are very confident it will be applied ethically in five years’ time.
To make that happen, however, companies must establish a sound framework for governing the technology.
Monitoring evolving regulations
In the U.S., widescale AI regulation has not yet developed, though New York City and a handful of states have recently passed laws governing the use of automated tools in hiring and promotion decisions.
At the national level, an act of Congress created the National Artificial Intelligence Advisory Committee (NAIAC), a coalition of experts from business, academia, and nonprofits. This organization is responsible for advising the Office of the President on AI and making recommendations for its use. The White House recently released an executive order on safe development and use of AI detailing eight guiding principles for responsible use.
Chakraborty, who is a NAIAC member, says government will play a key role in shaping the technology’s future.
“There is important work going on in government,” he says. “As companies adopt AI solutions, they need to understand how models are going to be used. What are the potential sharp edges? Do you have human oversight and accountability?”
Using AI responsibly
To get ahead of new and potential regulations, business and IT leaders should examine their procedures in light of NAIAC recommendations, practices suggested by the National Institute of Standards and Technology (NIST), and research from other reputable organizations, using the information as a framework to establish their own guidelines for using AI responsibly. Doing so is important for any company using AI, but especially for those that develop AI-based products.
“Traditionally, technology providers have viewed their responsibility as ending when they deliver a solution. That cannot be the approach here,” Chakraborty says. “Vendors must put in guardrails and auditing capabilities so that people understand how it’s being used, where it’s being used, and how to tune it and use it correctly.”
At Workday, this approach has translated to not only putting humans in the loop, but keeping them at the center of all AI decisions.
“We are grounded in our values as a company, and that dictates how we deliver artificial intelligence—to augment people and help them realize their potential,” Chakraborty says. “How do we make jobs better, easier, more fun, more fulfilling? How do we take some of the drudgery out, so instead of having to go to 20 different places to piece together information—some of which is out of date or inapplicable—it is all synthesized for you?”
Responsible use also extends beyond ensuring that solutions are helpful and well-functioning.
“It requires legal experts to ground us in the law and ethics experts to help us understand the ethical boundaries. We bring those cross-functional teams together anytime we are adding AI to an existing product or creating a new product,” Chakraborty says.
Another essential element to consider in deploying AI is data quality, which is critical to solution accuracy. At many companies, finding, collecting, and managing clean, useful data to incorporate into AI solutions is a constant challenge. Workday’s large customer base gives it an advantage here.
“Workday has an incredibly high-quality set of data from our 10,000-plus customers and our 60 million users,” Chakraborty says. “Our focus is on training our large language models on the high-quality data we have, versus reading a bunch of chat groups on the internet. At the end of the day, you have to have accountability.”
In a broader sense, accountability is the key not only to solution performance, but to the technology’s long-term viability. For AI to succeed, humans must work closely with the technology to apply solutions ethically and maintain control of any decisions recommended by the algorithms. Only then will AI earn the trust it needs to drive enterprises toward a more productive and rewarding future.
Learn how Workday’s finance and HR platform, with AI embedded at the core, can help your organization redefine how work works.