Generative AI isn't waiting for business to catch up - news.wpcarey.asu.edu

Generative AI isn't waiting for business to catch up

New research and a website give businesses tools for managing AI use — from the boardroom to frontline employees.

Walk into any meeting, mention generative artificial intelligence, and you may be surprised by the response. Some people don't want anything to do with it. Whether they see it as too intrusive, too likely to replace employees, or simply too risky to consider, their feet are firmly planted.

Others in your organization may have been using AI since ChatGPT launched in November 2022. They're enthusiastic about what multiple platforms are doing for them today — and what those tools could do for them in the future.

Wait. They've been using some form of AI in your organization for nearly four years? What for? In which departments? Where are you vulnerable? And who authorized it?

In many companies, the answer may be less reassuring than leaders expect.

Those questions are at the heart of what Associate Professor of Accountancy Scott Emett has studied since OpenAI introduced ChatGPT. He recognized early that AI would be disruptive. Today, he says, "it's a little dizzying to keep up with all the capabilities of generative AI models."

When businesses think about generative AI and how to implement it, "many feel a bit paralyzed," he says. "They don't know what governance process to put on top of the implementation."

That's why Emett and colleagues Marc Eulerich of the University of Duisburg-Essen, Jason Pikoos of Meta Platforms, and David A. Wood of Brigham Young University collaborated to create a framework that helps businesses put guardrails around generative AI — from boardroom strategy to frontline employees using it every day.

The team created the framework after seeing businesses struggle with both adoption and risk. The need has only grown. Emett observed that even if leaders could press pause on AI development, "using the models we have today, it would take several years to implement all the tasks they're capable of performing."

There is no pause button. And as the technology keeps improving, Emett says the growing use of AI across business processes is "really exciting and really scary."

That tension has recently become more visible as AI developers themselves wrestle with how fast the technology should move. "We're seeing what I call misalignment risk, where some of these AI models are going rogue," Emett says. "And the question is, what do we do about that?"

While we're used to technology doing what we've coded it to do, "one of AI's capabilities is it's probabilistic — which is also one of its biggest risks," he says. "There's always a chance AI won't behave the way we expect."

The governance tool helps businesses see where AI can be implemented effectively, responsibly, and ethically, "while addressing risks in a thoughtful way," he says.

Building guardrails

It's a lot to consider when your company is racing ahead with generative AI, slamming on the brakes, or landing somewhere in between. With so much at stake — and few leaders eager to admit they don't know where to start — Emett and his colleagues created GenAI Governance, a website that offers step-by-step guidance on safely integrating AI.

A research paper published in Accounting Horizons documents the steps the team took to build the framework. "We tried to make the framework accessible to a broad range of stakeholders," Emett says. "A board member or frontline employee can read this framework and understand what it's trying to accomplish." When the framework is thoughtfully implemented, frontline employees also understand what data AI can access and which organizational data they're allowed to expose to the technology.

The team turned to the accounting and auditing professions as a starting point because "accounting is the language of business," Emett says. "Accountants think broadly about organizations. They think about process, control, risk, and opportunity." From there, the framework drew input from more than 1,000 practitioners and academics, including generative AI specialists, auditors, regulators, and C-suite executives.

Identifying risks

The team identified five governance domains that organize the areas businesses need to consider. Before diving in, Emett recommends organizations determine their risk tolerance and define their AI objectives and goals. Once those are clear, Emett suggests businesses identify which governance domains should take priority.

Strategic alignment and control environment aligns generative AI initiatives with organizational goals, strategies, and risk appetites while establishing comprehensive governance policies. "Many companies spent a lot of time with this domain," Emett says. "When they saw how powerful these generative AI models were, businesses immediately knew they needed to implement them. But they didn't think about how the technology fit into their strategic priorities or how it would help deliver a better product or service." Because generative AI is powerful, "if you try to overuse it by putting it everywhere, you're going to run into all sorts of problems."

For Emett, AI alignment makes governance and control even more important. "Businesses need layers of control to guard against AI acting in ways it wasn't designed to," he says. "If companies aren't thinking about governance and control, they're going to expose themselves to these AI misalignment risks, which can do real damage within organizations."

Data and compliance management identifies, assesses, and mitigates data-related risks while maintaining legal and regulatory compliance.

Operational and technology management integrates generative AI into operational processes while managing technology, cybersecurity, testing, monitoring, and related IT risks.

Human, ethical, and social considerations address training, workforce effects, ethics and bias, reputational and social effects, and environmental concerns. "Social risks are incredibly important for organizations to think about," Emett says. "Are you implementing AI in a way that mitigates social or reputational risk? Are you implementing it in a way that will bring your workforce along or alienate them? Sometimes, technically oriented leaders don't sufficiently consider those risks."

Transparency, accountability, and continuous improvement makes generative AI decisions accountable and traceable, monitors changing technology and risks, and continuously updates governance practices.

The idea behind the framework isn't for organizations to dive deeply into every domain or address every risk within a domain. It is a guidance tool that helps businesses identify where AI fits their objectives, determine which risks matter most, and build guardrails around its use. "My hope is that leaders can take on that responsibility to thoughtfully consider generative AI, then put structure around its implementation," Emett says.

Seizing opportunities

One of Emett's biggest worries is whether society is prepared for the pressure the technology will place on organizations and workers. He believes generative AI will fundamentally reshape how business gets done through the opportunities it creates. "We also need to think about the extent to which organizations and our broader society are prepared for that rapid change."

Emett has seen that hesitation up close. His brother works for a large corporation where the IT department has decided not to implement generative AI because of risk concerns. "I think there's also a risk of underutilization," he says. "For companies that seem to be immobilized, this framework offers them a way to implement generative AI that mitigates risks so that they can seize opportunities."

Emett says the paper could be updated every week because of how quickly AI is advancing. In fact, the team has already launched a second version of the framework. "Although we've seen how much these models have improved over time, I feel we've just scratched the surface of what they're capable of doing within business."

Latest news

- Generative AI isn't waiting for business to catch upNew research and a website give businesses tools for managing AI use — from the boardroom to…

- The global supply chain is learning to heal itselfNew research finds that the global supply network is becoming more resilient not by getting…

- 5 wealth-building stocks to buy with an inheritance that will help build the next inheritanceAn ASU finance expert's research found that less than 50 stocks are responsible for half of the…