Artificial intelligence may be unique in the way it raises ethical and governance questions for boards—and may need equally original answers.

The response will include formal and structured governance models, but AI will also require people—humans in the loop—to ask deeper questions about the impact of using AI and the way in which the technology answers questions.

Those were the key observations of a special panel discussion hosted by Board Agenda and Diligent, the board technology providers, exploring AI and its increasing use by boards.

Maria Axente, a former head of AI public policy at PwC and currently part of an AI advisory board at NATO, said the technology’s ability to work without human supervision, adapt to the external environment, and its probabilistic method of operation made it a special case for governance.

“If we start to understand what would distinguish AI from previous technology—its probabilistic nature—we start to understand why we have to have a different approach,” she said.

Ethics and governance

Artificial intelligence—in all its forms—has become a fixture of business operations but now is increasingly part of boardroom deliberations and a tool fast becoming embedded in the governance process. Many companies now summarise board papers using AI while, at the extreme, some may be consulting AI on business strategy.

Either way, its use raises questions of ethics and governance, especially at a time when the information presented to boardroom members is vast.

Hasintha Gunawickrema, a non-executive with Skipton Group, a building society, and formerly a senior executive with HSBC, says the use of AI in boardrooms touches on fundamental questions of boardroom culture and accountability.

“We all have to operate in an ethical way—wherever we sit in an organisation and especially at board level because… you’re responsible for everything that happens within the firm,” she said.

One way of dealing with AI is through a formal and structured governance model such as an “approve to act” framework, which ensures AI systems used for high-impact decisions are reviewed, authorised and monitored by appropriate individuals or bodies.

In practice, it means identifying “use cases”, the AI systems that may be involved, the risks raised by the systems and who would be involved in authorisation and monitoring.

April Skipp, principal product compliance adviser and governance thought leader at Diligent, said that, without monitoring, “boards do risk those accountability gaps where no one owns the bad outcomes”.

This is especially important, she stressed, as AI moves to “agentic” models: proactive systems that can make decisions autonomously to achieve identified goals.

“When you’ve got autonomous agents, you need to have that ‘approve to act’ element built in to ensure that you’ve got a little element of control over the very end of the process,” Skipp added.

Human agency

Approve to act will help with regulatory compliance, Skipp noted but, because it keeps a human presence in the process, it also helps maintain skills and expertise necessary for judgement calls.

“It’s multifaceted,” she said.

According to Ruth Odih, head of secretariat at Centrica, a formal governance framework offers a number of benefits when using AI, including an audit trail. Centrica even keeps track of “prompts”—the instructions deployed to use genAI. This allows governance professionals to see if prompts on successive iterations can produce the same response.

But for Odih, there is also an emphasis on people in the AI process.

“One of the things our board is very conscious of…when decisions are made using AI, is: ‘who’s the human in the loop and have they done the thorough check they should have done?’”

But she also echoed the view that one ethical consideration of AI is the impact on employees and employment.

This, Odih notes, might be a question raised by section 172 of the Companies Act, covering directors’ duties.

“Companies need to be thinking broadly when they think about how they use AI to ensure that core knowledge, ways of working, ways of thinking through problems and finding solutions, are not lost because people are just moving straight to AI to find easy answers,” she said.

However, use of AI remains a work in progress for many senior leaders and organisations. One Diligent survey found that 66% of directors have uploaded board papers to open-source AI models such as ChatGPT, a clear security risk.

Maria Axente warned that experimenting with AI was necessary but business leaders would have to ensure it was done safely. “We should encourage people, on the one hand, to be creative and innovative; but, at the same time, provide them with the safeguards that allow them to innovate safely and don’t increase the overall risk.”

The issue raises a question about boardroom knowledge of best practices and governance policies.

Hasintha Gunawickrema said: “What is happening in some places I work with—either as a board member or a board adviser—is that the general counsel and the HR team are continuously sharing best practice ideas, reading materials, coaching them, getting third parties to do presentions.

“So, continuously improving their [board’s] knowledge.”

Enter the CIO

Gunawickrema also believes AI may demand adjustments in boardroom personnel.

“We need to bring the chief information officer into the board space now. They’re the subject matter experts,” she said.

For Ruth Odih, it may be early days yet for a board to include an AI expert because the technology’s use is still evolving.

But she said board members should have “access” to the right executive. At Centrica, non-executives have held AI discussions with executives outside board meetings. “It’s making sure we, as governance professionals, are making them aware of what’s available to them.”

Use of AI is still evolving as is its governance, and, in turn, the use of AI in governance.

Frameworks like “approve to act” are emerging but there remain broader questions about AI’s wider impact on workforces and institutional knowledge. Those questions may go to the heart of boardroom responsibilities.

One thing is clear: use of AI will appear on risk registers of companies in all industries. Boards will need to maintain their focus and replenish their knowledge constantly.

Diligent hosted this event, which was supported and moderated by Board Agenda. You can learn more about how Diligent is enabling the positive application of  AI in the boardroom here.