November 6, 2026, (Inside AI) — Corporate boards are experimenting with artificial intelligence in the boardroom, but governance experts warn that the technology could erode director independence and create new legal exposure. Harvard Business School professor Suraj Srinivasan, who studies both AI and corporate governance, argues that AI can address longstanding complaints about director inattention and information gaps. However, he cautions that introducing AI into the board-executive relationship requires careful safeguards.
The stakes are high. Boards already face criticism for directors who do not know the business, skip pre-reading, and contribute little. Directors themselves want deeper access to company data. AI, with its tireless research capacity and ability to offer varied perspectives, could bridge that gap. But without clear rules, it could also concentrate analysis, create evidentiary records that invite lawsuits, and weaken the diversity of thought that human directors provide.
Srinivasan recommends four steps. First, companies should bring directors inside their own AI environment. This means board analysis happens under company security protocols, not on external models that may lack those protections. Second, directors need education on AI's benefits and risks. The benefits include faster, deeper understanding of the company and industry. The risks include what Srinivasan calls "cognitive surrender," the passive acceptance of AI output without applying one's own critical thinking. Overreliance could also create homogeneity in board analysis, negating the value of diverse human experience.
Third, companies should get legal advice. U.S. boards generally limit the creation of evidentiary materials that could create liability. Using AI for research could leave a record of board engagement that introduces legal exposure. It could also jeopardize attorney-client privilege claims if AI tools touch materials linked to legal issues. Fourth, boards should enforce a clear policy with norms for approved AI use, ensuring consistent security and compliance.
Read: A&O Shearman commits $1.1 billion to AI strategy over five years
Despite these risks, some boards are already experimenting with synthetic directors. International Holding Co. (IHC), the United Arab Emirates' largest listed company, has an AI-powered, non-voting board slot named Aiden Insight. Introduced last year with backing from Microsoft, Aiden Insight analyzes data and provides real-time advice on risk assessment, compliance monitoring, and other strategic issues. Srinivasan says he has seen boards experiment with synthetic directors meant to deliver the wisdom of respected business leaders, such as Steve Jobs and Warren Buffett, or more generically, "someone who knows a lot about finance."
While one cannot know whether the advice would match what Buffett might provide, the approach can be a source of fresh thinking for executives weighing strategic options. AI presents a complicated range of opportunities, Srinivasan says, which will play out differently across boardrooms everywhere. "As one of my students put it, AI will make good boards better, and bad boards worse," he says.
The legal and governance questions are not theoretical. Boards that use AI to analyze documents may create a discoverable record of their deliberations. That record could be used in litigation to show what directors knew or should have known. Attorney-client privilege, which protects confidential legal advice, may not extend to AI-generated analysis if the tool is not properly ring-fenced. Companies that bring AI inside their own environment can mitigate some of these risks, but they still need policies that define what AI can and cannot do in the boardroom.
The push for AI-assisted boards comes as directors face growing pressure to oversee cybersecurity, AI strategy, and other technical risks. Many boards lack deep expertise in these areas. AI tools could help directors ask better questions and spot patterns they might miss. But the same tools could also lull directors into false confidence, especially if they treat AI output as authoritative rather than as one input among many.
Srinivasan's warning about homogeneity is particularly relevant. Boards spend significant effort recruiting directors with diverse backgrounds and perspectives. If AI analysis converges on similar recommendations, that diversity could be undermined. The risk is not that AI is wrong, but that it is uniformly persuasive. Directors may defer to the tool's output without challenging its assumptions. That dynamic could make boards more efficient but less skeptical, a trade-off that governance experts say deserves close attention.
The experience of IHC with Aiden Insight offers an early test case. The AI board member does not vote, which limits its formal power. But its real-time advice could shape how human directors frame issues and evaluate options. IHC has not disclosed how often the board follows Aiden's advice or how it resolves disagreements between the AI and human directors. Those details matter for understanding whether synthetic directors enhance or distort board judgment.
Read: AI Law Firms Bid for Weil Assets and Lawyers Amid Dissolution
For companies considering similar moves, the path forward likely involves clear boundaries. AI can handle research, data analysis, and scenario planning. It should not replace human judgment on strategy, risk appetite, or ethical trade-offs. Boards that treat AI as a tool for better questions, rather than a source of answers, may capture the benefits while limiting the risks. Those that treat it as an oracle could find themselves less capable of independent oversight.
The broader lesson is that AI in the boardroom is not just a technology deployment. It is a governance decision. Boards that adopt AI without updating their policies, training, and legal safeguards may find that the technology creates more problems than it solves. Srinivasan's advice, while aimed at boards, applies to any executive team considering AI-assisted decision-making. The technology can improve analysis, but only if the humans using it remain engaged, skeptical, and accountable.
Shamil has a low tolerance for AI that only works in controlled demos. He covers agentic systems and generative AI with the skepticism of someone who has seen too many “revolutionary” tools quietly disappear six months after launch.
