Agentic AI is expected to drive a leap in productivity and innovation. But how can companies harness its potential without risking their values, security or profits?
The Financial Reporting Council’s latest Review of Corporate Governance report reveals that 73 percent of companies are now addressing AI in boardroom discussions, whether it’s about risk, opportunities or operational integration.
Meanwhile, ISS Corporate’s analysis of S&P 500 proxy statements shows a dramatic increase in board-level AI oversight: there’s an 84 percent jump in AI-related board disclosures between 2023 and 2024, and a more than 150 percent increase from 2022 to 2024. Clearly, AI is no longer a back-office experiment: it’s becoming central to governance.
Sebastian Weir, AI, analytics and automation practice leader UKI at IBM Consulting, believes that AI has vast potential to elevate the quality of governance and decision-making ‘by streamlining data analysis, enabling faster and more informed decisions and enhancing risk assessment through anomaly detection and operational monitoring’.
He highlights how AI goes beyond analytics, saying that ‘it supports strategic foresight through more accurate predictive modelling based on historical data, strengthens regulatory compliance by proactively identifying concerns and ensures transparency with timely, automated reporting.’
‘Through sentiment analysis, AI can also boost stakeholder engagement, and it can facilitate ethical decision-making by integrating fairness principles into governance frameworks,’ he adds.
As for agentic AI, which can make independent decisions despite being public-facing, Weir says there is a need to evolve. ‘Current models of corporate governance may not be fully equipped to handle autonomous decision-making agents, as they often focus on human decision-makers and their accountability,’ he explains.
Moreover, the introduction of autonomous agents, such as those enabled by agent-based modelling techniques, ‘requires a shift in governance to ensure these agents align with business objectives, mitigate risks, and maintain transparency,’ he says.
That requires new decision frameworks, new boundaries of authority and updated evaluation metrics.
Sebastian Weir, AI, analytics, and automation practice leader UKI at IBM Consulting
One of the toughest questions to answer is where accountability lies when AI systems act with partial or full autonomy.
‘In the current stage of adoption, reinforcing existing accountability frameworks and regulatory requirements is critical – regulators will not absolve a board of their accountability because agents have informed their decision,’ he adds.
So, what should the next generation of governance mechanisms look like? Ones that provide full visibility into AI systems, implement robust guardrails across the entire AI lifecycle and ones that ensure transparency, security and ethical design and deployment are seemingly key.
Weir also sees automation playing a central role, with automated governance, risk detection, and mitigation – as well as enhanced agentic AI evaluation and lifecycle governance – as essential components of these new mechanisms.
‘By implementing these governance mechanisms, organizations can build trust in AI systems, mitigate risks associated with biased algorithms and maintain compliance with evolving regulations,’ he says.
But is a new shared model of liability emerging, or does accountability remain entirely with the corporation?
‘Today, corporations must be held fully accountable for the role and jurisdiction of “autonomy” they choose to enable their AI agents to hold within the business,’ Weir explains. ‘This forces boards to truly engage with the extent and remit AI agents will have within decision making.’
As for the future, he says that in time, we will see the concept of shared liability evolve and increasingly, fully autonomous organizations will emerge that ‘will test the customer sentiment of adoption and the regulator’s ability to react at the requisite pace to control and protect’.
McKinsey’s latest Global Survey on AI found that only 28 percent of organizations say their CEO oversees AI governance, and just 17 percent cite board-level involvement. On average, two leaders share responsibility, often with minimal AI expertise. This creates a gap and a growing need for a dedicated role.
With AI at the core of many organizations’ competitive strategies, the board’s tech literacy is increasingly under the spotlight. A deep understanding of these technologies is essential for fulfilling fiduciary oversight responsibilities.
So how can companies develop the skills and build the trust architecture needed to govern AI responsibly?
Weir believes that corporations can do this by ‘establishing robust governance frameworks that incorporate ethical guidelines, data integrity, security, explainability and fairness, all overseen by specialized teams including ethicists and technologists’.
But that approach must be strategic, not a full hand-off, so instead of outsourcing governance entirely, a hybrid approach is recommended ‘maintaining core oversight while tapping into partners for targeted and deep expertise, ensuring alignment with corporate values and regulatory compliance’.
The bottom line is AI is no longer optional in the boardroom, and neither is understanding how to govern it. Boards must catch up, or risk being left behind by the very systems they’ve enabled.