Agentic AI is reshaping professional liability risk, says Markel’s Desai

The emergence of agentic AI marks a significant shift from AI acting as a tool for professionals to it acting on their behalf. As businesses race to capture productivity gains, risk managers must grapple with new questions around accountability, governance and liability.

KEY POINTS:
Agentic AI now acts for advisers
Cognitive surrender’ can be a real danger
Governance more critical than ever

That’s the perspective of Bhavik Desai, director of professional indemnity at Markel International, who’s urging firms to get their governance in order before the liability issues end up in court.

“Traditional AI has largely been used as a tool to facilitate a professional delivering their advice or service,” Desai told AIRMIC Today. “But agentic AI, which is a more automated form, will change that dynamic. AI will be providing services on behalf of professionals, and that is where the liability landscape will change materially, along with the governance required to deliver it safely.”

The transition is already under way, he says. While many firms remain in the experimentation phase, others are embedding AI deeply into their operations. As adoption accelerates, so too do concerns around hallucinations, bias, privacy, intellectual property and accountability.

“Most of our insureds and our clients are at least experimenting with it, and some are further on in terms of creating enterprise-level solutions,” Desai said.

But the key question remains remarkably simple: when AI gets something wrong, who is responsible? 

Desai believes fundamentally accountability remains unchanged by agentic AI. If anything, it heightens responsibility.

“AI is being used as a tool to facilitate your professional advice. You as a professional, or as a business, are responsible for what goes out in your name,” he stressed. 

That principle might eventually be tested in court as regulators, lawmakers and judges grapple with questions around negligence, duty of care and acceptable levels of oversight. For now, however, firms should assume accountability remains firmly with them.

“All these things are very uncertain at the moment, but what’s not uncertain is that the professional adviser has a responsibility for the utilisation of AI in a safe manner,” Desai said.

“Agentic AI will change the liability landscape materially.”

‘Cognitive surrender’

One of the most striking risks he identifies is not technological but human. As AI systems become increasingly sophisticated, professionals might become less inclined to question their outputs and trust it like they would “trust a confident colleague”.

Desai calls this behaviour “cognitive surrender” – accepting AI outputs without sufficient challenge or verification.

“That concept of cognitive surrender will come through, and that is a real danger,” he warned.

The risk becomes more acute as models improve. Greater reliability might encourage confidence, but even a small error rate can still result in poor decisions and liability claims.

To counter that risk, he says, firms need robust governance frameworks, clear controls around prompts and data sources and audit trails regulators can examine when outcomes are challenged.

“Governance is really important in how risks are managed in this area,” Desai said. 

“Regulators will be looking at how professional services are being delivered, and especially when things go wrong, they’ll be looking back to see what decision-making happened.”

He also emphasised that businesses must understand “where their data is going and how it is being used”.

“Those that are able to adopt this technology safely and efficiently will be the ones that thrive and succeed.” 

Watch the full interview below.

Competitive advantage

Desai said organisations that successfully and safely adopt AI while maintaining appropriate controls could gain a significant competitive edge through greater productivity and efficiency. But he also warned that those that fail to do so might struggle to compete.

“Those able to adopt this technology safely and efficiently will thrive and succeed,” he said. “Those that aren’t will lose their competitive advantage and ultimately fall away.”

However, he stressed that successful adoption requires investment in both technology and people, including training staff to use AI tools effectively while retaining the expertise needed to challenge AI outputs.

Looking ahead, Desai expects agentic AI to transform business models, pricing structures and professional liability exposures.

“Agentic AI is the area which will change things materially,” he said.

As AI becomes increasingly embedded in professional services, he expects organisations to devote greater resources to governance and risk management.

“The productivity gains associated with AI are so great, but also the liability associated with it can be so great, depending on how you employ it.”

For risk managers, the challenge will be ensuring governance keeps pace with innovation.

Bhavik Desai is the director of professional indemnity at Markel International. He can be reached at: Bhavik.desai@markel.com

For more news from AIRMIC Today, click here.

Did you get value from this story? Sign up to our free daily newsletters and get stories like this sent straight to your inbox.