Counterpart's Hackett: Agentic AI is inevitable but human judgement remains crucial

By Mia MacGregor

April 4 - (The Insurer) - Agentic AI has the potential to transform the insurance industry but a strong “human in the loop” approach is required to mitigate risks, Counterpart CEO Tanner Hackett said during a panel discussion at the InsurTech NY conference.

“The future looks more agentic than it does just traditional insurance with human underwriters,” said Hackett on Thursday. “There's an insatiable thirst for more tools. Now people see the value, and it's existential for companies if they don't embrace this.”

Agentic AI refers to systems and models that can act autonomously to achieve goals without constant human guidance.

However, Hackett stressed the importance of human oversight in AI-driven insurance. “The key piece that is often missed, especially by many of my peers in insurtech, is the ‘human in the loop’, the validation aspect of insurance,” he said.

“Agentic insurance means bringing together technologists and industry experts, actuaries, underwriters, claims managers, who can recognise anomalies in the data and course-correct quickly. When combined with rich datasets, AI can synthesise information in ways that were previously unimaginable.”

Kasey Roh, head of U.S. business at Upstage, also noted the value of human validation in AI-driven processes.

“For AI to replace a lot of manual, repetitive work, we have to ensure there’s validation in the process,” Roh said. “Insurance is an industry where you care and can’t allow small errors.”

She explained that at Upstage, “we embed a very efficient, human-like review process within the automation stage".

"This ensures that as data is processed and synthesised, it still goes through the human mind process in terms of validation and confirmation,” she continued.

Daniel Bodnar, chief information and technology officer at Skyward Specialty, said that “knowing where to start is a huge challenge” when talking about data readiness.

“Our strategy is to let use cases guide the process while heavily incorporating feedback loops to enable rapid iteration,” he said.

Bodnar also highlighted security and privacy as foundational concerns.

“How do you adopt AI in a company if employees don’t trust that their data is secure and private? The initial challenge in adoption is building trust, having humans in the loop to ensure things are going the right way," he said.

Girish Arora, vice president at Nationwide, underscored the importance of regulatory compliance in AI implementation.

“Regulatory compliance is critical,” Arora said. “When it comes to ethical and responsible data usage, bias, and hallucinations in AI models, these will become crucial considerations as we deploy these tools and models.”

Hackett concluded that businesses must not resist the inevitable shift toward agentic AI.

“You shouldn’t obstruct the inevitable, which is agentic AI,” he said. “I’ve seen it before in underwriting, companies refusing to write policies in California because of a bad claim from 12 years ago. Now I’m seeing it again with data.”

He said that while AI adoption carries risks, the right guardrails and human oversight can mitigate them.

“There will be those who move forward, understanding that adopting AI requires a calculation of risks,” Hackett said. “But if you put in the right safeguards and have the right people overseeing the process, you can and should embrace it, because the consequences are very meaningful if you don't do it.

“Companies like ours, data-native and AI-native, are growing rapidly. The best outcome for customers is for everyone to move forward together. But for those who don’t, they will be left on the side of the road.”