Insurance's 212-Degree Moment: From Pilots to AI at Scale

Reimagining insurance with AI: Insights from the HFS roundtable with IBM

AI has moved past hype in insurance. The question is whether leaders can scale it beyond isolated wins and turn it into enterprise value.

At a recent roundtable co-hosted by HFS Research and IBM, industry executives cut through theory and focused on execution. Tony Filippone and Saurabh Gupta of HFS Research led the discussion with support from IBM Consulting leaders Shobhit Varshney, Neeraj Manik and Girish Ratnam. The message was clear: pilots are everywhere; scaled outcomes are rare.

Move beyond pilot purgatory

"AI is no longer an experiment, but scaling it remains the hard part," noted Neeraj Manik. The group agreed: too many pilots sit in silos, disconnected from end-to-end processes and measurable outcomes.

Claims automation is a perfect example. Point solutions speed up steps, but without integration into full workflows, carriers see throughput gains yet miss real cost takeout and customer impact. The fix: modern data and infrastructure, responsible AI guardrails, and change programs that align tech with people and culture.

3 hard truths holding insurers back

• Legacy debt: About USD 200 billion in tech, process and data debt is limiting agility and scalability.
• Governance gaps: Few carriers have enterprise frameworks for ethics, transparency and AI oversight.
• Customer experience plateau: Service is efficient, but it often lacks the sentiment and loyalty-building engagement policyholders expect.

Closing these gaps requires modernized foundations, stronger guardrails and experience-led transformation.

The industry's AI maturity check

Most leaders placed their firms at 2-3 on a 5-point scale. Experimentation is common; execution is fragmented. Only a handful have a Chief AI Officer or an enterprise AI strategy.

Momentum builds once governance matures, data becomes AI-ready and business leaders own a clear North Star. One insurer accelerated underwriting by creating a cross-functional AI council that aligned risk, compliance and operations, cutting through silos fast.

Proof from "client zero"

Keynote speaker Shobhit Varshney highlighted IBM's internal program: automated workflows, modernized HR and reengineered processes delivered measurable results, including a 79% NPS from AI-enabled experiences and billions in productivity savings. Large-scale AI transformation is already delivering outcomes, not case studies.

The 212-degree moment

As Girish Ratnam put it, the sector is at its "212-degree moment" - the point where incremental efforts must give way to systemic change. That shift moves AI from experiments to enterprise priorities.

The CIO remit is changing toward governance, integration and risk management. Technology leadership and investments must align with reinvention, not isolated tools.

The AI scale-up challenge: 12-month focus

• Invest in modern data foundations: Rationalize core platforms, reduce data debt, adopt cloud-first patterns and standardize APIs. Make data findable, accessible and reusable across claims, underwriting and service.
• Build trust at the core: Stand up enterprise governance with clear policies for transparency, bias, privacy, model lifecycle, and human-in-the-loop controls.
• Make AI a business imperative: Define a North Star owned by P&L leaders. Tie every use case to specific KPIs (loss ratio, FNOL cycle time, NPS, expense ratio) and funding gates.

Practical first moves (next 90 days)

• Stand up a cross-functional AI council (business, risk, compliance, IT, data) with decision rights.
• Map your top 10 processes by cost and cycle time; shortlist 3-5 AI use cases with line-of-sight to P&L.
• Publish an AI policy covering data usage, human oversight, model monitoring and incident response.
• Stand up a governed feature store and standard prompts/templates for core workflows.
• Pilot end-to-end, not point solutions: e.g., FNOL to payment in one stitched flow, not separate steps.
• Train frontline teams and adjust incentives so adoption is rewarded, not resisted.

Metrics that matter

• Claims: FNOL-to-payment cycle time, leakage, subrogation yield, LAE per claim.
• Underwriting: quote turnaround, hit ratio, risk selection lift, premium per FTE.
• Service: NPS/CSAT, first-contact resolution, average handle time, containment rate.
• Risk/Governance: model drift alerts, bias tests passed, privacy incidents, audit findings.
• Financials: expense ratio, combined ratio, revenue from new AI-enabled products.

Guardrails to keep you safe and compliant

• Use approved data sources with lineage; restrict sensitive data and apply minimization.
• Record every model and prompt version; monitor drift and retract models that degrade.
• Run fairness tests by segment; document mitigation steps for approvals.
• Keep humans in the loop for high-impact decisions; log overrides and rationale.

Where to go from here

The winners will be the carriers that modernize foundations, embed responsible AI across the value chain and rethink customer engagement for loyalty, not just speed. Those who stay stuck in pilots will fall behind.

For independent research and benchmarks, see HFS Research for industry perspectives here. For governance best practices, review the NIST AI Risk Management Framework here.

Build skills across your teams

If you need structured upskilling paths by role (claims, underwriting, actuarial, operations), explore curated programs and certifications by job function here. Equip product owners, adjusters and underwriters to work with AI, not around it.