Insurance leadership teams are accelerating the adoption of AI to improve decision quality and efficiency across underwriting, claims, compliance and third‑party risk. Yet a hard truth is emerging: AI is only as effective as the data, governance and accountability that underpin it.

At Dun & Bradstreet, we work with insurers to operationalise AI. What we see consistently is that insurers are not constrained by ambition, but by fragmented data foundations that limit their ability to scale AI safely and defensibly.

In the Lloyd’s market, where risks are syndicated, delegated authority is common, and portfolios are built through brokers and MGAs, fragmentation shows up quickly. The leadership challenge is no longer whether to adopt AI, but how to do so without compromising trust, regulatory rigour or decision quality.

AI maturity in insurance: ambition meets friction

The causes are consistent: irregular entity data with no persistent identifier, poor quality data trapped in siloes, limited interoperability between internal systems and external sources, and governance that does not scale across use cases. These are not AI model problems; they are data and decisioning problems.

Insurers that make sustained progress focus on operational readiness, including data pipelines, identifiers, entity resolution and clear ownership. Many embed Dun & Bradstreet’s mastered business data and the D‑U‑N‑S® Number at the core of AI workflows, enabling a consistent view of insureds, prospects, suppliers and third parties across markets, classes and lines of business.

Weak data foundations amplify AI risk

In delegated portfolios, incomplete bordereaux or inconsistent legal names only serve to multiply duplicates and obscure group structures. More concerning still, some insurers rely on unverified sources, including search engines and AI tools, to assess third‑party risk. This increases exposure to inaccuracy, bias and regulatory scrutiny.

The risk isn’t limited to individual underwriting or claims decisions. At the portfolio level, fragmented entity data can create silent concentrations, like hidden aggregation across related entities, schemes, brokers or dependencies, that appear diversified at policy level but become correlated under stress. Without reliable group linkage and entity resolution, AI can optimise point decisions while leaving accumulation, counterparty and even supply‑chain dependencies unseen.

Build versus buy: a critical AI design choice

Building and maintaining entity resolution, group linkage, lineage and multi‑market coverage is complex, costly and ongoing. Increasingly, Lloyd’s carriers, brokers and MGAs recognise that AI investment is better focused on pre-mastered data to support underwriting insights, claims outcomes and portfolio steering, rather than recreating foundational data infrastructure.

Governance: from constraint to competitive advantage

AI governance is no longer optional. Transparency and responsible usage are being interrogated. Insurers must be able to reconstruct and explain AI outcomes to regulators, auditors and clients using verifiable lineage, documented transformations and controlled model logic.

In practice, this means tracing a declined risk, adjusted terms or flagged claim back to the sources used, how inputs were standardised and enriched, which model version was applied, and what controls were in force at the time.

Assessing AI readiness

Senior leaders should ask direct questions to determine whether AI initiatives are truly ready to scale:

• Can entities be consistently resolved across systems, jurisdictions and products using a trusted identifier?
• Can AI‑driven decisions be traced end‑to‑end, from source data through enrichment and scoring to outcome?
• Can model performance be monitored over time, with confidence in detecting drift and responding appropriately?
• Can the organisation scale from one successful pilot to multiple use cases without rebuilding foundations?

What good looks like

In summary, senior insurance leaders should prioritise:

1. clean, consolidated, AI‑ready data powered by trusted identifiers
2. interoperability across core systems and trusted external data
3. resilience by design with embedded controls
4. measurable outcomes tied to loss ratio, quote‑to‑bind, cycle time and leakage reduction, and customer experience.
5. transparency so AI‑driven decisions remain lawful, compliant and defensible.

Insurers that build AI capabilities that are traceable and accountable will be best positioned to scale securely, with confidence.

Zulf Raja brings nearly a decade of experience working with insurance firms of all sizes. He has worked closely with industry leaders to deliver transformative solutions that streamline processes through automation and digitalisation, enhance decision-making, drive innovation, and generate measurable outcomes.  Zulf’s expertise lies in applying cutting-edge technology and data-driven strategies to optimise operations, strengthen risk management, and help the insurance sector meet fast-evolving market demands. He has successfully led projects that integrate advanced data analytics and automation, resulting in increased operational efficiency and cost savings.