This insights piece is by Joshua Burkhow, Chief Evangelist at Alteryx
Data readiness has become one of the biggest determinants of AI success in the insurance sector. Yet most firms are still far from the maturity required to support scalable, trustworthy AI. Our recent research found that 51% of insurance organisations rate their data maturity as moderate or low, while 47% cite poor data quality as a major barrier to progress. Despite the hype surrounding AI, the sector’s data foundations remain weak.
That gap matters. Fragmented datasets, ageing legacy systems, years of underinvestment in data resilience, and inconsistent governance continue to erode confidence in analytics outputs and make it difficult to deploy AI safely or at scale. Without trustworthy data, insurers cannot deliver trustworthy AI.
By bringing together their disparate data sources, modernising outdated systems, and strengthening governance and resilience, insurers can build the foundation needed to scale AI with confidence. With that groundwork in place, they can move beyond isolated pilots and unlock the full value of AI across workflows, such as automating claims triage and fraud detection or personalising underwriting decisions in real time.
There’s nowhere left for data quality issues to hide
A critical stumbling block for successful AI rollouts is data quality. The truth is that no insurer today runs on perfect datasets and pipelines. Data maintenance and quality assurance has always been resource intensive, particularly in environments where semi structured and unstructured data sits across multiple legacy systems and sources.
But scaling AI effectively requires a baseline of data quality and many insurers have yet to reach it. Some insurers have layered genAI tools directly onto data that is incomplete, duplicated or siloed, and received outputs such as hallucination-filled outputs in return. This compromises model accuracy and limits the ability to trace or justify how a decision was reached, eroding the explainability that robust model governance requires.
This drives a second crucial blocker: reduced trust in AI outputs. When underwriting teams, claims handlers or compliance leaders lose confidence in the reliability of AI outputs, adoption understandably slows. Senior leaders hesitate to approve AI‑driven changes to high‑stakes processes, and frontline staff become reluctant to use AI insights in their everyday decision‑making.
Efforts to centralise data and their limits
In response, many insurers are investing heavily in cloud data platforms to centralise enterprise data and improve standardisation. These initiatives are an important step towards improving data quality and creating more consistent foundations for AI.
However, centralisation alone rarely resolves the full challenge… Too often, using data in these platforms requires data extraction and knowledge of SQL – skills most underwriters, claims handlers and operations teams neither have and don’t need. As a result, AI use cases can remain concentrated in central teams, removed from the business context where decisions are made.
This reflects a traditional IT‑led approach to rolling out data and analytics applications. But that approach doesn’t suit AI rollout. The teams closest to policy workflows, customer journeys, submissions, claims files, and risk assessments are best positioned to identify high value use cases and define what “good” looks like operationally.
For insurers aiming to become next‑generation intelligent enterprises, data readiness requires more than centralisation. Teams need governed, accessible and context‑rich data close to where decisions are made so they can embed business logic, define guardrails, and shape AI workflows. This is critical in insurance, where decisions must be transparent and auditable.
The push to centralise data also doesn’t necessarily solve the issues of trust that hold AI deployments back. Even when strong data foundations limit the risk of inaccuracies and unpredictability in AI output, challenges can still exist without clear, enforceable controls over what data is used, how it is transformed, and how it flows into and out of AI workflows.
A data stack primed to scale use cases
Data lakehouses provide a unified architecture to store, manage and analyse enterprise data at scale (like a digital library). They can enable progress toward industrialising AI, centralisation alone doesn’t sector remove barriers that have historically limited adoption.
An independent analytics layer can reduce friction by enabling teams to work with governed data stored in these platforms without extraction. This makes it easier for employees to design the most effective, context‑rich AI use cases built on high‑quality datasets.
Crucially, these layers also serve as a foundation for capturing and operationalising the business logic at the heart of insurance decision‑making – underwriting rules, pricing formulas, claims adjudication steps, risk thresholds and regulatory checks. Embedding this logic ensures that AI workflows deliver reliable and repeatable outcomes. Governance rules set centrally at the admin level can then be applied consistently across internal workflows.
Analytics software can offer a low/no‑code interface for users to build AI workflows without any coding required. And, visual in nature, these workflows are simple for line‑of‑business leaders to review, understand data dependencies, validate business logic, and approve – an essential capability for high‑stakes processes where accuracy and auditability matter.
Together, data lakehouses and an analytics layer provide an effective stack for enterprises looking to close the AI pilot to production gap.
This requires sustained leadership commitment. It includes rolling out new tools and building data literacy through targeted training, regular cross-team sharing of delivered use cases, and transparent governance policies for AI.
Final thoughts
As insurers accelerate their pursuit of AI, only those willing to confront their data challenges head‑on will progress beyond stalled pilots, failed projects and scattered wins. Strengthening data quality, governance, and accessibility is essential to restoring trust in AI‑driven decisions. And when modern data architectures are paired with intuitive analytics tools, teams gain the confidence and capability to develop compliant, context‑rich AI workflows themselves.
The insurers that treat data readiness as a strategic priority will be best positioned to move beyond pilot programmes and realise the full value of AI, safely and at scale.
Be the first to comment