Fraud is a booming business: The FBI's 2025 Internet Crime Report found that losses surpassed $20 billion last year, a 26% increase from 2024.
Identity is at the heart of today's fraudulent operations and its being weaponized by criminals to make false claims and leverage fraudulent policies.
So-called "synthetic" identities are a growing threat to insurance firms and are present in nearly a quarter of all claims investigated for identity theft over the past five years, according to the National Insurance Crime Bureau. Ghost policies are created by bogus middlemen that, when exposed and canceled, leave insurers and customers out of pocket.
Identity-driven fraud has existed for decades but these new activities arrive as the industry reaches an inflection point. Having invested heavily in fraud analytics, firms are now turning to AI to help with the fight. It's a priority for 35% of organizations, according to Deloitte.
As organizations ramp up their AI use, reliable, high quality data will become one of their most important assets. The problem many firms are now realizing is that they lack the confidence in their data fitness to ensure that it can be used by AI to help fight fraud.
The customer journey is king
Fraud has traditionally involved exaggerated claims, staged incidents or acts of opportunism with investigators tying individuals to suspicious activity.
This defense, however, is compromised in a world of synthetic identities and ghost policies because criminals can hide behind plausible details while taking advantage of gaps between systems in the customer journey. Fraudsters live in onboarding, policy administration, claims and billing systems by appearing differently in each. They exist, too, in the third-party systems insurance firms rely on to validate customer identity and history — systems operated by credit bureaus, identity registries and law enforcement.
Existing defenses are undermined in two ways. First, by the way data is housed and managed. The systems in the customer journey exist as individual domains demarked by different architectures, data formats and types, and formatting and operational practices. These factors make it challenging for fraud teams and AI to find and use data of the quality needed to spot criminals and patterns of behavior.
When analytics or machine learning models used to train AI are applied to such poor-quality data, the results are compromised and cannot be trusted. Fraud is missed and false positives increase, creating noise that distracts from threats and let bad actors slip through.
The second problem is each system captures just a single moment in the customer journey — such as onboarding or claims — not the full picture. Customer information can vary across these systems with duplicate, incomplete or out-of-date records. Without a consistent, end-to-end view, investigators are left with partial and sometimes conflicting data about an individual or event.
This makes it difficult to spot patterns that emerge over time or during interactions. Without the ability to connect activity across the journey, insurers are left reacting to fraud at the point of claim rather than preventing it through pre-emptive analysis of individuals or group behavior.
Quality data play
Insurers that build a complete, connected customer view will be best positioned to detect and pre-empt fraud using AI. To understand individuals at every stage of the customer journey means rethinking how data is connected, processed and maintained — cracking the information silos and driving up the quality of data released.
The starting point in that process is connecting data across the journey in a consistent and repeatable way. That means creating a shared understanding of key data — what they represent, how they relate to each other and how they should be used. Fraud teams and AI systems can then draw on a complete pool of information rather than juggle competing versions of the truth.
Over time, this creates a reliable and trusted data environment — one where identity can be understood across systems and where the signals of fraud become easier to detect. Insurers can cross-check identity to confirm individuals are who they claim to be.
The next step is making that data usable at scale. It's not enough to know what data exists — teams need to be able to bring it together, shape it and move it quickly for analysis as suspicious activity unfolds. That means two things: First, creating data pipelines capable of combining inputs from those disparate systems — onboarding, policy, telematics and claims. And, second, giving those running fraud detection the power to build pipelines themselves rather than calling in IT, which slows delivery and delays response. When fraud teams can build and adapt pipelines themselves, teams don't become sidetracked by build mechanics and can act quickly on suspicious activity.
Giving teams the tools to build pipelines also breeds resilience. As events adapt and data evolves, existing pipelines can be updated to help teams avoid brittle pipelines that break under new conditions. With the right data flowing at the right time, teams can apply analytics and machine learning to patterns as they develop and uncover relationships between actors that would otherwise go unnoticed.
None of this works without establishing trust in the data itself. A well-connected data platform will fall short if the data flowing through it is incomplete, inconsistent or inaccurate. Profiling and checking data as it moves across the customer journey is essential to identifying gaps, removing duplicates and flagging the anomalies that will scupper AI.
Data management that performs at this level is vital. Small inconsistencies such as amended personal details or information that's been re-used can be easily missed in isolation but caught, quarantined or eliminated through constant and consistent assessment. By establishing data standard processes, insurers can pick out the bad data that pollutes AI. It's a working foundation, where fraud signals are less likely to be obscured.
Conclusion
Insurance fraud is evolving and traditional defenses are creaking as criminals weaponize identity. Firms that assess customers across their entire journey will be best positioned to stay ahead of fraudsters. This means rethinking the data foundations behind fraud detection, raising the bar on quality and delivering data fitness at the speed and scale demanded to pin down an elusive enemy.
Michael Donahue is the global field CTO for Pentaho.
Photo credit: ArtemisDiana/Adobe Stock
© Arc, All Rights Reserved. Request academic re-use from www.copyright.com. All other uses, submit a request to TMSalesOperations@arc-network.com. For more information visit Asset & Logo Licensing.