The Real World Isn’t Structured: Fixing Entity Resolution for FinCrime Investigations
Explore key insights from Lucinity’s expert webinar on the real-world challenges of entity resolution and data trust in financial crime compliance.
This Lucinity webinar covered the challenges of working with disorganized data and unstructured sources, focusing on how to link companies and individuals across fragmented systems using entity resolution.
Hosted by Luke Fairweather, VP of Sales at Lucinity, and joined by Alex Ridden, CEO and Founder of Knights Analytics, the discussion explored how financial institutions can better connect scattered data across systems to surface relevant risk and eliminate investigative blind spots.
Together, Luke and Alex shared hard-earned lessons on cleaning, connecting, and contextualizing data, especially across incomplete identifiers. The conversation covered structured and unstructured data, entity resolution frameworks, ownership networks, trust validation, and automation’s current and future role in compliance.
"We're talking about something that is pretty integral to the modern world of financial crime investigation… that being entity resolution and playing with structured and unstructured data." - Luke
Key Insights From the Webinar
Introduction: Why Does Entity Resolution Matter in FinCrime?
Luke opened the session by emphasizing the importance of entity resolution in financial crime investigations, particularly in areas like anti-money laundering and adverse media monitoring. He invited Alex to unpack how entity resolution affects detection quality, automation accuracy, and overall investigative efficiency.
Alex shared that her early work in fraud and AML quickly revealed that model performance hinged less on model complexity and more on data readiness.
What Is Entity Resolution, Really?
Alex defined entity resolution as identifying when two or more records refer to the same real-world person, company, or object. This process becomes especially important when matching across different systems, with incomplete identifiers.
Luke offered a simple KYC-period example: using two identifying elements (like name + address) from separate sources to validate a match. However, as Alex explained, financial crime investigations now demand more nuance, particularly when trying to resolve links across common names, complex ownership structures, or fragmented records.
“I didn't think the models we were building were necessarily better than our competitors' models. I don't think we were doing anything super unique. But what we were doing much better than them was preparing the data.”- Alex
Why Does Data Context Outperform Data Quantity?
Alex made a case that data context often trumps data volume. In one project, she found her team’s success wasn’t due to better algorithms, but their ability to combine multiple weak signals into meaningful patterns. Transaction flows, linked entities, and behavioral context helped them infer shared control across accounts, a relationship that direct identifiers alone rarely make clear.
“All financial crime investigations are making a judgment based on data expertise and precedent. So where the data can be woolly, it's all of that context around it that informs the decision that's ultimately going to be made.”- Luke
How Does Unstructured Data Affect Investigations?
Alex detailed two types of unstructured data in investigations:
- Submitted documents – where automation can extract structured data for validation (e.g., uploaded passports or addresses).
- External mentions, such as news articles or blog posts, where the subject might match a client but lacks clear identifiers.
She highlighted that the second group of uncontrolled sources introduces ambiguity. You don’t get dates of birth, registration numbers, or even consistent naming. This calls for systems that can handle partial trust and high variance, adjusting for uncertainty rather than assuming accuracy.
“If you think about the way that, for instance, search engine optimization and the way that Google ranks credibility of sources, it's a similar kind of approach.” - Luke
Can We Trust Adverse Media Mentions?
Luke brought up the “John Smith” problem in negative news: Is it the right John Smith, and does the news even matter to AML risk?
Alex clarified that while Knights Analytics doesn’t classify sentiment, their resolution engine ensures all mentions of a subject are linked, allowing institutions to apply their own filters. She emphasized the need to normalize mentions of different crime types, like how “fraud” or “embezzlement” may be phrased differently across articles, and resolve semantic variations at scale.
“You can either search that information, but then you have to be able to search it at scale all the time, or you can kind of semi-structure it... and then maybe do a second deeper dive.” - Alex
How Should Institutions Approach Dirty, Fragmented Data?
From duplicate records to post-M&A database merges, Alex outlined common data decay issues:
- Corrupted fields from default values or lazy data entry.
- Multiple onboarding events across products without cross-checks.
- No reconciliation layer, leaving identical clients with scattered, non-integrated profiles.
She emphasized that such fragmentation limits visibility and impairs investigations. Entity resolution allows organizations to consolidate information, avoid redundancy, and actually understand customer relationships holistically.
"It’s difficult to do that at scale if you haven’t had an Alex to bring in-house, tidying the data up." - Luke
What About Control and Hidden Ownership?
Luke and Alex turned to the challenge of UBOs (ultimate beneficial owners). Alex described fraud cases where primary perpetrators dropped off the radar post-investigation, while their close associates (often former co-owners) continued operating new shell companies. Detecting control relationships, not just legal ownership, becomes key.
This requires resolution systems that go beyond entity-to-entity linking, building hierarchical graphs of ownership and control, and surfacing indirect links based on co-founders, shared addresses, or transaction clusters.
“You don't only want to merge information about that single entity. You're actually looking at connecting the dots across quite a large and complex structure.” - Alex
How Do Graphs and Networks Support Investigations?
Alex explained how network analytics supports both pattern detection and social linkage understanding:
- Transactional graphs reveal how money moves, its timing, and its volume.
- Social graphs uncover hidden links between individuals or shell firms.
She gave an example of a laundering network where a few individuals were publicly charged, while their former co-owners quickly established new companies, pointing to ongoing activity. Graphs help investigators track those connections and anticipate risk, even before traditional alerts trigger.
Can AI Replace the Investigator’s Intuition?
Luke asked whether AI could replicate human intuition in investigations. Alex responded clearly that it can't yet, and perhaps never will.
While AI excels at parsing large volumes of data and identifying anomalies, it lacks the context and intent recognition humans use to infer risk. She framed human intuition as pattern-based judgment, developed from experience and understanding of criminal behaviors, not mere gut feeling.
“At the moment, I would say, I see AI today being a huge facilitator to investigators. I think it can reduce the time spent on doing the mundane side of retrieving information.” - Alex
What Happens When the Investigation Hits a Dead End?
Luke posed a difficult question, asking, “What if the trail runs cold?”. Alex advocated for perpetual monitoring, where AI systems continue scanning for new signals and correlations, even after a case seems resolved. She noted that new data, updated linkages, or even future media coverage can reignite stalled investigations.
Continuous screening, driven by graph enrichment and open-source data, allows teams to surface dormant risks without reopening every file manually.
“But now, obviously, we've got much better at being able to continuously listen for signals, and I think AI has played a huge role in the ability for us to do that as well.”- Alex
Wrapping Up: Final Thoughts on Data Trust and Resolution
As the conversation drew to a close, Luke thanked Alex for her insights and highlighted how the discussion reinforced the need to make complex, multi-source data both accessible and explainable for investigators. Luke emphasized that platforms like Lucinity, built to integrate and enhance diverse datasets, are critical in today's financial crime environment.
Alex reminded the audience that the ability to detect financial crime depends on how well the data can be connected and trusted. She stressed that investing in strong entity resolution capabilities is foundational to meaningful detection and smarter compliance. With mutual appreciation for the deep dive into entity resolution’s impact, both speakers signed off on a note of collaboration and future innovation.
“It is precisely why we work together… bringing together context that we have in networks, be they transactional or social... and providing a robust framework by which these teams can access, interrogate, and ultimately make decisions on this data.”
Lucinity: Making Complex Data Explainable, Actionable, and Trustworthy
Lucinity’s core platform is built on principles discussed throughout the webinar:
- Unified Case Manager: Consolidates structured, unstructured, internal, and third-party data in one actionable interface.
- Luci AI Agent: Transforms complex data into explainable summaries, visualizes money flows, and connects relevant entities across systems.
- GenAI Plug-In: Integrates Luci’s skills across CRMs, Excel sheets, and browser-based apps, boosting productivity by 90%+ with no infrastructure change.
- Skills Studio: Let's institutions customize entity resolution logic, automate news screening, and analyze customer networks in real time.
To see how these tools help organize monitoring, investigations, and reporting while supporting stronger compliance, visit Lucinity.
Meet the Speakers
Alex Ridden
CEO & Founder, Knights Analytics
Alex works with entity resolution and graph analytics using AI-based methods. With over a decade spent building production-grade ML systems for TMNL, ABN AMRO, and others, her work helps financial institutions connect disparate data and surface real-world risks.
Luke Fairweather
VP Sales, Lucinity
Luke leads Lucinity’s efforts to lower operational costs in financial crime work and has a strong background in enterprise FinCrime and RegTech sales. With experience at PassFort and Moody’s Analytics, his expertise lies in connecting customer needs with actionable AI capabilities.
Final Thoughts
The conversation between Lucinity and Knights Analytics highlighted that entity resolution is fundamental to effective AML and fraud detection, rather than just an added feature. Accurate matching across fragmented data depends more on data quality, context, connections, and clarity than on complicated models.
The discussion addressed the challenge of working with messy, unstructured sources, given that institutions use open data and adverse media. It highlighted the growing importance of identifying control and social connections, particularly in UBO detection.
Another important takeaway is that while AI helps uncover hidden risks and speeds reviews, human judgment remains irreplaceable. As Alex Ridden said, “It might feel like intuition, but it’s rooted in real and deep knowledge of crime.” Watch the full webinar for more details- https://youtu.be/lkWxw0MMQ9s?feature=shared