“Illicit actors’ objective is to dilute the capability to identify the sources of their funds. Given we know that they operate as part of a network, we need to target networks rather than individual accounts. That’s how you can actually move the needle to disrupt criminal enterprises and take effective action. This is the objective we must have in mind when we are implementing data analytics and the advanced technology that builds on it.”
– Seb Drave, Director of Customer Solutions and Innovation, Harbr
This article draws key observations and findings from a conversation between Seb Drave, Director of Customer Solutions and Innovation at Harbr, which has a mission to build next-generation data businesses, and the Institute for Financial Integrity’s Associate Managing Director, Catherine Woods.
“Data analytics” is a field that encompasses multiple disciplines focusing on leveraging data to make the most of the insights that can be gained. Previously referred to using terms such as exploratory data analysis, data mining, or applied statistics, more recently the discipline has come to be referred to as data science. This can be attributed to the integration of research and development approaches and experience to the enterprise workforce, and the recognition that as data analytics techniques and technology have advanced, they enable organizations to use mathematically and statistically-driven methods to gain increasingly accurate insights from less complete data. Data analytics intersect with machine learning and artificial intelligence including generative AI.
“You can’t predict what you don’t have experience of, and you can’t explain and understand things you don’t measure.”
– Seb Drave, Director of Customer Solutions and Innovation, Harbr
An organization that implements a data-driven approach to its business ensures it has situational awareness of what is happening within that business and in the wider market. Well-designed models built on appropriate and well-structured data can be used to advance workflows, optimize products and services, improve client interactions, and enhance decision-making. Without them, organizations are more likely to be reactive, less efficient with their resources, and only able to resolve issues after they have occurred – rather than gain early insights that can be actioned sooner.
In the financial crime context, data analytics enhance the ability to detect and take action against illicit activity. They also enable productivity gains and an increase in effectiveness. For example, by minimizing the number of false positive alerts, they reduce the resources required for validation and assessment. Commercial benefits can also be achieved too. For example, data analytics can be used to identify which vendors or suppliers present the most risk, are most cost-effective or provide the best performance outcomes.
“If you’re not capturing data and using that to make your decisions, then your business decisions and processes are being run on perceptions and best guesses. It’s like shooting in the dark: occasionally you might hit something, but a lot of the time you won’t. In an environment where resources are constrained and there’s competition for market share, you want to make sure that you are getting the maximum from the resource that you have and doing the best that you can for your clients.”
– Catherine Woods, Associate Managing Director, Institute for Financial Integrity
An organization seeking to implement a data strategy and analytical capacity can take the following actions to maximize successful outcomes.
The first step is to establish a data strategy. This includes both technical requirements – such as data architecture and security – as well as governance and organizational elements – such as roles, responsibilities, and processes for managing data. The strategy ensures data is managed consistently and appropriately throughout the organization in accordance with its strategic objectives.
In alignment with the strategy, data should be captured, monitored, and evaluated to inform and optimize operational processes, products and client interactions, and decision-making, to name just a few potential applications.
To gain support for data initiatives, it is most effective to identify a specific problem that can be solved use data-centric methodologies. Resources can then be focused to acquire the necessary data, apply relevant analytics, then demonstrate a quantifiable improvement in the original process or problem. Rapidly demonstrating measurable improvement over current operations will encourage support and adoption by the rest of the organization.
In contrast, many data initiatives begin by attempting to acquire and integrate all organizational data sources, an approach which is resource intensive, lengthy, and complex – and where measurable benefits are not evident for extended periods.
“There is undoubtedly information outside your organization that will make your capabilities more powerful. But there’s also a lot of information that may make your capabilities more prone to noise and if used in an unconstrained manner, they may overwhelm your systems, processes, and the people working through this.”
– Seb Drave, Director of Customer Solutions and Innovation, Harbr
While internal data is a key data resource, it can be enhanced by well-considered selection and integration of external data sources. Well-established examples of external sources used within financial crime compliance for due diligence and investigations include watchlists of adverse news, sanctions lists, and corporate registry information.
“The greatest and most complete view of the world will come with organization’s sharing and working together both at the operational level and potentially the data level as well. When anyone in one of those organizations is thinking about what is the most valuable external data, I wouldn’t necessarily limit that to commercial providers of certain types of information or particular public sources. Potentially, your peer group organizations are also going to have very kind of valuable resources from the industry as a whole.”
– Seb Drave, Director of Customer Solutions and Innovation, Harbr
In some contexts – financial crime being a key example – some the most valuable data is held in organizations where there are significant restrictions on sharing information. For example, governments and law enforcement hold information about investigations and networks, while financial institutions hold details on client profiles and suspicious activities. While restrictions on sharing are imposed for valid reasons such as data privacy, the effect is that each organization has a more limited view of the problem set and usually only has visibility of part of an end-to-end transactions flow or illicit network. In the context of data analytics, this means that each institution has fewer data points to train its models, with a particular shortage of “confirmed positives” such as validated suspicious activities.
To improve data capabilities, organizations should identify what data can be shared and how to balance regulatory and internal policy objectives. This may include incremental sharing of data points, starting with less sensitive information or attributes, or it may include using technology solutions to maintain confidentiality and ensure that each organization maintains custody of its own data.
By finding solutions, the performance of participating institutions can be enhanced, and most importantly so can effective action against financial crime.
Deep Dive: Privacy-Enhancing Technologies and Operational Checks and Balances
“Hopefully as an organization your controls are working effectively and during onboarding, you are preventing businesses and individuals from gaining access to your organization and misusing your products and services. If that’s working well, it means confirmed cases of actual illicit activity will be quite small. This has implications when training a model to ensure that the limited dataset does not impact its accuracy and reliability.”
– Catherine Woods, Associate Managing Director, Institute for Financial Integrity
To effectively train models, for example to enable machine learning, a set of validated outcomes is required. In many use cases, these validated outcomes are plentiful. Examples include a hospital using large volumes of patient records to enhance models to diagnose diseases, or an airline using aircraft sensors and service records to improve fleet maintenance schedules.
However, the financial crime context is different and validated outcomes are scarce. If a model is being trained to identify activity indicative of financial crime, the dataset of “confirmed positives” may be based on suspicious activity reports. However, within an institution, these known positives are rare and there may be systematic biases (such as defensive filings of SARs/STRs). The result may be that there may be insufficient data to train an accurate model, or that the model’s learning may be skewed by the few examples that are available.
While data sharing between institutions provides one way to augment these datasets (described above), another solution is “synthetic data”. A carefully constructed synthetic dataset can augment real data to enhance model development. Firstly, it can “boost the signal”, referred to as “up sample”, the amount of financial crime activity within the system, while still having a large population of normal transactional patterns. Secondly, synthetic data can enable a feedback loop of build-train-validate, which can be difficult to do in the real world where confirmed positives are infrequent and may not always be representative.
With a more comprehensive and well-constructed dataset, models can be trained for greater accuracy and effectiveness, enhancing confidence in their use.
Deep Dive: Synthetic Data
Deep Dive: Regulators Encouraging Innovations Such as Synthetic Data
“Data democratization is one of the key elements that allows you to foster a multidisciplinary ethos within a team… It also enables more technical members of a team to build, test, and deploy repeatable capabilities for use by other members of the team.”
– Seb Drave, Director of Customer Solutions and Innovation, Harbr
Data democratization refers to providing the skills and tools to enable data consumers, such as financial crime investigators, to more effectively interact with and access data. In a siloed model, capabilities and access to extract data and meaning from it are held by specialist data scientists, working in separate teams from those using the data to achieve operational outcomes. This can be described as a separation between “deployment and production” and “consumption and outcomes”. If more staff can be provisioned with and confidently utilize more advanced analytical capabilities, there will be a greater return on investment.
“This is a partly about increasing the kind of capability of the investigators and analysts, but also making the data more accessible so that they are more effectively able to access it directly and run queries… As an investigator or intelligence analyst, that would be extremely valuable because when you are following a lead, being able to query a data set directly rather than to refer it to a separate team and then it comes back and then you ask another question, would increase the effectiveness of everyone involved. And of course, the other advantage is that you have a common data picture and potentially the scope to more effectively feedback the outcomes that you’re identifying, rather than them being captured in a spreadsheet on a team share drive somewhere.”
– Catherine Woods, Associate Managing Director, Institute for Financial Integrity
Deep Dive: Data Democratization
“You could ask AI to generate an investigative case summary for someone who is not a financial criminal in any way. And the AI could build an incredibly compelling view of the world that you would read, but ultimately it would be totally false and not rooted in reality. From the point of view of anti-money laundering and counter terrorist financing, utilizing those AI capabilities in a very controlled way, ensuring that only the appropriate kind of the appropriate context is fed to those models, is necessary.”
– Seb Drave, Director of Customer Solutions and Innovation, Harbr
In the context of financial crime investigations, use of technologies such as machine learning and other artificial intelligence, must be carefully considered. Over recent years, we’ve seen the development of technologies such as data analytics and data science, graph analytics, synthetic data, and technologies to enable sharing while protecting privacy and custody. In the next few years, we are likely to see these different capabilities unifying together to enable rapid progress towards solutions which are more powerful and more targeted at the problems they intend to solve. In the financial crime context, these will empower intelligence functions to execute the mission in a way that hasn’t previously been possible, and at scale.
Deep Dive: AI Agents
Humans will continue to fulfil a critical role, particularly in contexts where the consequences of inaccuracy are so significant. They will train models, validate their performance, and tune their outputs. Humans will evaluate the results of the models – such as potentially suspicious findings – and ensure they stand up to scrutiny, before taking action on real customers with real lives.
“Even with machine learning, it is incredibly important to have the human in the loop. It is especially important in the financial crime use case, and that will continue to be the case even with generative AI because of the high stakes, the nature of the problem space, and the kinds of determination you’re making around members of your client base.”
– Seb Drave, Director of Customer Solutions and Innovation, Harbr
https://finintegrity.org/wp-content/uploads/2026/09/shutterstock_2645005987-scaled.jpg
1706
2560
IFI
https://live-black-pebble.pantheonsite.io/wp-content/uploads/2023/12/GIFI-Placeholder2.png
IFI2026-09-24 07:00:302026-09-24 11:44:27Dirty Barrels
https://finintegrity.org/wp-content/uploads/2026/01/bg-synethic-identities.jpg
500
1200
IFI
https://live-black-pebble.pantheonsite.io/wp-content/uploads/2023/12/GIFI-Placeholder2.png
IFI2026-01-08 07:00:252026-03-27 12:40:46Synthetic Identities
https://finintegrity.org/wp-content/uploads/2025/08/shutterstock_2599430089-scaled.jpg
1202
2560
IFI
https://live-black-pebble.pantheonsite.io/wp-content/uploads/2023/12/GIFI-Placeholder2.png
IFI2025-08-27 07:00:452026-05-14 14:14:01Deepfake Deep Dive
https://finintegrity.org/wp-content/uploads/2024/11/bg-ai-vs-human-judgement.jpg
1162
1800
IFI
https://live-black-pebble.pantheonsite.io/wp-content/uploads/2023/12/GIFI-Placeholder2.png
IFI2025-04-09 07:00:202025-09-25 15:17:41AI vs. Human Judgment
https://finintegrity.org/wp-content/uploads/2025/02/the-heat-is-on-bg.jpg
1157
1800
IFI
https://live-black-pebble.pantheonsite.io/wp-content/uploads/2023/12/GIFI-Placeholder2.png
IFI2025-02-24 07:00:152025-03-26 19:38:52The Heat is On
https://finintegrity.org/wp-content/uploads/2025/01/article-green-gold-rush.jpg
1261
1800
IFI
https://live-black-pebble.pantheonsite.io/wp-content/uploads/2023/12/GIFI-Placeholder2.png
IFI2025-01-13 07:00:472025-09-25 15:00:34The Green Gold Rush
https://finintegrity.org/wp-content/uploads/2024/12/bg-toward-financial-integrity-risk-management-program.png
835
1800
IFI
https://live-black-pebble.pantheonsite.io/wp-content/uploads/2023/12/GIFI-Placeholder2.png
IFI2024-12-19 07:00:412025-09-25 14:57:22Toward a Financial Integrity Risk Management Program
https://finintegrity.org/wp-content/uploads/2024/12/shutterstock_2472894163-scaled.jpg
1403
2560
Lauren Jack
https://live-black-pebble.pantheonsite.io/wp-content/uploads/2023/12/GIFI-Placeholder2.png
Lauren Jack2024-12-12 07:00:342025-09-25 14:57:44Safeguarding Trust – How to Balance Innovation and Security in Gen AI-Powered Compliance
https://finintegrity.org/wp-content/uploads/2024/10/beneficial-bg.jpg
1200
1800
Lauren Jack
https://live-black-pebble.pantheonsite.io/wp-content/uploads/2023/12/GIFI-Placeholder2.png
Lauren Jack2024-12-10 07:00:012026-05-14 14:51:01Webinar Recap – The Dos, Don’ts & Expert Insights on Beneficial Ownership
Optimizing Effectiveness