A data-driven approach to duty of care | Regulatory Risk | Deloiitte Netherlands


A data-driven approach to duty of care

Identify potential duty of care risks in a smart way

For financial institutions, upholding their duty of care to their clients is a formidable challenge. When the bar is raised, past dealings with your clients come under scrutiny and archived documents and data become crucial. How searchable are your archives should you be faced with disputes or even remediation suits?

Since legislation on duty of care was introduced, its meaning and implications have remained elusive. Duty of care provisions are based partly on open standards, and these are constantly shifting. With legislation, supervisors, public perceptions and court cases constantly redefining what duty of care means, conduct risk looms large.  What was considered acceptable a market practice five years ago may no longer be acceptable now. Risks in this area are therefore hard to anticipate and require an approach that allows a forward looking view. The risks are evidently considerable, both in financial terms and in terms of your brand and reputation.

Hiding in plain sight

Our experience with numerous remediation efforts shows that client records are often hidden in a mass of unstructured data. PDFs, emails, digital forms, reports from all kinds of legacy software, scans of handwritten notes, and even paper files. In some instances an army of analysts and subject matter experts had to go through the archives manually to ascertain what the financial institution had communicated. Dreary and error prone work that takes years.

Technology: an extra set of hands, eyes and brains

But what if there is a way to speed up this process and get better quality results at lower costs? A technology-based solution? The good news is that this technology exists in the form of Natural Language Processing, and is steadily becoming better, faster and sufficiently cost effective to be used in a business setting. 

Natural Language Processing (NLP) uses machine learning and other types of algorithms to extract meaning from unstructured data. NLP can semantically analyze huge volumes of text in a reasonable timeframe, and can even deal with complexities like figurative language and synonyms. In simple terms, it does this by translating words into mathematical equations and then looking for patterns via neural networks – a set of algorithms modelled loosely on the human brain. 

Smart and smarter

Data extraction engines based on NLP technology can extract specific data entities that you need, like contract clauses, start and end dates, or disclaimers. It can recognize such entities in all kinds of files, even if the format and wording are not identical in all of them.

Moreover, these data extraction engines are swiftly becoming smarter. While early versions needed to be fed thousands of documents before they learned enough to search intelligently, these days good results can be derived from data sets as small as 50 documents. As more documents are fed to the extraction engine they pool what they learn from previously processed documents.  Essentially they compile a common body of knowledge from many analysts and experts to become even smarter. 


Coming to grips with your data means more than just being ready when duty of care issues arise. It means as a company you can actually be proactive, protect your client relationships and nip potential conduct risks in the bud. Before a new regime comes into force or acceptable business practices change, you can dive into your client portfolio and easily check whether contracts you engaged in long ago are compliant, find out the related terms, conditions and characteristics, and take measures. 

Have you recently started offering better terms for certain products? It is now relatively easy to ensure that existing clients with similar products are identified, made aware and offered an upgrade. This will lead to higher levels of trust and customer satisfaction. In contrast to the situation, that they notice the discrepancy themselves and contact you.


Staying on top of duty of care risks is a lot easier if you have all the data you need available in a structured format and at your fingertips. Today, thanks to advances in NLP technology and falling costs, this is no longer just an academic exercise. 

Deloitte is a pioneer in NLP-based solutions and has – among other – developed its own data entity extraction engine, DocQMiner .
It’s just one aspect of our extensive expertise in supporting our clients with their
duty of care and other challenges. Interested in how we can help your organisation gain insight into its duty of care risks? Get in touch with one of us.

More information

Do you want to know more? Please do not hesitate to contact Frank Boerkamp or Wendy Brink-van den Nieuwenboer via the contact details below.

Did you find this useful?