AI innovations unlock digital industrialisation opportunities

Artificial intelligence-led process digitalisation offers outsized benefits to post-trade.


For more than five years we have been experimenting with artificial intelligence, robotics and more, identifying what works and what does not. From this we know that few initiatives offer such outsized benefits to the securities servicing industry as artificial intelligence-led process digitalisation. With industry wide consensus on the potential and advantages of these technologies clear, we are in the deployment stage.

The goals of our investment and long-term plans are threefold; continue to improve our client experience, increase efficiency and reduce risk. We recognise that talent is key to successful use of technology, training is vital. As part of our investment, we have a global data and digital training programme run out of our International Operations Centre in Warsaw, to guarantee the readiness of our entire staff as to the capabilities in place and being rolled out.

Within BNP Paribas Securities Services, we have rolled out multiple technologies to achieve our aims and continue to develop powerful new capabilities using AI. We are transforming how we work and the experience we deliver to clients.

Live AI transformation

Automating processes as much as possible is vital. We are incorporating a range of artificial intelligence-based technologies into our post-trade processes, combining machine learning, natural language processing (NLP) and natural language understanding (NLU) technologies.

Document management – a task that affects every domain and activity across the bank –gains particular benefits from digitalisation. Using rules-based technology for structured documents and machine learning for more complex unstructured ones, we automatically extract information from a wide range of documents, including SWIFT messages, email, paper forms and PDFs. To date, our collection of AI-driven tools manage approximately 300,000 documents annually. The goal is to expand that to over 1.5 million documents in the next two years. Coupled with our robotics efforts, managing 2.8 million tasks per year, this brings enhanced accuracy and reliability, accelerated processes, reduced risk, increased productivity and improved responsiveness to client requirements.

We apply these AI-fuelled information extraction capabilities across multiple areas:

Fund prospectuses

By leveraging NLP-based capabilities, our tool can read the language in the documents and identify data we want to retrieve. We use the extracted information to improve how we set up our systems and manage the risks related to the fund.

Contract management

When, for example, a new product is to be launched, checking which clients have signed an affected contract is vital. We can now easily extract the data to determine which clients are impacted, and smoothen the change.

Fee schedules

Our machine learning technology extracts fee conditions, possibly unstructured with multiple scenarios, and automatically uploads them to our systems. It conducts ongoing reconciliations to help us ensure clients receive high quality and accurate billing.

Trade management

With the help of optical character recognition (OCR) technology, we can convert documents – faxed trade instructions or confirmations, corporate action announcements – from image to text format. From there, we extract the necessary information and convert it to a proper instruction, creating an efficient end-to-end process.

Report generation

Powerful natural language generation capabilities enable us to produce automatic executive summaries of huge reports with lots of data. Clients receive fast and easy-to-understand report summaries using the data on their global activity, helping them quickly understand the main performance points.

Corporate action translation

A translation tool, built and trained in-house, using machine learning technology, allows us to swiftly create corporate action announcements in multiple languages. Corporate event announcements often include heavy narrative sections. Since we provide a global custody service, narratives must be translated for local clients. The tool ensures data sent to clients is high quality and on-point, translating on average 30,000 messages per month in 2021.

Combined technologies to drive synergies

We are integrating AI tools with other technology innovations to streamline our end-to-end processes further. Through these combinations, we are enhancing operational servicing and automating the routing of information.

Automated tax reconciliation is a perfect example of a successfully implemented solution. Using OCR and data extraction technology to extract information from PDF reports, the data format is then transformed using an ETL (Extract Transform Load) solution. The data is then reconciled through RPA (Robotic Process Automation), and a user friendly view is provided thanks to a visualisation tool. This combination of multiple technologies allows us to create an end-to-end solution bringing security and efficiency for our clients.

Regulatory intelligence is a key activity for many services within the bank, analysing texts that impact services, products and customers. We have implemented a third-party digital solution to streamline that monitoring process. The solution automates the search for, and selection of, relevant regulatory texts. It optimises collaborative work and helps disseminate the information via newsletters to clients both quicker and safer. We also leverage the tool for market intelligence to generate automated reports on market changes.

E-signatures, another prime example, are vital for digitising end-to-end document management processes. E-signature demand has rocketed since the pandemic; we have e-signed more than 3,000 documents since January 2021. The technology is mature and user comfort is growing. The main barrier to further take-up is the residual legal hurdles, which need to be analysed country by country and contract by contract to ensure e-signatures are permitted.

Our email classifier delivers additional efficiencies for our clients. Allied with Hobart, our client query management system, the tool examines the content of emails, classifying them automatically and speeding up our response times. The system analysed 220,000 emails in 2021, allocating them to the person best positioned to provide our clients with a timely answer.

Advanced analytics initiatives focus on reducing risk across diverse business areas. The technology’s core purpose is to anticipate anything that could go wrong. By combining a rule-based, statistical approach with machine learning, it generates algorithms to detect potential areas of failure, thereby lowering risk for our clients and us.

Application stability is a key area of focus, identifying hidden patterns with this technology combination. We have launched two major studies, one classical data science R&D project and a PhD mission with the Pantheon-Sorbonne University Lab. Both will scientifically assess client platform stability, to identify innovative solutions to predict IT incidents and allow pro-active action.

Analytic technology applies well to cash payments, using a statistical approach to look for suspicious activity, or payments that do not follow a client’s standard patterns. Flagging unusual payments helps guard against illegitimate transactions, whilst protecting our clients’ interests.

Workflow management – the next step?

A newer area of focus for our AI-related technology initiatives is process modelling, diagnosis and workflow management.

Here we are employing the modelling capabilities offered by process mining, a method of applying specialized algorithms to event log data to identify trends, patterns and details of how a process unfolds, allied with a workflow management orchestration tool, to accurately map our system processes. The technology flags bottlenecks and potential areas of inefficiency, and helps redesign and monitor our processes securely and efficiently. We will be rolling out this capability this year.

None of the technology initiatives which we are undertaking – whether it is machine learning-based document management or robotic process automation to ease routine tasks – would be effective without data quality. Digital industrialisation demands high quality data, which is why we assess the consistency of the data flows and introduce controls to enhance its quality every time we implement a solution.

Embrace the AI future

The last five years have confirmed that the link between technology and talent is vital; the technology is only as good as the talent working on it. At BNP Paribas Securities Services we’re instilling digital as part of our DNA. Technology is one of three key pillars of BNP Paribas’ 2025 strategic plan, putting industrialisation and client experience at the heart of our model.

As we continue to deploy these technologies, sustained experimentation is part of life. New use cases for predictive analytics and process mining will be a focus as we move forward. Speech recognition technology, although not yet mature in post-trade, is on our horizon for potential developments.  

Client experience, increased efficiency and reduced risk are shaping the integration of industrialised digitalisation into our end-to-end processes. The advances we have made have had a major impact across the business. By embracing emerging technologies, we will deliver further efficiencies and process improvements to our clients in the months and years to come.