19 Feb, 2021 • 2 min readArtificial Intelligence Natural Language Processing Machine Learning Communications Mining
Reimagining Post-Trade Operations with Machine Learning and Intelligent Communications
“I’m totally convinced that the battleground of banking is not the front office. The battleground is the back end.” Sergio Ermotti, CEO UBS
Banks’ Legacy Operations Transformation Challenges
A decade on from the global financial crisis, many banks are still struggling with profitability and high cost-income ratios. Alongside efforts to increase revenues, banks have rightfully identified back-office operational efficiency as the key to unlocking the next level of cost savings.
To enact change, most banks have continued to pull traditional cost levers such as outsourcing, process excellence, and workflow management. Other much-hyped technological solutions, like Ops 4.0 and DLT, promise productivity improvements through the comprehensive transformation of legacy operations. In practice, successful implementations are rare and their results superficial.
The same is true for robotic process automation (RPA). Real-world outcomes have been underwhelming with multi-year change programs failing to deliver expected benefits. Operating costs remain stubbornly high, and in some cases, cost-income ratios are increasing.
Where these other solutions have failed to deliver, advances in natural language processing (NLP), machine learning, and communications mining are already accelerating opportunity discoveries with remarkable results.
90% of enterprise data is unstructured, and a significant portion of this data is text in the form of emails, customer support chats and notes.” International Data Corporation
A Proven Solution – Communications Mining
By combining unsupervised and supervised machine learning techniques, communications data can be mined for key concepts and intent without complex heuristics.
To begin the multi-stage learning process using a dedicated platform, businesses simply need to create a data taxonomy of business-specific labels with which to tag their conversations. For post-trade operations, these will typically describe a hierarchical processes model which identifies function and exception type, e.g., “Settlement > Fails > Counterparty Lack”.
The insights gained from the subsequent automated process facilitate hyper-streamlined operations, allow for the automation of time-consuming manual tasks, and enable new levels of productivity and efficiency savings.
How Advances in Communications Mining Benefit Post-trade Operations
The most effective of these communications mining platforms now only require 15-20 data samples to detect new concepts with high confidence. This significant step forward in the fields of NLP and machine-learning already provides two major business benefits:
1) The Detection of High-Risk, Low-Frequency Events
Inferior data-mining solutions which require thousands of data samples simply cannot detect some of the events that risk the most harm. With just a handful of samples, advanced machine learning software can now automatically detect rare events such as GDPR breaches which risk catastrophe.
2) Short Training Cycles
Whereas less efficient NLP can take weeks or months to effectively train a model, the best communications mining solutions can complete the process in a matter of hours regardless of the size of the data set. After this hyper-brief training period, the taxonomy can be applied in real-time.
These advances are only the beginning of what communications mining can achieve. By implementing automated preventative measures rather than manual reactive ones, processing and exception management can eventually become fully automated.
Case Study – Communications Mining Benefits in Practice
Re:infer are at the forefront of research in the intelligent communications sector and are already working with a select group of early-adopter banks with staggering results.
Re:infer’s Co-Founder and CEO Dr. Ed Challis and CCO Stephen Mackintosh supported UBS Enterprise Architect Martin Swanson’s authorship of the whitepaper – “How Machine Learning is Enabling New Cost Levers in Post-Trade Operations”.
As well as more in-depth studies into banks’ legacy operations transformation challenges, the whitepaper provides ground-breaking insights on the machine-learning developments that are accelerating large-scale change and actualizing the “Zero-Ops” vision.
The whitepaper also covers:
- How communications mining highlights root-cause errors, quantifies failures, and provides a bridge to downstream automation
- Why Google’s deep-learning solutions do not deliver on unbounded communications
- A detailed case study of a successful implementation in a large European bank
- Ten key findings after mining the bank’s 300 shared mailboxes and five terabytes of communications data
- The identification of problems in fixed income settlements and their related treasury functions
- The case for embracing “Zero-Ops” – How operations can be transformed into process functions
- Prototype and platform design