The leading knowledge platform for the financial technology industry
The leading knowledge platform for the financial technology industry

A-Team Insight Blogs

Opinion: How Machine Learning Solves Client Onboarding, by Adam Devine, WorkFusion

Brute force can’t solve onboarding. Machine learning can. Here’s how.

Throwing headcount at the ubiquitous client onboarding problem hasn’t worked. It’s still a mess. Offshoring just sweeps the mess under the rug, and the expression, “misery loves company” is a good way of describing the impact of utility plays. Every onboarding solution on the market today just shifts the burden and expense of data analyst headcount to different places and P&Ls, and because every bank has a slightly (or very) different onboarding process, there will never be a single off-the-shelf solution for every bank. The key to reducing the risk and cost of onboarding is automating the manual, repetitive human work within every unique process. Are we suggesting that every financial institution kick off its own enormous custom IT project? Certainly not. That’s a well-worn page from 1998’s playbook. Cloud-based machine learning is the answer in 2015, and it’s radically improving the way data is collected, validated, and enriched.

What are the challenges of traditional automation?

Let’s first talk about machines that don’t learn, otherwise known as rules-based automation. Scrapers, optical character recognition (OCR), and parsers are common Rules-Based Automation (RBA) point solutions for data collection. They work well if the underlying business process and data sources never change, but change is the only constant in an onboarding process. RBA requires upfront programming and configuration by IT, and it’s impossible to program every potential variable in a data collection process or account for variations in client documents. When the process or sources change, IT must re-write the rules. The process halts while the point solution is repaired, and business continuity is compromised. This is why outsourcing begins where RBA ends – it’s just easier to have headcount do the work than it is to constantly repair and maintain automation.

What are the challenges of headcount?

A human-only data analyst workforce is an expensive, fixed and finite resource, unable to elastically scale up to meet bursts of demand or scale down in troughs. Skilled as they may be, people naturally make mistakes, and sometimes these mistakes are incredibly costly. Supplementing or replacing a full-time equivalent (FTE) workforce with outsourced workers provides moderate cost relief and incrementally more scalability, but as both data volumes and global labor rates rise, the benefits of business and knowledge process outsourcing (BPO / KPO) fall.

What is machine learning?

You’ve heard of it, and you may even be able to describe it, but let’s define it here just to be thorough. Machine learning is computer programming that can program itself by watching humans perform a task. With sufficient repetition of a method of work, machine-learning algorithms can identify a consistent pattern and program itself to replicate it. Machine learning and its parent field, artificial intelligence, will significantly change and improve knowledge work. The most relevant improvement for banks will be the incremental automation of data collection, validation, and enrichment, which will reduce reliance on offshore labor, improve risk profiles, increase controls and data accuracy, speed up processes, and radically reduce the cost of run-the-bank.

Why hasn’t machine learning solved onboarding yet?

Machine learning is a little like a Formula 1 racecar. It’s incredibly sexy and thrillingly fast, but it’s also very complex and difficult to maintain. Machine learning first and foremost requires a lot of high quality data to train algorithms. G-SIBS generate plenty of data for algorithm training, but then there’s the problem of selecting and testing the right general purpose learning algorithms and modeling them in an effective way for a given task – like plucking IBANs from standard settlement instructions or monitoring SEC documents on EDGAR for changes. The selection, modeling, and training process can take years, and no bank has that much time (or budget) for an IT project.

How can machine learning solve onboarding today?

Rather than every bank building its own racecar and test track, the speed, accuracy and efficiency benefits of machine learning are now available as a software service. WorkFusion does for machine learning what Amazon Web Services did for cloud computing, turning an expensive, challenging capability into an on-demand service. WorkFusion is an end-to-end platform for optimizing and automating data collection. The software invisibly pairs machine-learning algorithms with FTE or outsourced human data analysts doing business-as-usual work. This pairing, sometimes referred to as human-in-the-loop computing, incrementally automates the repetitive tasks within processes.

17 of the top 20 global data vendors use WorkFusion to perform mission critical data collection work for entities and instruments, corporate actions, and risk and compliance. Global banks have begun to use WorkFusion for a wide variety of data collection processes, from SSI extraction to BCBS 239 compliance to, most importantly, client onboarding and AML/KYC. WorkFusion spent four years and a considerable amount of its investors’ money building a world-class, enterprise-proven software platform that brings the benefits of machine learning to the financial industry.

Onboarding is only one of thousands of applications that banks will discover and deploy in the coming years, but it is certainly the one of the first big messes that machine learning will clean up for the financial industry. For more information about WorkFusion, swing by the company’s booth at A-Team’s Data Management Summit (DMS) in London this Wednesday and New York on April 16th, check out www.workfusion.com, or email info@workfusion.com.

By Adam Devine, VP, Product Marketing & Strategic Partnerships, WorkFusion

Related content

WEBINAR

Recorded Webinar: Best practice approaches for corporate actions automation

Demand for timely and accurate corporate actions data is growing as volumes and complexity rise, and financial institutions acknowledge the increasingly costly gap between accurate corporate actions processing in real, or near-real, time and faulty processing caused by poor data and resulting in missed opportunities to optimise revenue. While many firms currently process parts of...

BLOG

The Great and the Good Gather for A-Team’s RegTech Virtual Event This Week

The global RegTech community gathers virtually this week for A-Team’s RegTech Summit Virtual, which kicks off on Wednesday. Over its two-day programme of must-attend sessions, the conference will offer an assessment of the global RegTech landscape, plus discussions ranging from the ongoing regulatory reporting challenge, the emerging conduct compliance requirement, addressing financial crime and other...

EVENT

Data Management Summit New York City (Redirected)

Now in its 8th year, the Data Management Summit (DMS) in NYC explores the shift to the new world where data is redefining the operating model and firms are seeking to unlock value via data transformation projects for enterprise gain and competitive edge.

GUIDE

Regulatory Data Handbook 2021/2022 – Ninth Edition

Welcome to the ninth edition of A-Team Group’s Regulatory Data Handbook, a publication dedicated to helping you gain a full understanding of regulations related to your organisation from the details of requirements to best practice implementation. This edition of the handbook includes a focus on regulations being rolled out to bring order and standardisation to...