Injecting Artificial Intelligence into Financial Analysis
As with many other industries, there are many processes in banking that we have to go through on a daily basis. These traditional processes form the ‘business as usual’ for many organisations and its functions — but what if usual can be better?
Developments in artificial intelligence (AI) and machine learning (ML) have opened new doors to improving traditional processes and making work more efficient. One such example is the automation of financial analysis, which has saved valuable time and effort.
We catch up with Don Ng, Executive Director from the DBS Risk Management Group Credit COO Office, and Dr. Luan Huanbo, co-founder and CEO of start-up 6Estates, to find out more about their proof-of-concept (POC) for the Automated Financial Analysis (AFA) project through DBS Startup Xchange and how emerging technologies are creating new opportunities.
Hello, Don and Huanbo! Shall we start by sharing more about the AFA project that you guys worked together on?
Don: Sure! For the DBS Risk Management Group, understanding the financial credit standing of our customers is an important aspect of corporate lending, and we typically start by reading the annual report. A large corporate annual report could run anywhere from 100 to 400 pages, with credit information spread across these pages. We wanted to explore harnessing the power of natural language processing (NLP) to quickly read, understand and correlate information from annual reports into financial analysis write-ups. This would not only reduce effort required by relationship managers to obtain relevant information, but also enable a more consistent analysis approach using AI.
Huanbo: In response to the challenges faced by Don’s team, we built a NLP-powered automation solution that can take a target company’s historical annual reports and produce a draft financial analysis report within minutes according to customisable templates. With this, the team will be able to draft a credit memo within minutes, rather than the usual days. This can enable faster loan processing, which will greatly improve the customer experience for the bank’s clients. The solution also provides fully traceable references to original content behind every sentence and data point, which facilitates quick and easy cross-checking if any doubts arise.
As AI solutions rise in popularity, what sets the AFA solution apart from other intelligent process automation solutions?
Huanbo: While we do see many AI solutions in the market, our Automated Financial Analysis solution is unique as it taps into our key competency in Document AI. It uses our proprietary comprehension-based correlation and reason-finding technology, which can accurately locate and extract information from lengthy complicated documents. Moreover, the solution can process not only documents that are in English, but also those in Chinese and Bahasa Indonesia.
We have also purposely designed the solution to be an AI-human feedback loop, so that we can constantly capture the modifications done by humans and feed those “learnings” back to the algorithm to continuously improve the model.
That’s really cool! So what excites you the most about working with emerging technologies like AI, ML and NLP, and what can we look forward to in this area?
Huanbo: The most exciting part is definitely the intrinsic disruptive nature of frontier technology. That is precisely why our data science team keeps a close eye on the latest technology breakthroughs and research publications across the spectrum. For example, when Microsoft released their new LayoutLM model, we immediately investigated and integrated it into our existing AI engine. The team is also an active contributor to frontier NLP research with several research papers published in top conferences and journals every year.
We have also observed significant development in Computer Vision (CV) in the past few years. In the NLP domain, large scale self-supervised pre-training methods also show promising improvements. While its performance is not quite as proficient as the CV domain, we are sure it won’t be long before they catch up!
What role do you think AI technology will play in the future of finance and risk management?
Don: AI technology will be disruptive and game-changing. I watched a video recently about a conversation between Alibaba’s Jack Ma and Tesla’s Elon Musk on AI. Interestingly, they disagreed on the threat of AI in the future. Jack Ma’s view was optimistic while Elon’s view was that AI will make jobs obsolete. I echo Jack Ma’s view that we will have less work for each person but there will be different type of work to replace those which were lost. That being said, we will need to be prepared and continuously improve ourselves in the rapidly evolving landscape of the industry.
Huanbo: AI will become a revolutionary enabler of smart decision-making in finance and risk management by making processes more efficient and effective. The first step to efficiency and full automation is data digitalisation, and conversion of unstructured to structured data. This can be achieved through AI technologies like NLP and optical character recognition (OCR). In terms of effectiveness, with advanced cognitive AI technology like Knowledge Graph, we can achieve deep knowledge accumulation.
Could you tell us more about your experience collaborating through the DBS Startup Xchange programme?
Don: It was great! The Startup Xchange team was really knowledgeable about the landscape of the solutions that are available. Through their contacts, we were able to quickly shortlist fintech companies with expertise in the space. They also facilitated the engagement with 6Estates and definitely contributed to a wonderful working experience for everyone.
Huanbo: It was definitely a pleasant and fruitful experience working with Startup Xchange. We were put into contact with frontline project members from the risk management team, all of whom were patient and insightful. The Startup Xchange programme also provided great support for the project in terms of focus groups and training sessions on relevant domain knowledge.
What is the biggest challenge you face when it comes to innovating, and how do you overcome it?
Don: A big challenge for us is that we are often bound by our existing processes. Most innovative ideas require us to change the way we work and think, and to challenge our existing processes and workflow. We are constantly challenging the status quo and making sure to ask the right questions. Only by reviewing our processes in a critical manner can we facilitate further development and achieve true innovation.
Huanbo: Another challenge that all AI startups face is bridging the gap between people’s expectation of emerging technologies and its actual capabilities. This challenge goes both ways. In application fields with clear definitions, simple objectives and sufficient high-quality data, such as facial recognition, today’s AI technology can actually perform better than people expect. In many other tasks and scenarios where such pre-conditions cannot be met, such as deep understanding of domain specific text, some people have unrealistic expectations of AI solutions. The only way to overcome this challenge is to improve communication and cooperation between the technology and business worlds.
What is your biggest takeaway from this experience?
Huanbo: This experience affirmed our belief that effective cooperation between deep technology start-ups and leading banks like DBS will create huge value to the financial industry. We are also highly impressed by the Innovation team from DBS, where we see the true ambition to go full speed on digital transformation. They have become the most innovative bank regionally and perhaps even globally, and we applaud them for that.
Don: In our landscape assessment of the available solutions in the market, we noticed that the NLP technology was available, but there was no real-life application to credit financial analysis. After the POC with 6Estates, there was still a lot of work that needed to be done in order to achieve our goal in delivering a viable and functional automated financial analysis tool. We decided as a team to take up the challenge together and built this innovative solution. Our biggest takeaway is thus that common goal can galvanise and gel different team to achieve better results.