Banking and financial industries are leveling up their internal processes with Artificial Intelligence (AI).
Freepik
Share in
In the 90s, with the introducing of internet and various forms of technology, companies began reengineering themselves, integrating technology into their operations to gain a competitive edge. This brought about significant changes, including faster communication, accurate calculations, and improved efficiency. However, despite the power of technology, large-scale integration remained a challenge due to the rigidity of the available software solutions like SAP or Oracle, which provided smooth operations only within small teams or departments.
The arrival of Artificial Intelligence (AI) has revolutionized business process management (BPM).
What does AI offer that is not provided by existing automation technologies?
AI can study vast data pools from a business, provide real-time analytics and insights, and generate customized responses in human language, a capability known as generative AI (GenAI). It not only understands natural language but also provides accurate predictions in mere seconds, a feat that would take humans hours or even days of studying hundreds of pages. AI can analyze data, generate novel ideas, and explain complex concepts using high schooler’s terms.
The power of AI lies in its ability to process and learn from massive amounts of data, these could be companies’ sales data, customer buying behavior, or internal processes data. AI can identify patterns and insights that could take a long time or be very hard for humans to discern.
How do companies take advantage of AI to improve business processes?
- Machine learning: When considering the use of machine learning (ML) to improve business process management (BPM), it’s important to carefully consider the trade-off between generating a more responses versus ensuring high accuracy with less responses. The best approach may vary depending on the specific application and use case. For example, when using machine learning to process internal tasks/issues and mark them as done, understand the criteria to close issues might not be so important, we should prioritize the highest accuracy of the outcome which is closing all the issues that are done. On the other hand, for high-stakes decisions like approving bank loans, it’s necessary to generate detailed explanations, ensuring transparency and justification for the outcomes.
- Natural Language Processing: Natural language processing (NLP) allows businesses to improve internal communication through smart chatbots. The early versions of chatbots are mostly for customers support with multiple choices, the chatbot then generate pre-coded response for each of the choice that customer select. Companies can now create virtual assistants tailored (fine-tunning) to specific departments by feeding the language models their internal data and documents.
- In the video, IBM launched Watson X HR, an AI assistant for HR that learns from a company’s HR guidelines and carries out internal 24/7 conversational support. The virtual assistant is able to understand complex requests like “Can I use PTO (paid time off) for my child’s early dismissal from school every Wednesday?” or “I want to save my paid leave because I am getting married next year and I want to add the leaves to my honeymoon”. The conversations went on with multiple complicated requests from the employees, and the virtual assistant handled it very well, all from the data it was trained on by the company. One of the key importance factors when integrating NLP into a business is the accuracy of the response. What often happens with many large language models (LLMs) is that they generate inaccurate responses (or “hallucinations”). This was handled very well by IBM Watson X HR, as when an employee asked a question that wasn’t covered in the guidelines, the virtual assistant responded that it did not know the answer.
Findings
According to Yahoo Finance, the market size for BPM is projected to reach $60.49 billion by 2031, with a compound annual growth rate (CAGR) of 18.5%. Companies are increasingly using software and automation tools to reengineer, redesign, and optimize their business processes, boosting efficiency and increasing flexibility.
In the banking sector, machine learning has been widely adopted since the 2000s for data structuring and analysis. According to S&P500 Global, by the end of 2022, machine learning accounts for a significant 18% of the total market in finance securities, banking, and insurance.
Beyond machine learning, the adoption of AI has become increasingly prevalent across the banking sector. JPMorgan Chase, for example, has incorporated AI in multiple areas, including marketing, fraud detection, and direct communication with clients, integrating over 400 total use cases in production. Notably the launching of their cloud base home-made Artificial Intelligence (AI) platform, OmniAI. OmniAI helps engineers in JPMorgan Chase streamlining their data and make it ready to use, cutting down the time that engineers need to sort and organize the data, it also helps prevent duplicating processes across the company. Moreover, OmniAI is assisting their employees by proactively providing technical supports in the form of articles by bringing it them in real time.
Some other use cases of AI in banking and finance
Barclays, has deployed an AI-tool for fraud detection that predicts potential instances through real-time monitoring of merchant payment transactions.
Another bank, Banco Santander’s Corporate and Investment Banking division developed an AI-tool called Kairos, which shows how a corporate client could be impacted by economic events, enabling more informed investment and lending decisions.
Bank of America’s platform, Glass, helps sales and trading employees uncover hidden market patterns and anticipate client needs by consolidating market data with the bank’s in-house models and machine learning techniques.
Discover Financial Services, through a partnership with an AI software provider, is using AI for credit underwriting to improve the process and reduce default rates.
Similarly, DBS bank in Singapore leverages AI to reduce false-positive transactions, improving anti-money laundering and fraud detection efforts that were previously challenging with the old rule-based systems required by banking regulators.
Potential damages from bad use cases of AI
The potential of Artificial Intelligence (AI) is undeniably immense for the finance and banking industry. However, if AI is misguided, trained on inadequate or biased data, or exposed to lopsided opinions, it can lead to significant issues.
- Biases: There are several ways AI can produce biased outcomes. Firstly, if the training data itself is biased, the AI model will inevitably generate biased results. Alternatively, even with unbiased data, the design algorithm of the AI model could be inherently biased, potentially leading to unfair or discriminatory outputs. Such biases in AI-driven decision-making can have severe consequences, such as incorrect report presenting or providing financers with bias opinion summary articles. Unfair treatment can also erode trust in banks and financial institutions, undermining financial inclusion and stability.
- Privacy violation: As companies integrate AI for business process management, leveraging large datasets containing employee information, it creates potential vulnerabilities that must be addressed. The more employee data consolidated for AI analysis, the higher the risk if a data breach were to occur, exposing sensitive personal details. Malicious hackers could exploit vulnerabilities in AI systems to gain unauthorized access to this trove of employee records, enabling identity theft, financial fraud, and other criminal activities targeting staff. Furthermore, AI algorithms profiling employees could inadvertently enable discrimination in areas like promotions or compensation if biased data is used. Beyond tangible risks, the perceived surveillance from AI constantly ingesting employee data could erode trust and morale. To mitigate these threats, robust cybersecurity, strict access controls, data encryption, and transparency around AI data practices are crucial. Companies must also follow strict data privacy regulations. Failing to safeguard employee data from AI system vulnerabilities can severely undermine workforce
- Hallucination: A concerning risk in the finance and banking industry’s adoption of AI is the potential for “hallucinations” – incorrect or misleading information generated by AI models. These AI hallucinations can stem from various factors like insufficient or biased training data, flawed assumptions made by the model, learning incorrect patterns, overfitting issues, excessively complex models, or processing errors. In high-stakes financial applications, such hallucinations could have devastating consequences. Imagine an AI system providing inaccurate investment advice, wrongly approving or denying loans, or mishandling fraud detection – the ramifications could include catastrophic financial losses, damaged reputations, regulatory penalties, and erosion of public trust.
Action Plan
Short-Term (0-6 months):
- Define clear objectives and prioritize high-impact internal processes where AI can optimize efficiency, such as document processing, data entry, or customer inquiries.
- Establish a cross-functional team with IT, process owners, and subject matter experts to lead the AI-driven BPM initiative.
- Conduct a data audit of existing processes to identify gaps and prepare relevant data for AI modeling.
- Develop an AI data strategy focused on collecting, cleaning, and structuring process data.
Mid-Term (6-12 months):
- Cultivate a data-driven culture by promoting collaboration between data scientists, process experts, and business analysts for data-driven optimization.
- Invest in AI/data science training for employees to build internal AI capabilities.
- Launch AI pilot projects in selected high-priority processes, closely monitoring performance and iterating.
- Prioritize ethical AI by ensuring transparency, fairness, and addressing potential biases in algorithms.
Long-Term (12+ months):
- Scale successful AI pilots across relevant processes organization-wide, using lessons learned.
- Continuously monitor, refine, and retrain AI models based on evolving process needs.
- Implement robust security for process data used in AI models to safeguard sensitive information.
- Promote responsible AI adoption by providing ongoing training and engaging stakeholders for trust.
Get the best tech tips, AI tool reviews, and more to your email.
You will receive the best editors’ picks from all of our stories, stay tune!
