Autonomous CRM Agents: Using LLMS for Loan Processing, KYC and Regulatory Adherence in Salesforce

Abstract

Customer relationship is a key area where advanced technologies such as Large Language Models (LLMS) are being adopted extensively in the realm of the financial services industry. The following paper focuses on the practice adoption of self-sustained Self-service CRM Agents rooted in LLMS for Salesforce ecosystems for loan processing, Know Your Customer or KYC operations and/or compliance management. Salesforce, as a leading CRM platform, offers extensive capabilities for customer data management and interactions. The introduction of LLMS brings in new chances to automate work processes, enhance decision-making and guarantee compliance more flexibly, effectively and at a larger scale. We have employed LLMS to undertake activities such as loan application processing, customer identification through the KYC process and conforming to the regulatory compliance standards and requirements. The features of these autonomous agents are presented below: These are self-contained and self-managed programs that use NLP and machine learning for data analysis on the customer data and production of the required letters, generation of risk profiles and generation of decision support to the financial institutions. Besides, they are made in a way that would enable the attainment of full regulatory compliance with the use of checklists against the existing national and international laws in real-time. Lastly, this paper also describes the advantages associated with the implementation of LLM-powered CRM agents, including less failure rates, time-saving and overall better customer satisfaction. In addition, it talks about the threats, including data security issues, the system of laws that is difficult to navigate and the requirement for the model update because of the constant changes in the rules. New findings are taken from case experience and pilot undertakings in different financial organisations utilising Salesforce’s stage and the importance of efficiency and customer satisfaction is discussed.

Keywords: Autonomous CRM agents, Large language models (LLMS), Loan processing, Know your customer (KYC), Regulatory compliance, Salesforce CRM, Natural language processing (NLP), Machine learning, Financial services

1. Introduction

1.1. Importance of autonomous CRM agents

Figure 1: Importance of Autonomous CRM Agents.

• Increased operational efficiency: Another advantage of utilising autonomous CRM agents includes time-saving, since some of the many activities or day-to-day operations that are usually handled include answering customer inquiries, scheduling and order processing. These include performing simpler repetitive and consumptive tasks that would otherwise allow human agents to attend to more important issues relating to service provision, such as handling of complaints or strategising. This is because businesses can be able to manage numerous clients’ complaints and inquiries at a time, hence increasing efficiency and decreasing operational expenses of having to recruit and train a bigger number of employees. Thus, the businesses can offer the services that are prompt and effective to their consumers and this effect is very important in today’s world as it defines the competitiveness of different organisations.
• Enhanced customer experience: Such CRM agents based on AI and machine learning and more specifically the LLM, have the potential to serve the customers at any time throughout the day and interact with them at a very personalised level. They have the potential to operate with customer data and engagements, to be able to present recommendations and solve inquiries, as well as predict the needs of customers. Such a high level of personalisation and availability is helpful to the customers since they are promptly provided with an accurately responding person, as they are satisfied. In using self-assist services in industries such as banking, retail and telecommunication, where customers can be repeatedly exposed to the company, the organisational performance will be greatly improved through self-assist support agents.
• Improved decision-making through data insights: Self-sufficient CRM agents have the potential to generate a lot of customer information and data. It is because they can monitor customers' activities, choices and buying habits so that the business gets the full view of customers and market trends. This way of analysis can be helpful to assist businesses in making proper and proper decision-making processes in the context of marketing or product management, as well as for the division of clients. The knowledge derived can also be used effectively for improving the business’s CRM tactics and even predicting future trends, changing customer expectations and thus sustaining a competitive edge.
•  Cost reduction: The use of the autonomous CRM agent cuts out a lot of cost that would otherwise come from the use of manpower and operations. They can help shift workers’ attentions from mindless, repetitive work to higher-level or valued-added or customer-oriented work. On the same note, automation also entails the ability in actual business operations to interact more often with their customers without having to employ more people. In addition, automation of activities like billing order processing and customer support has its way of minimising mistakes and enhancing the flow of processes, hence lowering costs. In every organisation, the cost incurred by agents during their initial stage of establishment overpowers the numerous benefits that will be accorded by their independence in the later years.
• Scalability and flexibility: The general advantage of other intelligent customer relations management agents is the scalability factor, which is not present in other models of customer service. That is, when businesses expand or when there is fluctuating demand in the market, agents can be expanded at the same proportion in order to cater for the increased business flow. Whether it is a situation such as a turn in the volume of customer calls during the winter, summer, spring or the festive season or situations such as diversification of operations with the realisation of other markets, these agents can flexibly adapt to enhance their performance in the company. Again, the flexibility enables the business to deliver the same level of service regardless of the level of operation and the opportunities in the market will always be exploited with ease.
  
  

1.2. Using LLMS for loan processing, KYC and regulatory adherence in salesforce

The use of Large Language Models (LLMS) in Salesforce for automating sensitive operations in the financial services industry, like loan processing, KYC verification and compliance and regulations, is making this shift progress in the financial services industry. Integration with LLMS, GPT models by Openai, introduces an enhanced level of NLP into the Salesforce environment, which helps to automate high-level, data-based operations that used to involve agents^4,5. In the processes of loan, LLMS can perform well in processing loan forms, identifying risks from historical data and reporting to the loan officers to support quicker and better decisions in loan approvals. Therefore, through operationalisation of the decision-making process, LLMS ensure effectiveness through efficiency gains, decreased workload, increased accuracy and less variation. In Know-Your-Customer verification procedures, LLMS can verify the customer details mechanically by comparing the documents provided by them, which include passports and drivers’ licenses from their official databases and records. It is a result of integrating the LLMS in this process enables speeding up the identity verification process and decreases the possibilities of mistakes, effectively minimizing fraud. Another benefit identified included the efficiency of the KYC procedures in preventing money laundering and other unlawful acts due to automation. With regard to compliance, LLMS can constantly monitor and update it with the latest on regulations such as AML and data protection laws (such as GDPR). It means that instead of frequently intervening with checks and balances, as is usually done in ordinary organisations, they stay loyal to their overall legal requirements and can adapt to changes in those requirements within the shortest time possible, thus avoiding fines and legal troubles that may be incurred in the process of non-compliance. Through the use of LLMS in the integration of Salesforce, the financial institutions can enhance the following: Operational efficiency, Compliance, especially given the dynamism in the financial market regulation.

2. Literature Survey

2.1. Autonomous CRM agents in financial services

There is a vast literature regarding the idea of autonomous Customer Relationship Management (CRM) agents in financial services. Other scholars, also limited their prior studies to explain how customer services can be automated in a way that may include responses to inquiries, complaints and product suggestions^6-9. These early pioneers provided the basis for more complex implementations where the agents’ role involved performing more extensive interaction with customers. Modern improvements, as it was described, have included the application of AI technology in CRM systems in the financial industry. Here, the idea is the performance of mundane activities like client solicitation, credit appraisal and customer treatment. With the technological advancement in NLP and adoption of ML, current self-sufficient autonomous agents have capabilities of analysing the behavior pattern of the customers, personalising services and even increasing the efficacy at the customer touchpoints. Among the leading CRM systems, Salesforce has also been instrumental in such innovations with the implementation of AI functionalities that perform customer interactions independently. In each of these applications, Salesforce is responsible for the ability to provide faster responses to customer queries, automated loan applications, as well as efficient marketing campaigns. But there are still relative issues such as the struggle to increase the speed and accuracy of the systems through the application of automation and the necessity to adhere to highly strict financial services regulations, for example, data privacy and protection.

2.2. Large language models in financial services

Nowadays, it has gained enormous power with models like GPT series from OpenAI and BERT in a variety of text processing tasks such as language understanding, generation and summarization. These models prove helpful when it comes to handling tasks that consist of dealing with human language, which in one way or the other is helpful within the financial service sector. As stated, LLMS have benefits in the way of improved customer service since it is possible for an AI chatbot to help in responding to the basic questions that a client or customer may pose, answering FAQS and offering essential financial-related information and handling routine queries. Also worth mentioning, LLMS has also been used in automating some of the functions of the loan processing system, thus increasing its speed of data analysis of financial documents and their data extraction, as well as its lending decision-making. Nevertheless, its usage in even more critical domains such as KYC/KYC and other compliance tasks is still in its developmental stage. As stated by Wu, et al. in the area of compliance regulation, LLMS can be trained to search for risk factors and to identify fraudulent activities; this would involve comparing clients’ data with the provisions of the law and regulations. Despite all these prospects opened by LLMS, the integration of LLMS in the financial services sector has its issues, including data secrecy, ethical implications and mostly the question of satisfying regulatory requirements to standards set in the financial industry.

2.3. Regulatory challenges and adherence

When it comes to the application of AI and autonomous agents, the issue that social actors pointed out as particularly significant is the issue of compliance with regulatory bodies in the financial industry. Financial organisations continue to work under different legal requirements and standards, for instance, the General Data Protection Regulation (GDPR), the Anti-Money Laundering (AML) policies, among others, with the main aim of protecting the customer's data and financial sector systems. The process of applying artificial intelligence in the automation of compliance with all these regulations is not easy, bearing in mind that such regulations are dynamic and they differ from country to country. The ever-changing financial regulation is a factor that keeps the AI systems under stress as they need to update themselves. The application of autonomous agents, especially those using LLMS for automating adherence, requires frequent updates and recalibration of algorithms to accommodate the dynamicity of the law. In addition, LLMS should perform an understanding or rather the ability to analyse numerous authorities that involve significant consumer laws and regulations, many of which consist of legal language and the changeable meanings of legislation. This constant demand for system upgrades and configurations presents a continuous concern to the financial institutions that wish to incorporate AI while being fully compliant with laid down rules and regulations, not to neglect the sound security and confidential management of customers’ data.

2.4. Salesforce CRM and AI integration

As one of the pioneers in the CRM sector, Salesforce has been implementing the use of artificial intelligence in its products to assist companies in delivering efficient customer service. Salesforce has several artificial intelligence integrations, the most popular of which is called Salesforce Einstein, which helps to make predictive analytics, sales forecasting and automate many processes. Argues that Salesforce Einstein has advanced in achieving such automations, including the lead scoring, customer segmentation and marketing automation that enhances the efficiency in CRM. In the financial services domain, it started revolutionising the industry when it was implemented in different areas, including loan processing and automated customer service. This would make the financial institutions respond to the customers swiftly, automate the approval of loans and enhance customer interaction, hence improving their satisfaction levels through the application of Salesforce’s intelligence. Nevertheless, there has been a minimal amount of work done in the application of self-acting agents based on LLMS for important operations, including compliance with regulatory requirements and the KYC process. By considering the Salesforce CRM framework, the following key findings can be made about the use of LLMS in an unexplored role for LLMS in relation to automating the following compliance, financial analysis and KVC processes within Salesforce CRM systems. There is potential that incorporating these technologies into Salesforce’s ecosystem will bring a lot of benefits, may be almost all the advantages listed above, but they come with proper problems that have to do with data security, ethical implications and compliance that must be considered and overcome for the solutions to be good and right.

3. Methodology

3.1. System architecture

Figure 2: System Architecture.

• Salesforce CRM platform: Salesforce is the primary CRM system and contains customers’ personal data, loan requests and KYC data. Salesforce is largely based on low code, which means that custom applications built in this platform can easily interface with third-party applications such as LLMS and compliance modules. Salesforce also has some degree of automation already dispersed throughout the CRM, thereby allowing financial institutions to track information regarding customer relations, loan progress and even regulatory data. This way, it offers a convenient tool to the interface for the employees; at the same time, all data is safely stored as well as readily available for analysis and preparing the necessary reports.
• Large language models (LLMS): LLMS or large language models for dipping, such as Openai’s GPT series or similar products of competition, are interconnected with the system through APIS for purposes of processing the customers’ queries, loan applications and other documents. Concerning loan processing, LLMS cleans, sorts and summarises loan application data, credit risks and creates reports for financial companies. In KYC verification, these models compare the documents submitted by customers with official records as a way of ascertaining the truthfulness of provided information. The skill of the LLMS in natural language processing makes it possible to complete complicated client exchanges and contract-related documentation much more efficiently than by checking them manually.
• Compliance module: The external compliance module focuses on the legal compliance of the loan processing and the customer verification performed during the pre-processing step by reconciling the standards of AML, KYC and the GDPR, inter alia. This particular module is engaged in the process of ensuring that the system complies with the existing laws and regulations, up to the amendments made. With this module incorporated in the system, the system is capable of ensuring that every customer interaction is compliant with the current regulations on processes such as loan approval, data handling, among others. By using the implemented module, the institution can have real-time control and observe any deviations or infringements and thus minimise legal liability and implications.

3.2. Data flow

The Data Flow Diagram (DFD) demonstrates the flow of data throughout the system, stressing specific transactions that take place between the activities that deal with customer data and loan applications, as well as with correspondents that are applicable in ensuring compliance with regulations. Customer-related data is the first data in the flow process; this consists of several aspects, for instance, personal information, loan documents and the KYC, which stands for know your customer documents. This data can be entered by the customer or can be uploaded through the CRM interface. The information can exist in simple written text, electronic images of documents like ID and even structured data such as filled-out forms and application fields. When the data is obtained, it is passed to the next level, the Autonomous CRM Agent. There are two main functions in the case of the Autonomous CRM Agent that are performed with the help of a Large Language Model (LLM). First, the agent takes customer information from the loan application, where the agent parses data from customer input such as income, credit score and loan amount. Through the application of LLMS, the person's KYC checks are also conducted through comparison with official information from documents filed and databases to ascertain that the customer’s identity is legal and compliant with the set regulations. Also, the CRM agent can recommend potential risks to the loan approval by comparing the customer's characteristics and their credit score against certain risk models. Through the use of LLMS, most of the work that is involved in decision-making on the value-added services for customers and management of loans is done by this agent. After this, the data proceeds to the next unit, depicted as the Regulatory Check, which is operated by the Compliance Module. This module helps to compare and identify discrepancies in the processed data against the given legal requirements as per AML norms, GDPR norms and other financial laws. The compliance module enables all the customers’ interactions, loan approvals and verifications to be within the legal requirements. If there are any complications or problems, they can then be highlighted for a supervisor to go through and therefore at the same time ensuring that legal and other important requirements are met in processing the information. Such integration of customer interaction, risk assessment and regulation check algorithmically showcased how the CRM system and the LLM-powered autonomous agent come with the compliance module to create a continuous cycle of efficient workflow.

Figure 3: Process Flow for Loan Processing,

• Customer submits loan application: The logistics for the processing of the loan start from the time when the customer applies for the loan^15-18. This can be done through the CRM interface and affords them the necessary personal details, financial details and the amount of the loan they would like to borrow. It is accepted that the additional documents could include pay stubs, a credit check or an identity card (like a passport or driving license). The submission process is also streamlined to the extent that if any of the required boxes are not checked or any of the files are not uploaded, they cannot proceed to the next step.
• Extract customer information (LLM): The next step that follows the submission of the loan application is to utilise large language models to extract possible key details from the application, as well as any documents that accompany the application. The LLMS browse over the traffic and extracts information such as name, address, income, employment and the amount required by the applicant. Hence, the models employ OCR and NLP methodologies for reading the contents of documents such as bank statements or identification. This makes the process faster and makes the data to be accurate as compared to when it is processed manually because it may include human errors.
• Risk assessment: After the details extraction, the process proceeds to the risk identification part. Analysing the application, different machine learning algorithms help in determining the financial stability and creditworthiness of the borrower. This score is arrived at through such factors as credit score, income and debts, employment record and any other necessary financial factor in order to establish the risk level. This score is useful in identifying the likelihood of the borrower to default on the loan. Machine learning is based on historical datasets and can thus update its risk assessment going forward in a continuous manner in the loan application assessment to increase the accuracy of the model utilised.
• Document generation: After the risk assessment has been made, the system also prepares all the papers that are required for the actualisation of the loan release profile. This may encompass the loan agreement, any terms and conditions formulated and any other documentation that may be necessary for legal purposes. Such documents can be produced by the LLMS of the system that analyses the application information and the output of the risk assessment stage. These are defined and attached as annexes to the particular loan and its particular customer to guarantee that all legal and financial requirements are proper. One of the advantages of automating a document is that it takes less time to produce when compared to manual writing, since the often-used blocks of text can be simply inserted into the document.
• Compliance check: To reach this type of loan, pre-approval tests are run to see whether the loan meets with credit, legal and other regulatory frameworks. The new feature works as compliance built into the loan application and checks for AML and KYC policies or another regional requirement, such as GDPR. It compares the customer’s information and documents provided with the corresponding database and legislation to avoid issuing a loan to a person who does not meet legal requirements. This particular step is a critical one which will help to maintain the stability of the financial institution and help the company to avoid any legal troubles due to non-adherence to legal requirements.
• Final decision: The functionality of the program, when used in the decision-making context, refers to the ability to approve or reject a loan application after going through all the data gathered throughout the business process. This decision, therefore, depends on the extracted customer information, risk score and the results of any compliance check as may be conducted. It is imperative that a loan only proceeds to this stage if it fulfils all the laid-down qualifications and fits the set guidelines and regulations. Upon being approved, the relevant documents are forwarded to the customer for signing. If the application does not fit the requirements, it will not be accepted and the customer is informed about the rejection with the reason given. It has reduced chances of humans interfering, hence making it fast and effective with little chances of their interfering with the set standards and legal requirements.

4. Results and Discussion

4.1. Case study: Loan processing in a financial institution

This case involves loan processing in an unspecified financial institution and its employees were suspected of being involved in the embezzlement of some of the borrowers’ funds. Since this paper looks specifically at the Bartram Group’s case of its mid-sized financial institute customer employing Salesforce’s CRM platform with LLM-driven autonomous agents in the loan application processing, the case study was undertaken in this setting. During this period, the system served a total of 500 loan applications (Figure 4). This case study made their experience in using an automated system useful for understanding how efficient it is, how much time was saved and how much independent work was done. The outcomes prove that LLMS can optimise the procedures of loan operation, with the necessary level of efficiency and precision achieved (Table 1).

Table 1: Loan Processing Results.

• Number of applications: 100%. However, from the time of the automation to the end of the study period, the number of applications processed was still 500, meaning that the automation did not affect the quantity of loans. However, the system improvement efforts were geared towards achieving greater speed and efficiency of the system in handling a constant number of applications. The 0%, as the one based on the number of applications received during the training period, is expected, as it was not aimed at increasing the number of applications, but at enhancing the speed of their processing and limiting the interference of people in that process as much as possible. This proves that the purpose of the automation system in its development was more directed towards the bureaucratic internal flows rather than towards more submissions from the customers.
• Average processing time: 40%. Some of the most expressive gains mentioned within the case study are related to the mean processing time. Before the use of the innovative technological system, the common time for appraisal of loans was a few hours because of the data entry, document review and assessment of risk. From here, the involvement of LLM-powered CRM agents was able to cut the processing time in half. This means that the average time taken in the processing of each loan application was reduced, hence shortening the time taken in loan approval. It also ensured that the company was able to respond to the needs of the customers in a faster way, thus increasing customer satisfaction levels as well as customer operational satisfaction levels year-on-year.
• Human intervention rate: 80%: Another was the cut of the human interferences needed during the loan application processes by 80 per cent. Previously, these service functions relied on agents who checked entries and verified information, evaluated risks and adherence to the rules, etc. Nevertheless, many activities, including information extraction, risk assessment and document creation based on machine learning algorithms, were automated and human input was required in high-risk cases, such as different financial discrepancies or fraud. Besides, it therefore saved human resources for more important tasks and also made the financial institution work effectively and eliminate many instances of human error.

4.2. KYC verification

The task of implementing LLMS in automating the KYC verification process, in particular, revolves around covering operational value in terms of ensuring the accurate validation of identity documents, including passports, driver’s licenses and utility bills. The system was created with the capacity to compare documents received with data contained in government databases and other official documents. Using techniques in NLP as well as OCR, it was possible to validate customer identity and identify any discrepancies. The main aim of the study was to optimise the rates of accuracy of documents and lower the levels of False positives or acceptable purchasers who are identified as fraudsters and false negatives or the fraudsters who are not identified as such (Figure5). The following is the evaluation of the changes that occurred in the various indicators (Table 2).

Table 2: KYC Verification Accuracy.

Figure 5: Graph representing KYC Verification Accuracy.

• Accuracy rate: 95% → 98% (Improvement: 3%): Originally, the accuracy of the KYC verification was at a 95% rate, which implies that only 5% of the verification process was erroneous, while 95% of the results were accurate. As a result of introducing LLMS, it improved to 98%, which is a 3% improvement. This enhancement can be regarded as showing that the system, powered by LLM, worked better in successfully verifying identity documents, thereby increasing the efficiency of the method used in KYC. Accuracy rose, leading to fast and correct stamping of the identity of the customers, rather than resulting in wrong validation of the identity of the customers.
• False positive rate: 5% → 3% (Improvement: 15%): The missed detection rate, besides, is defined as a rate where some of the genuine customers are identified by the system as suspicious or invalid and rejected. Previously, without the use of automation, the level of false positives was 5%, which implies that 5% of the customers were wrongly flagged for further checks or rejection. With the LLM integrated for document validation, this rate came down to 3% for the whole procedure, thus indicating an improvement of 15%. This is important from the aspect that there are fewer good customers held for additional checks, thus increasing customer satisfaction due to a reduction in delays. Limitation of Work: While using the described methods for facial recognition, it is crucial to limit the number of false positive results to provide customers with fast and efficient onboarding.
• False negative rate: 3% → 2% (Improvement: 33%): The false negative is the actual number of fraudulent or suspicious customers that, for some reason, had not been identified by the system. The automation impact of the false negative before the change was 3%, which implied that 3% of the customers writing cheques with the intent of fraud went through the verification without being flagged. On the other hand, the false negative after LLMS integration was reduced to 2 per cent, signifying a reduction of 33 per cent. This has a positive impact as it makes the security and integrity of the KYC process tighter and fewer identity fraudsters will be overlooked. Lower false negatives is a necessity for reducing financial crimes and minimising the legal and regulatory loss for the institution.

4.3. Regulatory adherence

It remains a significant factor for any organisation to adhere to the laid-down laws throughout the loan processing and customer verification, especially for firms in the financial sector. It has been thoroughly checked for compliance with the AML and KYC standards for it to be used in the company. These restrictions are aimed at fighting fiscal offences, including money laundering and fraud, as well as maintaining the customers’ confidentiality by enabling the financial institutions to identify as well as authenticate their clients. Another advantage of the system is that it is designed to check compliance in real time and since the financial regulation is a dynamic field, this is an important aspect. Exclusively, compliance processes have been revised every quarter or every year, which can cause gaps and failures. However, with the application of the mentioned automated system, the institution could guarantee that all the transactions and loan applications were processed to adhere to the existing legal provisions. Apart from AML and KYC compliance, this real-time checking mechanism helped the institution avoid any legal implications of typos, legal cases or penalties that may result from conducting a compliance check done online that may be either outdated or contain typographical errors. The system remained updated with the current regulatory updates, for instance, changes in the AML laws or current guidelines from the regulatory authorities and any of these changes were implemented in the loan processing system. This strategy of the institution proved effective in the early identification of emerging regulations and the institution was reputable for its compliance. This paper found that by automating this process of compliance with the regulation, a lot of dangers that could lead to fines or detriment to the worth of the institution are eliminated. In the long run, the systems helped to ensure that each loan that worked through the channel and each identity verification was compliant with the necessary regulations to reassure both the regulators and the customers.

5. Conclusion

In this paper, a detailed discussion has been made about the successful implementation of Large Language Models in Salesforce CRM to improve the key operations of the financial services industry, including loan processing, Know Your Customer (KYC) verification and compliance. Thus, the actual work done with the help of LLMS was the automation of several time-consuming processes that were previously required to be performed by human operators and bureaucrats. Apart from such, the use of these solutions also made application for loans relatively efficient with a reduced chance of having human errors lead to a noncompliance situation and reduced efficiency in customer service.

The use of autonomous CRM agents that enhance their operation through LLMS has evidenced well-expected advances in various aspects. Regarding loan processing, the efficiency increased to 60% as compared with the previous processing time means that more applications can be processed through the system in a shorter period of time; throughput. Consequently, the automation of KYC verifications came with increased accuracy of 98%, thus a low false positive and false negatives, meaning that most of the genuine customers did not escape being flagged and most of the fraudsters did not escape being declined. These enhancements not only benefit the customers and the institution but also impact the security and compliance with regulations. Furthermore, the system’s centralised and real-time performance update of the regulatory checks made it possible for the DRIVE system to adhere to AML and KYC regulations to correspond with the most up-to-date legal requirements on the completion of all transactions and customer verifications. This effectively means that it can easily adjust to these changes and that it can do so promptly, thus helping the institution to avoid scenarios whereby it is non-compliant with these regulations and thereby faces the consequences of the same.

Thus, the further development of LLM models heading into the next stages of this research will be aimed at increasing their efficiency in terms of accuracy and the model’s expanded capability to take into account such aspects in the regulation of international trade. This will involve using the models on new and more extensive datasets that will be developed and applying new regulations as they crop up. Thus, further enhancing the work of the system’s flexibility and effectiveness, the financial institution will be able to use AI as a capable tool for reducing costs and increasing customer loyalty while maintaining compliance with the most stringent industry regulations. In conclusion, the constant advancement of autonomy in these systems is going to be transformative in the automation of the financial services industry.

6. References

1. Wulf J, Meierhofer J. Exploring the potential of large language models for automation in technical customer service, 2024.
2. Akyüz A, Mavnacıoğlu K. Marketing and financial services in the age of artificial intelligence. Financial Strategies in Competitive Markets: Multidimensional Approaches to Financial Policies for Local Companies, 2021: 327-340.
3. Antão J, Pereira R, Ribeiro R. Mobile CRM development for real estate agents. Property Management, 2022;40: 457-483.
4. Ibrahim R. Formalisation of the data flow diagram rules for consistency checks, 2010.
5. Rajola F. Customer relationship management in the financial industry: organisational processes and technology innovation. Springer-Verlag, 2019.
6. Tsai WC, Chen AP. Service-oriented architecture for financial customer relationship management. In Proceedings of the second international conference on Distributed event-based systems, 2008: pp. 301-304.
7. Olszak C. Multi-agent framework for social customer relationship management systems. In Proceedings of the Informing Science and Information Technology Education Conference. Informing Science Institute, 2013: 367-387.
8. Yang Y, Uy MCS, Huang A. Finbert: A pretrained language model for financial communications, 2020.
9. Li Y, Wang S, Ding H, et al. Large language models in finance: A survey. In Proceedings of the fourth ACM international conference on AI in finance, 2023: 374-382.
10. https://360degreecloud.com/loan-processing-qualification-kyc-automation-salesforce/
11. Huang AH, Wang H, Yang Y. Finbert: A large language model for extracting information from financial text. Contemporary Accounting Research, 2023;40: 806-841.
12. Liu XY, Wang G, Yang H, et al. Fingpt: Democratising internet-scale data for financial large language models, 2023.
13. Kumar MA. AI-Driven CRM: How Salesforce Einstein is Revolutionising Customer Engagement, 2023.
14. Parmar D. Enhancing customer relationship management with Salesforce Einstein GPT, 2023.
15. Ibrahim R, Yen SY. An automatic tool for checking consistency between Data Flow Diagrams (DFDS). World Academy of Science, Engineering and Technology, 2010;69.
16. Ghasemaghaei M, Ebrahimi S, Hassanein K. Data analytics competency for improving firm decision-making performance. The Journal of Strategic Information Systems, 2018;27: 101-113.
17. Cheema SM, Tariq S, Pires IM. A natural language interface for automatic generation of data flow diagrams using web extraction techniques. Journal of King Saud University-Computer and Information Sciences, 2023;35: 626-640.
18. https://www.salesforce.com/in/financial-services/ai-in-banking/
19. Blevi L, Delporte L, Robbrecht J. Process mining on the loan application process of a Dutch Financial Institute. BPI Challenge, 2017;448: 328-343.
20. Darapaneni N, Kumar A, Dixet A, et al. Loan prediction software for financial institutions. In 2022, Interdisciplinary Research in Technology and Management (IRTM), 2022: 1-8.