The landscape of artificial intelligence and its integration into various industries is evolving rapidly, with one prominent player, DeepSeek, making waves particularly in the financial sectorThe financial services industry, comprising banks, investment funds, insurance, and securities, has recognized the potential of AI and is racing to implement DeepSeek's series of models to enhance their operationsMuch like the initial buzz surrounding ChatGPT, there is a lively discourse among financial professionals regarding whether DeepSeek will take over jobs traditionally held by financial workersHowever, the consensus among experts is that rather than replacing these workers, AI will foster a metamorphosis within the industry, driving professionals to become "AI-augmented" talentsFinancial specialists who acquire the skills to collaborate effectively with AI, maintaining an interdisciplinary view that combines technical know-how with industry knowledge, will find greater career opportunities awaiting them.

In a recent commentary, a financial practitioner articulated that in their view, DeepSeek serves not as a disruptor but as an enabler for the finance sectorAs these technologies gain traction, discussions are now aimed at how they will enhance operations instead of rendering human roles obsoleteHowever, experts, when interviewed, cautioned that while deploying such models holds significant promise, the nature of the financial industry, particularly banking, demands stringent adherence to security and compliance protocolsThe outputs generated by AI models, including DeepSeek, can sometimes lack the reliability needed for direct customer interactions, leading to risks if not appropriately supervised

Advertisements

This results in a collaborative approach where AI generates content, but human oversight is required to ensure accuracy and compliance.

The "AI Upgrade Battle" in Finance Is Poised to Begin

DeepSeek has recently unveiled its innovative V3 and R1 models, emphasizing a trifecta of low cost, high performance, and extensive accessibility; these features have significantly lowered the barriers for enterprise-level deploymentThis initiative has sparked a rush across various sectors seeking early adoption of DeepSeek's capabilities, with technology giants leading the wayKey cloud service platforms such as Huawei Cloud, Tencent Cloud, Alibaba Cloud, Baidu Cloud, and JD Cloud have all announced their integrations with the DeepSeek series of models.

Significantly, many enterprises and institutions have begun to integrate DeepSeek into their operations, particularly those in the financial sector, which has historically been one of the most eager adopters of AI technologiesFor instance, on February 2, Haian Rural Commercial Bank publicly lauded DeepSeek via its social media platform, highlighting insights on their market strength, service quality, risk management, and technology support, essentially leveraging DeepSeek for strategic self-promotion.

Taking a more hands-on approach, Jiangsu Bank engaged deeply with DeepSeek

Advertisements

Their digital finance platform, leveraging the "Smart Xiao Su" language model, successfully localized and fine-tuned the DeepSeek-VL2 multi-modal model and the lightweight DeepSeek-R1 inference model for specific practical applications such as quality inspection of smart contracts and automated valuation reconciliationBy incorporating DeepSeek's language model, the Smart Xiao Su has demonstrated enhanced capabilities in handling complex multi-modal tasks, saving computational resources while improving overall efficiency.

Insiders have noted that the DeepSeek-R1 inference model boasts robust reasoning capabilities necessary for processing complex financial datasets and tasksMoreover, its proficiency in context handling and task management, coupled with relatively lower training costs and high performance-to-cost ratios, positions it favorably in applications like intelligent customer service, automated investment advisory, and risk management, enhancing the efficiency of banking operations significantly.

From the data provided by Jiangsu Bank, the integration of the R1 model has facilitated comprehensive automation processes, achieving over 90% success rates in email classification, product matching, transaction entry, and valuation reconciliationThis automation is estimated to save close to ten hours of manual effort each day, reflecting a significant optimization of operational workflows.

In parallel, representatives from state-owned banks have expressed growing interest in DeepSeekThey recognize that the R1 model's open-source nature presents new opportunities, leading to the commencement of relevant research within the institutions

Advertisements

After a short holiday break, discussions surrounding potential applications in intelligent investment advisory, customer service, and regulatory compliance are set to accelerate.

The enthusiasm for AI is also palpable within the public fund management sectorIn recent weeks, over a dozen public fund firms, including Huitianfu, Fuguo Fund, and Noan Fund, have announced successful deployments of DeepSeek's financial models.

In particular, Huitianfu Fund revealed it has completed the private deployment of the DeepSeek series of open-source models, planning to utilize them across critical business areas such as investment research, product sales, risk management, and customer service.

Similarly, Noan Fund has successfully localized the deployment of DeepSeek’s financial models, launching its proprietary "Noan AI Assistant," developed on mainstream AI open-source frameworks and targeting research analysis, customer service, and risk management as initial pilot applications.

Insurance companies are also getting onboardChina Ping An stated it has been strongly committed to ongoing research and application of artificial intelligence and big data technologies to enhance its overall digital transformationCurrently, it is actively exploring the deep integration of its open data platform to bolster the construction of an "integrated finance and healthcare" ecosystem.

In the securities industry, firms such as Guotai Junan, Guojin Securities, and Guangfa Securities have also recently announced the successful localization of DeepSeek-R1 models.

Guojin Securities has finished tests for the localized deployment of DeepSeek, aiming to utilize the model for various scenarios including information retrieval, document processing, industry research, and market analysis, with plans to further expand into intelligent services and investment analysis.

In a transformative move, Xinyue Securities has established a data intelligence platform that accommodates various open-source models, having recently completed the integration of both DeepSeek V3 and R1 into its existing model matrix

This integration facilitates a holistic empowerment upgrade across multiple business scenarios, assisting in efficient knowledge acquisition for employees, enhancing customer service quality, tailoring personalized solutions, and improving R&D productivity.

In the financial technology sector, several firms are making strides as wellRecently, Finance One Account announced the launch of its independently developed intelligent platform, integrating DeepSeek and other open-source models to deliver comprehensive AI solutions for the banking sector.

Accelerating Intelligent Transformation in the Finance Industry

According to analyst Li Bolun from Guotai Junan, local deployment of large models may become a mainstream choice among financial companiesHe points out that given the unique characteristics of the finance sector, there is often a heightened demand for data security compared to other industriesAs a result, financial enterprises typically opt to store their data locallyThe recent release of the DeepSeek-R1 model allows for high-end capabilities to be deployed at a lower cost, enabling firms to combine their local data with the large model to create proprietary models that cater specifically to various business scenarios.

For financial IT companies, Li Bolun notes they possess the necessary capabilities to assist financial institutions in processing their vast accumulated data through cleansing and classification, vectorization, and fine-tuning for local model deployment to develop customized proprietary models

Additionally, these companies have accumulated significant industry knowledge, aiding clients in integrating their business needs with specialized model capabilitiesThis allows for the creation of tailor-made AI workflows, RAG (retrieval-augmented generation) processes, and unique agent solutions.

The development of AI large models and their practical applications is a focal point in the financial sector, particularly within banks, which are striving for accelerated digitalization against a backdrop of challenges such as efficiency improvements, risk management, and customer experience optimizationTraditional technological solutions are often inadequate in meeting increasingly complex demandHowever, the rise of AI large models presents a new breakthrough opportunity for the industry.

Shen Wenbin, the AI Product Director at Finance Cloud’s Banking Division, expressed that the release of the V3 and R1 models from DeepSeek possesses the outstanding general capabilities of a MoE (Mixture of Experts) architecture, leading to top scores across various evaluationsTherefore, the scenarios currently implementing large models in banking are universally applicable to DeepSeek.

DeepSeek-R1, for example, is lauded for its high-performance reasoning, reinforcement learning capabilities, and long-chain reasoning support, particularly excelling in fields involving mathematics, programming, and scientific inquiryThese characteristics are particularly relevant to banking, where extensive data processing and decision analysis are integral to operations.

In the realm of risk assessment, for instance, DeepSeek-R1 can effectively integrate multidimensional client data to create more accurate risk modeling, enhancing the evaluation of factors like credit and market risk

In customer service and marketing sectors, DeepSeek-R1 can precisely understand customer intentions and identify potential needs to smartly recommend suitable financial products.

The Foundation of Artificial Intelligence Remains Human Ingenuity

Despite the compelling capabilities presented by DeepSeek, there are growing concerns regarding its reliability as its applications become more prevalentUsers have reported instances of DeepSeek providing inaccurate outputs, particularly in generating academic content such as referencing non-existent materials or incorrectly linking to unrelated studiesIn the context of the high-stakes finance sector, reliance on these inaccurate outputs could lead to disastrous scenarios, such as flawed market analyses leading to misguided investment strategies.

Moreover, the application of large models necessitates managing vast amounts of personal and corporate data, which heightens the risk of information leaksSuch breaches could potentially result in the misuse of customer information for fraudulent schemes, tarnishing the reputation of financial institutionsIndustry professionals have noted that the integration of AI technologies like DeepSeek, while enhancing operational efficiency, could instigate multifaceted risks, including issues related to data security and privacy infringements, data contamination, regulatory challenges, and ethical concerns regarding trust.

In addressing these concerns, DeepSeek acknowledges that while the benefits of integrating AI with financial operations are significant, they come alongside complex risk factors

Challenges such as the interpretability of AI outputs and the inherent unpredictability of large model outputs pose potential hazards when interfacing directly with clientsConsequently, some current applications continue the practice of AI generating content under the supervision and verification of human operators.

Wang Peng, a researcher at the Beijing Academy of Social Sciences, points out that deploying DeepSeek involves overcoming challenges related to data quality, interpretability of models, and market adaptabilityBanks need to establish rigorous data quality management systems to ensure accuracy and completeness, while also improving their understanding and transparency regarding DeepSeek modelsSimultaneously, banks must continuously adapt their application strategies to align with changing market conditions, ensuring that the solutions remain relevant to their operational essentials.

Many industry insiders predict that an increasing number of banking institutions will adopt advanced language models, primarily to enhance efficiency across employee operationsGao Feng, the Chief Information Officer of the China Banking Association, highlighted that the current utilization of large models in the banking industry chiefly revolves around internal scenarios such as smart coding, AI-driven office tasks, and enhanced customer service management, rather than core business operations like account transactions.

Notably, the well-regarded economist and information communication expert within the Ministry of Industry and Information Technology, Pan Helin, remarked that the distillation algorithm employed by DeepSeek has sparked interest within the banking industry

Advertisements

Advertisements