No matter in the past or in the future, the role and meaning of finance will not be subverted, but will only be clarified, and will only return to the original.
Author | Sihang
Editor | Pi Ye
Produced | Industrialist
"This year, we were forced by the market to do digital transformation. Everything was disrupted, and such a transformation was unprecedented. By August, more than 20 additional budgets were made. Every time a new platform and new business line are opened It all means budget redoing," a northern food brand told us.
In the process of digital transformation, all problems, large and small, will be reflected in financial management. For a company to undergo digital transformation, what needs to be changed is not only the business operation method, but also the financial management concept.
Taking the northern brand mentioned above as an example, in the process of moving from offline stores to online e-commerce, corporate customers need to face changes in many aspects. From e-commerce platforms to online sales platforms, the entire chain must be digitized.
The digital transformation has been going on for more than half a year, and the launch of various new platforms and channels, as well as the development of product lines and business lines, have overwhelmed the traditional financial management methods in the past. Previously, the company would prepare a comprehensive budget management every year. However, more than 20 budget adjustments have been made in the past six months, revealing the fragility of corporate financial management methods in the past.
What this northern brand has experienced is becoming a microcosm. Behind this is a common problem encountered in the financial aspects of all domestic enterprises in the digital transformation.
In this regard, industrialists communicated with a number of financial and taxation SaaS manufacturers, and the unanimous conclusion was that due to various factors, the current degree of financial digital transformation of small and medium-sized enterprises in China is not high; traditional financial management methods are not only inefficient, but also risk-resistant. Difference.
“Digital transformation remains a difficult issue due to technical, process and cultural challenges. Even more SMEs still rely on traditional paper-based reimbursement and manual financial management processes, which are not only inefficient, but also have high error rates and delays . " The enterprise expenditure management platform "decibel pass" revealed to industry experts.
In addition to the above-mentioned problems, there are more new problems emerging in the field of financial management, such as fee standardization, data compliance, privacy protection and so on. The intertwining of old and new problems not only brings challenges to financial management software, but also affects the implementation of large models in the financial field.
Financial digitization, travel to the depths of the water.
1. A test of the financial model
In this competition of large models, the big manufacturers are racing rapidly. In specific vertical scenarios, To B manufacturers have entered the market one after another, and finance and taxation are no exception.
In the financial field, Kingdee, which started from financial software, was the first to enter the poker table.
"No company can make a complete large-scale enterprise model, so Kingdee first chose finance, which is the field we are most familiar with." Zhao Yanxi, executive vice president of Kingdee China and general manager of R&D platform, once told the media.
As an enterprise management software service provider, the reason why Kingdee chooses the financial field as its first choice is based on its practical experience in domestic replacement of business and financial integration for 177 major customers within two years.
It is understood that Kingdee has made many internal explorations for the large financial model. For example, since March and April, it has cooperated with Baidu, Tencent, Huawei and other large model manufacturers to select models.
Another example is Kingdee's prefabricated financial knowledge base, which has already embedded domestic tax incentives into the model for direct use by enterprises. For small and medium-sized enterprises, it can be used out of the box. For large enterprises, Kingdee's Sky GPT can also allow them to customize.
In addition, in addition to comprehensive management software vendors such as UFIDA and Kingdee, leading SaaS companies in the enterprise expenditure management track such as Fenbeitong and Hesi (formerly Yikuaibao) are also practicing their own large-scale model attempts and explorations.
Wu Rongbin, head of Fenbeitong Big Data and Algorithm, believes that in the long run, the financial large model can help enterprises to carry out real-time intelligent cost control, make more scientific decisions based on the enterprise's financial data, and predict future trends based on historical data and algorithms. Financial status.
However, in the short term, Wu Rongbin believes that "the low degree of digital transformation of small and medium-sized enterprises, the superposition of problems such as data security and privacy protection, as well as the challenges brought about by data quality, technology and resource requirements, and the intertwining of old and new problems make it difficult for large financial models to Realize its real industrial value."
Wu Rongbin told the industry, "We have implemented several application scenarios internally, but we are still in the initial stage of exploration. The direction of exploration is at the application layer, mainly depending on where the boundary of the large model is in the application. In addition, in order to improve the user experience , we also explored and optimized the overall technology stack of the large model, so as to vertically optimize the application scenario.”
"Frankly speaking, none of the technologies we use now is the technology of the past five years or even the past ten years. At present, Workday is using the theory of 25 years ago to guide the current software development, and this is actually the vast majority of The challenges faced by enterprise management software. Similarly, whether it is cloud native a few years ago or the current AIGC, when these technologies are maturing, the most important thing for enterprises to think about is how to combine technology with business depth. ” Tong Peize, Chief Product Architect of Si, told the industry experts.
In his opinion, the current technology has developed to a very high position, but whether the technology can be combined with the challenges and problems encountered in the business operation process to jointly improve and manage is a matter for the vast majority. The most important thing for most ToB manufacturers.
To B is a "long-termist". Through the integration of technology and business, it can bring long-term value to the industry. It is something that high-quality To B companies must consider when any technological wave sweeps across. In Tong Peize's view, the large financial model is valuable, but more importantly, through years of service experience, he can gain insight into the problems and challenges that corporate customers are facing. On this basis, how the large model can solve these problems is the most pragmatic thinking.
According to Tong Peize, "Currently, we are trying to use general large-scale model training to convert the business demands described in natural language into computer language and into the configuration in the product, so as to build an automated process. Because most of our customers use low-level The code platform configures the product, and configuring the product requires IT personnel and financial personnel to learn and understand the modeling method. The same is true for the large model. In this way, the customer can then import the large model in advance according to the prompt we provide data to generate automated processes.”
During the conversation with Tong Peize, the industrialist learned that Hesi is very cautious about the financial model. "The financial model must be the icing on the cake at present, and it is difficult to do it right away."
It is not difficult to see that the two sides have very similar attitudes towards the financial model. In their view, in the short term, the large financial model cannot penetrate deeper into the industry, let alone combine with the business to empower the business itself. But in the direction of large models, the attitude of both companies is "active research and development", hoping to find and verify the true value of financial large models in enterprise customer scenarios.
Whether it is Kingdee's first financial large-scale model, or Hesi and Fenbeitong's test in this direction, they all send out a signal: in the field of financial management, the value of large-scale models is unquestionable, but the current specific capabilities that can be achieved still need to be developed. Keep exploring.
2. High-quality financial data to speed up the solution
Judging from the large financial model released by Kingdee, what Kingdee has done is a prefabricated prompt project, which is convenient for enterprises to call directly. Behind the prompt project is Kingdee's input of professional knowledge in the financial field, such as tax incentives, into the large model.
Similarly, the reason why Hesi and Fenbeitong have no new actions on the product side of large-scale models is because they have not really verified the value of large-scale models in real financial scenarios.
"We are full of confidence and hope for the application of large models in the direction of financial management, but there are indeed some challenges and limitations. For example, the quality and integrity of data are crucial to the accuracy and reliability of models, and the acquisition and collation of high High-quality financial data may be a challenging task. In addition, the application of large models also requires a large amount of computing resources and labor costs, which may be difficult for some small and medium-sized enterprises.” Wu Rongbin said.
High-quality data is a factor that is repeatedly emphasized in exchanges with Decibel and Hesi . Because, in the training process of large models, as an enterprise service provider, providing high-quality data is the most important step in building a large financial model. In fact, large models in any field require high-quality data as a "base".
But in the financial field, obtaining and organizing high-quality data is a very challenging task. On the one hand, high-quality data needs to be provided by enterprises, while financial data has extremely high requirements for security, and corresponding technical and management measures need to be taken to prevent data leakage and abuse.
On the other hand, in practical applications, financial data may be missing, wrong or inconsistent, which requires data cleaning and sorting. Cleaning data requires certain technical means and investment.
It can be said that when data security and privacy protection have not been resolved, it is difficult to make breakthroughs in large-scale financial models.
Tong Peize told us when referring to the limitations of the large financial model, "From the perspective of business and financial integration, enterprises pay more attention to efficiency. In terms of efficiency, the large model does have a large application space. But on the other hand, a certain vertical Large models in the field need to be learned, trained, and fine-tuned, which depends on industry data. But the problem is that there is not enough public data training, which also makes it difficult for large models to be deeply cultivated and implemented in a certain vertical field.”
In fact, how to obtain high-quality data is not only a problem that financial large models will encounter, but also a challenge that all enterprise-level large models will face.
In the process of building a large enterprise model, data is an extremely important part. It can be said that the quality of data directly determines the effect of enterprise-level large models. Among them, industry data collection and data cleaning are the two most important steps.
Regarding this, a discussion in the industry is that, taking a manufacturer that develops a large financial model as an example, it can choose to cooperate with a manufacturer that collects data. In terms of data desensitization, providers of large financial models can also use federated learning and privacy computing to allow data to be safely applied to large financial models.
Take Hesi as an example. In 2017, Hesi chose to cooperate with an overseas manufacturer to conduct financial audits through AI technology. In the end, in terms of approval efficiency, it was possible to achieve 14% exemption from audit. In this process, the data desensitization technology of this overseas manufacturer was applied.
In addition, another example is the complete isolation of local deployment and public cloud deployment, and local proprietary deployment based on container and micro-service architectures, respectively realizing data internal circulation and data external circulation, and finally achieving the effect of making the financial model smarter.
3. Financial management, stepping into the era of large models
Today, the water temperature of financial digitization is also accelerating.
"The arrival of the fourth phase of the Golden Tax is accelerating the financial digital transformation of enterprises. But in this process, it will definitely be accompanied by the process of handover from the old to the new, that is, one leg has entered the digital age, and the other leg is still in the paper age. Correspondingly Therefore, finance must not only cope with the traditional paper-based operation mode, but also be compatible with the new digital mode." Tong Peize told us.
This kind of problem is exactly what the food brand at the beginning of the article is facing, and it is also the epitome of the digital transformation of most corporate finances.
In this context, the implementation of a large financial model can accelerate the digital transformation of enterprises. In Wu Rongbin's view, "The application of large financial models requires enterprises to carry out digital transformation, including the improvement of digital capabilities in data collection, storage and analysis. This will promote the digital transformation of enterprises in other fields and improve overall operational efficiency and competitiveness."
For the financial model, improving efficiency is only one aspect, and more importantly, it is the subversion it brings to financial management software.
In this regard, Wu Rongbin's thinking is, "The dialogue model can allow enterprises to realize the interaction with App and financial system through man-machine dialogue. Since financial management is very dependent on data-driven, the current BI tools cannot be used on mobile phones. It’s easy to operate, so using natural language to interact and get data and charts is a better way to get data insights.”
From a longer-term perspective, the big model changes the financial management software not only in the front-end UI layer, but also in the back-end application layer. As for this point, SaaS companies like Hesi and Decibel are constantly trying and exploring inside.
In addition, the promotion of the large model for the entire digital transformation is to reduce the cost of understanding and learning.
Tong Peize told us, "Large models can reduce the difficulty of enterprises in adapting business scenarios and products. In fact, the current enterprise management software is very deep, especially the management software in the vertical field. The functions it provides are of great value . However, limited by the cost of understanding and learning, for example, those who understand the scene may not fully understand the product, and those who understand the function of the product may not have a deep understanding of the business demands or business pain points of a certain company."
Therefore, in his observation, when the match between the scene and the product is low, it makes it difficult for the tool to be used in depth, let alone how to integrate it with the business. And reducing the cost of understanding is precisely the core value brought by the large model.
It can also be said that the subversion of financial management software by Shida Model, or the promotion of digital transformation, is essentially to empower the development of "people". In the conversation with Tong Peize, he mentioned that no matter in the past or in the future, the role and meaning of finance will not be subverted, but will only be clarified, and will only return to the original.
This is the value given by the big model to the financial field, and it is also the relationship between technology and people.
In the long run, the industrial value provided by the financial big model includes real-time intelligent cost control, data-driven decision-making, forecasting and optimization, which will also be the endgame of the future financial management field. In this regard, decibels give a more detailed interpretation.
First of all, through the application of large models, enterprises can monitor and manage expenses in real time, so as to better control and optimize the use of resources. The large model can automatically identify abnormal expenses and waste, provide real-time warnings and suggestions, and help enterprises adjust and improve expense management strategies in a timely manner.
Second, the large model can integrate and analyze financial and expense data from various departments to provide a more comprehensive and accurate financial situation and trend analysis. Based on these data, enterprises can make more scientific decisions, optimize the allocation of financial resources, and improve the profitability of enterprises.
Finally, large models can use historical data and algorithms to make predictions, helping companies more accurately predict future financial conditions. Enterprises can make more targeted optimization and adjustments based on these forecast results to achieve more stable and sustainable financial development.
However, before the above-mentioned "endgame" comes, there are still more thorny issues waiting to be solved in the financial field.