Major Tech Giants' Agents Are Collectively Breaking Into the Core of the Financial Sector

The desktop of financial practitioners is becoming a new entrance that major tech giants are competing for. 

On September 3, Tencent officially launched the financial version of WorkBuddy, rolling out more than 80 financial experts and expert teams for institutions including banks, securities firms, and insurance companies. From due diligence for corporate credit and fund research to customer operation for insurance agents, Tencent is trying to make Agents truly integrated into the business processes of financial institutions. 

Less than a month ago, Baidu just named its general-purpose agent GenFlow with the Chinese name "Kuku AI" and launched an independent office client. The financial sector was selected by Baidu as the first key scenario for Kuku AI to expand from general office scenarios to professional office scenarios. 

Earlier, Coze under ByteDance had already introduced financial Agents and Skills from institutions such as Huatai Securities, GF Securities, and Guosen Securities, bringing capabilities including market quotes, financial data, ETF screening, and fund comparison into the Agent platform. 

From ByteDance, Baidu to Tencent, the competition of general-purpose Agents is rapidly sinking into the financial industry. 

It is not difficult to understand why major tech giants choose the financial sector. In businesses such as investment banking, investment research, credit and insurance, a large amount of high-cost professional time is still consumed in checking data, reviewing announcements, verifying statistical calibers, updating models and producing materials. As long as Agents can take over part of these tasks, there is an opportunity to directly convert into efficiency that financial institutions are willing to pay for. 

However, the financial sector is also a tough market to crack. Data accuracy, traceability of conclusions, execution permissions and other aspects must stand the test. A general-purpose Agent that performs well on the open network may not necessarily remain reliable after entering the actual financial business scenarios. 

At present, major tech giants have not yet reached a consensus on whether the financial sector is worth developing a separate Agent version.

Segmented Customer Groups

Although ByteDance, Baidu and Tencent have all extended their Agent layouts to the financial sector, they are currently targeting different markets. 

At present, Coze under ByteDance and Kuku AI of Baidu in financial scenarios are still more inclined to target the C-end market. 

The strategy of Coze is to directly integrate professional financial capabilities into Agents. At present, a series of financial Skills from institutions such as GF Securities and Guosen Securities have been launched on the Coze skill store. For example, the 8 Skills provided by GF Securities cover high-frequency investment research scenarios such as financial comparison, top stock list, ETF screening, ETF capital flow anomaly monitoring, and regular fund investment. 

This means that these actions that originally required to be completed on securities firm Apps, financial terminals or different data pages are split into capabilities that Agents can call directly. 

Users only need to tell Coze what they want to research, and it can call corresponding Skills to inquire about market quotes and financial data, compare companies, screen ETFs or organize fund information. Multiple Skills can be further combined into continuous tasks for pre-market information sorting, intraday monitoring and post-market review. 

Baidu also cuts in from C-end users. Relying on the content, files and storage capabilities accumulated by Baidu Wenku, Baidu Scholar and Baidu Netdisk, Kuku AI is built in with financial data including listed companies, stock quotes, financial reports and research reports. 

After users put forward a research task, Kuku AI can continue to complete information retrieval, data processing and content generation, and finally directly deliver Word research reports, financial analysis PPTs or financial model Excel files. Tasks such as long-term market monitoring can also continue to run in the cloud. 

The financial version of Tencent WorkBuddy targets the B-end market: financial institutions including banks, securities firms and insurance companies. 

In the due diligence of corporate credit, WorkBuddy can generate a list of materials, identify missing items, call data from industrial and commercial, financial and judicial systems to cross-verify risks, and then form a preliminary due diligence draft with retained basis according to the internal template of the bank; in the fund research scenario, it needs to complete fund screening, portfolio diagnosis and other tasks; in the insurance scenario, it also needs to connect customer information, product information, and ensure compliance of marketing content. 

According to Tencent, since March this year, WorkBuddy has successively entered more than 100 financial institutions including CICC, SDIC Securities, Ping An Bank and China Taiping.

However, this application may still be limited to a small scope within financial institutions. According to verification by All-weather Tech with a number of securities firms, office Agents such as WorkBuddy have not been widely used internally at present.

Consensus Not Yet Reached

Major tech giants have not actually reached a consensus on whether the financial sector is worth being made into a separate Agent version. 

At least from the current product form, except Tencent, other major tech giants including ByteDance and Baidu have not launched independent "financial version" Agents for financial scenarios in their Agent layouts. 

An insider of Kuku AI told All-weather Tech that the positioning of Kuku AI is still a general-purpose office Agent, and the financial sector is more regarded as a "sample" demonstrating the ability to handle complex tasks.

The reason is not difficult to understand. The financial sector has strong professionalism and a large amount of data, and a task often needs to go through multiple steps including retrieval, data processing, cross-verification, analysis and final delivery. If an Agent can run through complex tasks like those in the financial sector, it can also more intuitively prove to what extent a set of general-purpose Agents can complete work. 

The above person said that Kuku AI will continue to expand professional knowledge and Skills for more industries in the future, rather than continue to develop in depth only around the financial sector. 

This actually reflects an unsolved problem when office Agents move towards vertical industries: whether general-purpose Agents should continuously expand their capabilities relying on data connections, knowledge bases and Skills, or dig deeper to develop a set of more in-depth products respectively for high-value industries such as finance, law and healthcare. 

The answer has not yet appeared, but the financial sector is indeed one of the industries that most easily arouse manufacturers' impulse to "go one level deeper". 

On the one hand, positions such as investment banking, investment research, asset management and credit have high value per working hour, but a lot of time is spent on processes such as searching for announcements, extracting data, updating models, checking industrial and commercial and judicial information, sorting out meeting minutes and producing materials. 

On the other hand, a large number of financial tasks are originally based on digital materials. Most of the market quotes, financial reports, announcements, research reports, industrial and commercial and judicial information, and even internal research materials of institutions already exist in the form of structured data or electronic documents. 

Perhaps it is precisely seeing these advantages that WorkBuddy directly launched the financial version this time, developing a separate set of products for institutions including banks, securities firms and insurance companies. 

This is also a more radical trial: if financial institutions are willing to pay for a dedicated industry Agent, it means that office Agents have the opportunity to further enter the industry-priced market from selling general productivity tools.

However, the difficulty of the financial industry also lies in that it is a heavily regulated industry.

In June this year, the State Administration of Financial Regulation issued the Guiding Opinions on the Safe Development and Application of Artificial Intelligence in the Banking and Insurance Sectors, listing fund transactions, asset evaluation, credit approval, underwriting and claim settlement, and risk management as high-risk artificial intelligence applications, and requiring the establishment of manual supervision and intervention mechanisms in key links. 

For key decisions that involve customer rights and interests or have substantial financial impacts, the regulator requires setting up manual review nodes and retaining records such as original data and reasoning paths. The document also specifically proposes that agent systems need to prevent risks such as data leakage, unauthorized identity access, tool abuse and operation out of control. 

All these put forward higher requirements for financial Agents. On the one hand, the financial sector may be a market where it is relatively easy to calculate the commercial return of Agent implementation, and on the other hand, it is also the first industry that forces manufacturers to answer questions about data security, model reliability, permission governance and responsibility boundaries.

Wall Street Takes the Lead in Pricing

Overseas tech giants have not reached a consistent answer on whether finance should be a professional scenario in general-purpose Agents or further developed into a separate industry product.

OpenAI is more inclined to the former path, that is, bringing ChatGPT into investment banking scenarios through the Investment Banking plugin, which can complete tasks such as company overview, comparable company analysis, pitchbook and due diligence materials; at the same time, it cooperates with PwC to extend Agents to CFO workflows such as forecasting, reporting and month-end closing. 

Anthropic also builds financial scenario capabilities on the basis of Claude's general-purpose product system.

This year, Anthropic launched 10 sets of financial Agent templates, covering scenarios such as pitchbook, KYC, financial modeling, valuation review and month-end closing, and connected Claude to professional financial data sources such as Excel, PowerPoint, Word, FactSet, S&P Capital IQ and MSCI. 

Differently, Google directly chose to move towards the industry-specific version.

In August this year, Google launched Gemini Enterprise for Financial Services, which is specifically oriented to capital market and corporate banking businesses, built in with more than 50 financial Skills, and connected to professional data sources such as FactSet, Moody’s, MSCI and PitchBook.

Although the industry has not reached a consensus, Wall Street has taken the lead in pricing for this. 

In April 2025, financial AI company Rogo completed a $50 million Series B financing with a valuation of about $350 million; its valuation has risen to about $2 billion this year, nearly 6 times the original level in about one year. 

Behind the continuous increase of valuation by capital, Rogo actually solves the breakpoint between the basic model and financial work.

Downward, Rogo connects to professional databases such as Capital IQ and FactSet; upward, it can generate directly usable Excel models, PPTs and research reports, while retaining data sources and original citations, allowing analysts to go back to financial reports, conference calls and investor materials to verify results. 

At present, Rogo has served more than 300 institutions, covering more than 40,000 financial professionals. 

Overall, there is no standard answer yet to what product form financial Agents should ultimately exist in. It can be a Skill in a general-purpose Agent, a set of financial Agent templates, or a separate industry-specific version like Google and WorkBuddy. 

If financial institutions are ultimately willing to pay extra for deeper data connections, workflow adaptation and security governance, the financial version of Agent may evolve from today's trial product to an independent software category, and more high-value industries will also be re-segmented along this path. 

However, if general-purpose Agents with stronger models, richer Skills and data connectors are sufficient to cover most financial demands, the so-called "industry-specific version" space may also be compressed again.

This article is from the WeChat official account "All-weather Tech" (ID: iawtmt), author: All-weather Tech, published with authorization from 36Kr.