ByteDance, Baidu, and Tencent's AI Agents Collectively Move into Financial Services
nashnova research
China's three biggest tech firms moved AI agents into financial services within a single month, but they picked opposite paths — ByteDance and Baidu target retail investors while Tencent sells directly to banks and brokerages — and the outcome will decide whether office agents can evolve from generic tools into industry-priced software.
What did each company launch?
Tencent released WorkBuddy Finance Edition on September 3, offering over 80 specialist financial agents to banks, brokerages, and insurers — the first standalone, finance-specific agent product from a major Chinese tech firm.
Baidu gave its general-purpose agent GenFlow the Chinese name "Kuku AI," launched a dedicated desktop client, and designated finance as the first vertical for its push from general to professional office work.
ByteDance's Coze platform has already integrated financial agents and skills from Huatai Securities, GF Securities, and Guosen Securities, bundling market data, financial statements, ETF screening, and fund comparison into its agent ecosystem.
Are they targeting the same customers?
No — the split is clear. ByteDance's Coze and Baidu's Kuku AI primarily serve retail investors (C-side); Tencent's WorkBuddy Finance Edition targets institutional clients (B-side).
Coze breaks professional finance into callable skills — GF Securities alone contributes 8 skills covering financial comparison, top-trader rankings, ETF screening, and more. Users can chain multiple skills for pre-market prep, intraday monitoring, and post-close review.
Kuku AI draws on Baidu's document library, academic database, and cloud storage. After receiving a research task it can deliver a Word report, a financial-analysis PPT, or an Excel model, and long-running market-monitoring tasks keep running in the cloud.
Tencent's WorkBuddy Finance Edition goes deeper into institutional workflows — credit due-diligence (cross-checking a borrower's business registration, financials, and legal records to draft an initial report), fund screening and portfolio diagnostics, and insurance compliance review. Tencent says more than 100 financial institutions — including CICC, SDIC Securities, Ping An Bank, and China Taiping — have begun onboarding since March.
Should agents be general-purpose or industry-specific?
This is the core question the big tech firms have not agreed on. This means → companies on the same playing field have reached opposite conclusions about the right path.
A Baidu insider said Kuku AI's positioning remains a general-purpose office agent; finance is treated as a "showcase" for complex task capabilities, and the product will expand to other industries rather than drill deeper into finance alone.
Tencent took the more aggressive step: shipping a standalone finance edition. In plain terms = Tencent is betting that financial institutions will pay a premium for a purpose-built industry agent — and if the bet pays off, office agents can shift from selling generic productivity tools to industry-priced software.
What are overseas tech giants doing?
There is no consensus abroad either. OpenAI leans toward the general-purpose route, using an Investment Banking plugin to bring ChatGPT into deal workflows while partnering with PwC to extend agents into CFO operations.
Anthropic released 10 financial agent templates covering pitchbooks, KYC (know-your-customer verification), financial modeling, valuation review, and month-end close, and connected Claude to FactSet, S&P Capital IQ, MSCI, and office tools like Excel and PowerPoint.
Google went furthest — in August it launched Gemini Enterprise for Financial Services, purpose-built for capital markets and corporate banking, with over 50 built-in financial skills and connections to FactSet, Moody's, MSCI, and PitchBook. This reflects the same general-vs-vertical debate playing out globally.
What do regulators and investors think?
The regulatory bar is rising. In June, China's National Financial Regulatory Administration issued guidelines classifying fund transactions, credit approvals, underwriting claims, and risk management as high-risk AI applications, requiring human review checkpoints for key decisions and full retention of raw data and reasoning trails.
This means → a financial agent cannot simply be built and deployed — every decision must be traceable and human-supervised, and the compliance cost is significant.
Capital markets have already priced the opportunity. Financial-AI firm Rogo was valued at roughly $350 million at its Series B in April 2024; that figure has since risen to approximately $2 billion — nearly a six-fold increase in about a year. Rogo's core value is bridging the gap between foundation models and real financial work: it connects downstream to Capital IQ and FactSet, generates ready-to-use Excel models, PPTs, and research reports upstream, and preserves source citations throughout. It currently serves over 300 institutions and more than 40,000 finance professionals.
How will this play out?
There is no standard answer yet. If financial institutions are willing to pay for deeper data connections, workflow integration, and security governance, industry-specific agents could evolve from pilot products into a standalone software category.
Conversely, if general-purpose agents with stronger models and richer skills can cover most financial needs, the space for "industry editions" will shrink again.
In plain terms = Tencent's aggressive bet is one of the earliest pressure tests for this strategic debate — its success or failure will directly shape how fast rivals follow.
市场有风险，内容仅供研究参考，不构成投资建议。