Written by: Web4 Research Center

Anthropic, the company that has swept the developer community with AI programming tools in recent years, announced on May 5th, U.S. time, the launch of 10 AI Agents specifically designed for financial services, officially charging into Wall Street.

According to a report by Sina Finance, the task list of these 10 tools almost covers the core areas of daily work for financial professionals: drafting client meeting materials, reviewing financial statements, and escalating cases for compliance review. Target users span professionals in banking, insurance, asset management, and financial technology. This is not a chatbot nor a simple Q&A tool. This is a group of digital employees that can be directly embedded into financial institutions' workflows to perform specific tasks.

The market's sharp fluctuations convey signals that are far more complex than news headlines. The actions of investors voting with their feet have exposed a consensus deep within the industry: as AI Agents begin to take over the previously irreplaceable work of financial professionals, the whole value chain of financial services may be on the brink of a turning point.

1. From Writing Code to Writing Reports: The Same Logical Business Path

Anthropic's entry into the financial field mirrors its conquest of the programming market. More precisely, this is almost a replay of the same script across different industries. Before the launch of the financial AI Agents, Anthropic had already established a dominant position in the programming tools market. According to a research report published by Zhejiang Merchants Securities in April 2026, Claude Code, a subsidiary of Anthropic, has captured 54% of the enterprise-level coding Agent market. As of February 2026, 4% of public submissions on GitHub were written by Claude Code, and analysts predict this ratio will exceed 20% by the end of 2026. In terms of enterprise-level large language model spending, Anthropic holds a 40% market share, with 80% of the world's top ten wealthiest companies as its paying clients.

Relevant data shows that Anthropic's overall share of the U.S. AI market has skyrocketed to nearly 70%, while the previous 90% share of ChatGPT has been significantly eroded. It took Anthropic less than a year to go from being a follower to surpassing market shares. The logic behind the disruption of the programming market isn't complex: AI Agents do not help programmers type more efficiently, but directly generate code, debug code, and deploy code, compressing development tasks that once took days into mere hours. According to a survey conducted by Economic Observer from October 2025 to January 2026 of 201 financial service providers in mainland China and Hong Kong, 81% of financial enterprises have incorporated AI into their workflows, yet the pain points remain pronounced—insufficient talent reserves, aging systems, and lagging regulations. These pain points are precisely the leverage points that AI Agents can exploit.

However, there is a subtle point worth noting here. Nicholas Lin, head of financial services products at Anthropic, made a remark that seems understated but carries deep meaning. According to a report by Tencent News, he stated that AI applications in the financial sector are "only a few months behind programming applications," the latter of which have already seen significant acceleration. A few months of time difference. Not years, not a generational technology cycle, but just a few months. This judgment hides a deep logic—if the demand structure for AI Agents in finance is essentially the same as that in programming, then the dominoes of an already disrupted programming market falling in finance is just a matter of time.

From the perspective of specific work scenarios, these 10 Agents are allocated to two types of tasks: five for financial research and client coverage, and five for finance and operations. In research and client service scenarios, Claude Agents can establish target lists, perform comparable company analyses, draft presentation materials for meetings, and organize background summaries of clients and counterparties before calls. In finance and operations scenarios, they can check if valuations align with similar company metrics, execute closing checklists, prepare accounting entries, and produce closing reports. Reports from TechOrange reveal more details: Claude can now operate directly through plugins in Excel, PowerPoint, Word, and Outlook, which means financial analysts no longer need to leave their daily work software, as AI Agents are already embedded within.

However, when AI Agents are embedded deeply enough, a more fundamental question arises: if these Agents do not just draft memos but are beginning to make financial decisions on behalf of institutions or clients, where can their "hands" reach?

2. The Same Battlefield, Two Offensive Routes

Anthropic is not the only "knocker" on Wall Street. Almost concurrently, OpenAI also launched its financial offensive. According to reports from Bloomberg Law on May 5, 2026, OpenAI announced a collaboration with PwC to develop AI Agents for CFO teams, covering core processes such as planning, forecasting, reporting, procurement, payments, finance, taxes, and settlements. More intriguingly, OpenAI has positioned its finance team as "Client Zero"—testing a procurement agent tool in its own financial operations before replicating the experience for enterprise clients.

Going back to earlier, on March 6, 2026, Zhitong Finance reported that OpenAI released the GPT-5.4 model along with a set of financial service tools that can connect to financial data sources like FactSet and Third Bridge and create and check financial models directly in Excel and Google Sheets. On April 14, Wedbush released a research report revealing that OpenAI officially acquired the startup Hiro Finance, specializing in autonomous personal finance.

The differences in the paths of both firms are becoming increasingly clear. Anthropic has chosen a bottom-up approach: starting from the analysts' workbenches and addressing those repetitive tasks that consume a lot of human resources daily, gradually penetrating into the operational systems of financial institutions. OpenAI, on the other hand, is leveraging consulting giants like PwC to push from the top down, focusing on the key control processes surrounding financial management. One path addresses "efficiency gaps," while the other targets "control heights."

This speed is worth pondering. This is not a gradual infiltration over several years but a market encirclement completed within months. As the largest financial institutions begin to define AI Agents as "digital colleagues" rather than "efficiency tools," the shift in terminology reflects a deeper identity confirmation—these Agents are transitioning from "auxiliary tools" to "semi-autonomous participants."

From augmentation to participation, each step seems smooth. However, moving from participation to autonomy requires entirely different infrastructure. An Agent that screens comparable companies for analysts, and an Agent that holds funds for clients and executes payments face almost two different species of technological challenges.

3. The Stakes Behind the Stakes: Where Money Flows

The market responds to the arrival of AI Agents with falling stock prices, while another market expresses its belief in a more primitive way: money. The timeline does not need to be drawn too long. In February 2026, Anthropic completed a $30 billion financing at a valuation of $380 billion. Just two months later, Bloomberg and CNBC reported on April 29, 2026, that Anthropic is in talks for a new round of financing of about $50 billion, with a valuation target of $900 billion. If this is successfully realized, this figure will surpass OpenAI's valuation of $852 billion set at the end of March, making Anthropic the highest-valued AI startup in the world.

In two months, the valuation jumped from $380 billion to $900 billion. Such a large magnitude is rare in all of business history. But what is more noteworthy is the direct catalyst driving this round of financing—Anthropic's Claude Mythos Preview model released in April. This cutting-edge model with advanced cybersecurity capabilities is only available for restricted access to about 50 institutions like Apple and Microsoft, and it quickly sparked high-profile meetings in Washington and Wall Street. Merely a preview version has propelled a leap in valuation of hundreds of billions, fundamentally changing the market's pricing logic for "trustworthy vertical industry AI."

The presentation of capital bets is not limited to valuations. According to IT Home reports on April 30, 2026, Anthropic's annual recurring revenue has reached $30 billion, whereas a year ago, its annual revenue was about $10 billion. The growth curve is almost vertical.

Meanwhile, on May 5, Anthropic CEO Dario Amodei discussed AI alongside JPMorgan CEO Jamie Dimon at a corporate event in New York, with high-level executives from Wall Street banks in the audience. What thoughts the Silicon Valley founder, standing in the spotlight of Wall Street's power center, might evoke in the bankers below is perhaps not hard to guess. When asked about the surge in AI infrastructure spending, Dimon responded, "Overall, it makes sense. If you try to pick winners and losers, that would be difficult." This seemingly bland comment precisely reflects the anxiety of the entire industry—not that they don't want to pick sides, but rather they dare not gamble recklessly on the wrong one.

However, a recurring question remains unresolved: if AI Agents are no longer just "digital colleagues" but need to directly hold assets, authorize expenditures, and sign contracts, how far can the existing financial infrastructure support them?

4. When Agents Are Not Just "Report Writing Assistants"

This question is not science fiction; it is already at the door. In May 2026, Odin Group officially launched the OwlPay Agent Wallet, a digital wallet designed specifically for AI Agents. According to a report by China Times on May 5, this is not a traditional digital wallet—it is constructed for AI rather than humans. After user authorization, the AI can send, receive, and manage stablecoins without direct operation. The wallet adopts a self-custody architecture, allowing users full control of private keys and funds, with all credentials generated and stored on local devices, supporting mainstream blockchains like Ethereum, Stellar, and Solana.

On the same day, GlobeNewswire also released related reports. Odin Group stated that the wallet utilizes the company’s payment licenses held in 40 states in the US, extending regulated stablecoin access to the AI Agent economy. This is not a concept validation. It is a formally launched product with regulatory qualification in 40 states.

So the question arises: why does a wallet designed specifically for AI Agents need to utilize stablecoins and blockchain? Can AI Agents use credit cards? Of course they can. As one analyst noted in an article published in late April: if an AI Agent only helps users book a flight, reserve a hotel, or renew a SaaS service, it can fully invoke existing payment systems like Swift, credit cards, or virtual cards without any essential barriers. But the real issues arise in other scenarios, where an AI research Agent may need to continuously call multiple databases, purchase several paid materials, access different model APIs, pay for chart generation tools, or even buy an analysis from another Agent to complete an industry report. In this series of operations, there might not be a single traditional store entry or a standard checkout page. Agents face a series of APIs, data interfaces, model services, and computing nodes.

When the transaction entities become machines, the traditional financial system finds a missing piece at its base. From a broader perspective, this is not an isolated business observation. AI Agents are evolving from auxiliary tools into true economic participants at a pace that far exceeds other infrastructures. Although Agents can currently execute tasks and transactions, they still lack a standard way to verify "who am I," "what am I authorized to do," and "how to receive payment" when operating across environments. "Identity is non-transferable, payments are not yet universally programmable, and collaboration remains isolated." Blockchain, as a public ledger, a portable wallet, and a programmable settlement layer, is being seen by some technical teams as a crucial infrastructure to fill these gaps.

This is not a one-sided wish for a blockchain narrative. As reported in PwC's early 2026 report, financial institutions are gradually positioning AI as a "strategic transformation engine" rather than merely a tool for efficiency enhancement. When Agents evolve from "doing tasks for you" to "managing assets for you," a "verifiable execution record" becomes a survival threshold rather than just an added value—on-chain records are not meant to replace traditional audits but to provide a trustworthy trail at the Agent level that human audits cannot cover in real time. This implies that in the future financial ecosystem, Agents may require both the compliance channels of traditional finance and the auditable on-chain identity and payment infrastructure, coexisting on two tracks.

However, it must be frankly acknowledged that while the OwlPay Agent Wallet has obtained payment licenses in 40 states, its overall adoption remains in the early stages; the x402 protocol and various Agent identity proposals are still in the standard discussion phase; and the concept of "Know Your Agent" (KYA), while gaining attention, is still a distance away from widespread implementation. This is not a story that has already run smoothly, but rather one that is stumbling forward. Its value lies not in proving any undeniable conclusions but in exposing a real existing problem: within the closed loop of the traditional financial system, machines remain tools, not subjects. Today, they are learning to do more.

5. Reinterpreting the Anchor of Value

This may sound like a depiction of AI replacing human jobs. But if we pause to think, the real change might occur on another dimension. The core value of traditional financial information services is built on a form of information asymmetry. The value of FactSet and Morningstar lies not only in their data but also in how they organize that data into formats that are callable, comparable, and modelable by professional users. This layer of "organizational cost" constitutes their moat. The logic of AI Agents, however, is completely different. They are not organizing data but executing processes—they are operators, not databases.

This distinction is critical. FactSet's stock price fell by 8.1% at one point after the announcement, while Morningstar dropped by over 3%, according to data from Sina Finance reported by Dongfang Wealth. However, the reason for the decline is not merely that "AI can replace human analysts"—it reflects a market repricing: when AI systems can directly connect to FactSet and Morningstar's data sources for real-time analysis, data services themselves shift from being endpoints to becoming raw materials. The pricing of raw materials is always below that of finished services.

This also explains why Anthropic announced plans to establish a $1.5 billion joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs to accelerate the integration of Claude's AI capabilities into more enterprise scenarios at the same time it launched the financial AI Agents. During the same period, Claude became able to directly connect with market data platforms like FactSet, S&P Capital IQ, MSCI, PitchBook, and Morningstar, and deeply integrate credit rating and corporate data sources with Dun & Bradstreet and Moody's—over 600 million pieces of information about public and private enterprises can flow through Claude's pipeline. The deeper meaning of this layout is that Claude is not here to compete with data providers but to redefine the decision-execution layer above them.

However, when the information processing and decision-making stages are compressed and merged into a continuous automated process, the human roles along the entire chain face redefinition. From briefings to compliance upgrades, these 10 AI Agents have effectively inserted themselves into three previously considered the most "irreplaceable" segments: information organization, professional judgment, and risk assessment. Each segment is being partially disintegrated. Analysts no longer monopolize the work of information organization, compliance teams no longer monopolize the initial work of risk screening, and investment banking VPs no longer monopolize the drafting of presentation materials.

This does not imply that "humans" will be completely replaced. It certainly signifies that human roles are shifting from operators within the process to designers and supervisors beside the process. This transformation is not merely a simple "job loss anxiety" issue. It resembles a river changing its course—the volume of water hasn’t decreased, but the riverbed has changed, the original docks may be abandoned, while new ones are rapidly constructed downstream.

This evokes a classic metaphor in the philosophy of technology. Heidegger, when discussing technology, was never concerned with the tool itself but rather with technology as a "scaffolding" that reorganizes our relationship with the world, changing how we view things, others, and even ourselves. The process of AI Agents embedding into financial workflows is indeed weaving a new scaffold. This scaffold not only processes data and writes reports but also redefines what the core value of financial work is.

6. Not the End, but a Watershed

The stock prices of FactSet, Morningstar, S&P Global, and Moody's have taken a hitting, and this market reaction itself carries meaning. According to reports cited by Dongfang Wealth, FactSet once fell by

8.1%, and Morningstar by over 3%. On Wall Street, such numbers mean that the market is betting real money on a judgment—traditional financial information service providers' competitive barriers are indeed more fragile than people imagine in the face of AI Agents. However, this "fragility" does not imply that traditional institutions will disappear immediately. More likely to happen is a different evolutionary path—a restructuring of the value chain. FactSet and Morningstar have irreplaceable data assets that are the fuel upon which AI Agents rely. The issue is that when the fuel itself is no longer scarce, what becomes scarce is the engine that can accurately inject that fuel into the combustion chamber. The manufacturers of engines are taking a larger share of the value chain.

A noteworthy detail is that according to an analysis in the April 2026 report published by Zhejiang Merchants Securities, one of the keys to Anthropic's success lies in its focus on an auditable rule framework. Compared to competitors like OpenAI and Google, Anthropic emphasizes traceable reasoning processes and transparent compliance systems, making it naturally adaptable to high-regulation industries such as finance, law, and government affairs. In a financial field where trust is already the core currency, the barriers created by AI firms' positioning on safety and compliance may prove to be more lasting than their model capabilities. It is not about who is smarter, but who is more trustworthy. The latter carries far more weight on Wall Street than the former.

AI Agents are evolving from tools that write code to behaving subjects that enter real economic circulation. When they begin to take on subject roles in the economic system rather than merely being tools, the grammar of economic infrastructure is rewritten. Concepts like payment, identity, responsibility, and auditing—pillars of modern finance—are facing redefinition when confronted with an "invisible participant." This redefinition occurs internally within traditional finance but also spills over into traditional frameworks, giving rise to new explorations of infrastructure.

Wall Street's financial AI Agents are only the beginning. As Goldman Sachs and JPMorgan roll out Agents into their core workflows, as FactSet and Morningstar are compelled to redefine their value propositions, and as scattered projects like OwlPay create dedicated wallets for Agents—these seemingly isolated events are in fact stitching together a larger picture: Agents are no longer just "doing tasks for people"; they are beginning to participate in value distribution.

In conclusion: Agents have entered the game, and the rules are just beginning to be written.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。