AI, know thyself. That’s kind of what we’ve asked of ChatGPT in this not-so-brief industry brief. OpenAI’s ChatGPT was tasked with writing a piece about the launch of ChatGPT for Financial Services, discuss a previous release from OpenAI’s closest competitor, Anthropic (Claude for Financial Services) and compare them. Can it do so fairly and objectively? We still have our doubts, but we felt that for reasons of personal and public interest, we would run the piece as generated below. Enjoy!
The competition between OpenAI and Anthropic has moved decisively from the general-purpose artificial intelligence market into financial services. On September 10, OpenAI introduced ChatGPT for Financial Services, a specialized offering aimed initially at investment banking and equity research and developed with Morgan Stanley and Evercore as design partners. The release arrives roughly 14 months after Anthropic introduced Claude for Financial Services and after Anthropic spent much of the intervening period adding financial data connections, Microsoft Office integrations, specialized skills and, more recently, ready-to-run financial agents.
That sequence matters. Financial institutions are no longer being asked simply to decide whether employees should have access to ChatGPT or Claude. Increasingly, they are being offered AI environments designed around the actual machinery of finance: proprietary and licensed data, Excel models, research documents, pitchbooks, investment memos, compliance processes and internal information systems.
OpenAI describes ChatGPT for Financial Services as a tailored ChatGPT Work experience combining built-in financial data with the reasoning capabilities of its GPT-6 Astra model. Its initial workflows include valuation analysis, leveraged-buyout modeling, buyer screening, earnings analysis and pitchbook preparation. OpenAI says it built the product around lessons from Morgan Stanley and Evercore, particularly the industry’s need for reliable access to financial data and the ability to turn analysis into high-quality work products.
That makes the launch something more consequential than a finance-themed chatbot. It is another attempt to make the large language model itself a financial-services workspace.
And Anthropic is attempting much the same thing.
What Is ChatGPT for Financial Services?
At its simplest, ChatGPT for Financial Services packages OpenAI’s frontier models, financial information and enterprise controls into a product specifically designed for financial professionals.
One of its most important distinctions from ordinary ChatGPT is data. OpenAI is including premium datasets from providers including Daloopa, PitchBook and LSEG News, covering areas such as company fundamentals, financial statements, earnings transcripts and private-company information. Rather than forcing firms to negotiate separate contracts and construct connectors for every included source, OpenAI says some of this information is indexed and hosted on its own infrastructure. That allows the company to optimize retrieval and latency while providing granular citations that let professionals trace figures and assertions back to their sources.
That last capability may sound mundane compared with increasingly powerful AI reasoning, but it is crucial in finance. An analyst does not merely need an answer to “What was this company’s operating margin?” The analyst needs to know where the number came from, which reporting period it represents, whether adjustments were made and whether the underlying source can be inspected.
OpenAI is also accommodating institutions that already spend heavily on financial information. Existing data subscriptions can be connected through integrations involving providers including FactSet, S&P Global, Preqin and Datasite, while OpenAI has been developing additional relationships involving S&P Capital IQ, MSCI, Dow Jones Factiva and Moody’s. The broader objective is to put both licensed external information and a firm’s internal information within reach of the model.
The result is a potentially important change in the financial analyst’s interface. Instead of moving repeatedly between a data terminal, browser, research database, Excel, PowerPoint and an AI chatbot, the professional can increasingly ask an AI system to retrieve the evidence, conduct analysis and create the finished artifact.
ChatGPT for Financial Services can produce spreadsheets, documents, slides, charts and other outputs from its analysis. OpenAI’s earlier ChatGPT for Excel product similarly allows professionals to build and update models, run scenarios and generate outputs directly within workbooks. Financial-data integrations for FactSet, Dow Jones Factiva, LSEG, Daloopa and S&P Global were already moving ChatGPT toward finance before the September release.
Security and governance are equally important. ChatGPT Enterprise supports role-based access controls, SAML single sign-on, SCIM provisioning, audit logs, DLP and SIEM integrations, encryption at rest and in transit, data-residency controls and regional processing. OpenAI says enterprise customer data is not used to train its models by default.
The significance is clear: OpenAI is trying to eliminate the reasons financial institutions might conclude that a consumer AI assistant is interesting but unusable for institutional work.
Claude Got There First
Anthropic’s strategy is strikingly similar, although the company has had longer to develop its finance-specific product.
Anthropic launched Claude for Financial Services in July 2025 around what it called its Financial Analysis Solution. It combined Claude models, Claude Code, enterprise capabilities and pre-built Model Context Protocol, or MCP, connectors with financial and internal corporate information.
The original ecosystem included FactSet, Morningstar, PitchBook, S&P Global, Daloopa, Databricks and Snowflake, among others. Anthropic positioned the system for due diligence, market research, competitive benchmarking, portfolio analysis, financial modeling, investment memos and pitch decks.
Anthropic then steadily pushed Claude deeper into everyday financial work.
In October 2025 it announced Claude for Excel, additional market-data and portfolio-analysis connectors and pre-built skills for tasks such as discounted-cash-flow models and initiating research coverage. The Excel implementation was particularly telling: Claude could read, analyze, modify and create workbooks while explaining its changes and directing users to the cells behind its conclusions.
Then Anthropic pushed beyond assistants toward agents.
In May 2026 the company announced 10 ready-to-run financial-services agent templates covering labor-intensive work including pitchbook production, KYC screening and month-end close. Those agents can operate as plugins within Claude Cowork and Claude Code or through Claude Managed Agents. Anthropic has also been extending Claude across Microsoft Excel, PowerPoint and Word, with Outlook integration planned.
Its data ecosystem has expanded as well. Claude can work with information from providers including FactSet, S&P Capital IQ, MSCI, PitchBook, Morningstar, Chronograph, LSEG and Daloopa, while newer integrations include Dun & Bradstreet, Fiscal AI, Financial Modeling Prep, Guidepoint, IBISWorld, SS&C Intralinks, Third Bridge and Verisk. Moody’s has also developed an MCP application providing access to its proprietary credit information.
Thus, Claude for Financial Services in September 2026 is considerably broader than the product Anthropic announced in July 2025.
ChatGPT Versus Claude for Finance
Comparisons between ChatGPT and Claude frequently focus on which underlying model is “smarter.” For financial institutions, that is increasingly the wrong question.
The more important comparison is becoming which platform can assemble the best combination of models, proprietary data, internal data, applications, governance, agents and financial workflows.
At launch, ChatGPT for Financial Services has a particularly interesting advantage in its treatment of built-in premium financial data. Instead of requiring every information source to be reached through a connector, OpenAI is hosting and indexing certain datasets itself. The company argues this improves retrieval, latency and citation quality. That approach could reduce integration friction and give users something closer to a finance-native AI product immediately after deployment.
Anthropic, however, has developed a formidable connector ecosystem and has had more time to embed Claude into actual financial workflows. Its strategy has evolved beyond financial analysis into agentic work spanning front, middle and back offices. Its 2026 agent push includes banking, asset management, compliance, accounting and other operational functions.
Anthropic also emphasizes Claude’s ability to work across enormous collections of documents and complex analytical tasks, a characteristic that has made it attractive for due diligence, research and document-intensive finance. OpenAI, meanwhile, has historically built a broader tool ecosystem around ChatGPT, coding, data analysis, research and artifact generation.
But the distinction is narrowing quickly.
Claude now works directly with Excel and other Microsoft applications. ChatGPT has its own Excel and Google Sheets integrations. Claude can execute sophisticated analytical and coding workflows through Claude Code. ChatGPT can conduct research, manipulate data, create models and produce finished documents, spreadsheets and presentations.
In other words, many of the simple “Claude is better at documents, ChatGPT is better at computation” comparisons are becoming obsolete. The products are absorbing one another’s advantages.
The emerging differentiators are more likely to be reliability, data entitlements, integration quality, model performance on specific institutional tasks, governance, latency, cost and the ability to create auditable workflows.
Finance Becomes a Battleground
There is a reason both companies are concentrating so much attention on finance.
Financial services is almost ideally suited to contemporary AI. It is an enormous industry populated by highly compensated knowledge workers who spend tremendous amounts of time searching for information, reading documents, reconciling numbers, writing reports, building spreadsheets and creating presentations.
The industry also generates and purchases huge quantities of structured and unstructured data.
That creates an unusually compelling return-on-investment calculation. If an AI assistant saves a junior investment banker several hours building a comparable-company analysis, or helps an insurance underwriter review a file five times faster, the economic value can quickly become substantial.
Anthropic has already cited evidence of such productivity gains. AIG CEO Peter Zaffino said early deployments incorporating Claude into underwriting compressed review times by more than fivefold while improving data accuracy from 75% to more than 90%.
Citadel says investment professionals are using Claude for Excel to build and update coverage models and pressure-test their work, while FIS has been working with Anthropic on agents intended to compress anti-money-laundering investigations from days to minutes.
OpenAI has its own powerful proof point in Morgan Stanley Wealth Management. More than 98% of advisor teams use the firm’s internal AI assistant, according to OpenAI, while the share of documents effectively accessible to advisors increased from about 20% to 80%. Morgan Stanley also reports that AI-assisted follow-ups that once took days can occur within hours.
These examples demonstrate why the battle is moving beyond experimentation. The important question for executives is becoming less “Can generative AI work in finance?” and more “Which processes should we redesign around it?”
What It Means for Financial AI
ChatGPT for Financial Services and Claude for Financial Services represent another stage in the maturation of financial artificial intelligence.
The first stage involved general-purpose models being used experimentally by individual employees. The second involved enterprise deployments protected by corporate security and governance. The third is now producing industry-specific platforms connected directly to professional data and applications.
The next stage is agentic.
That transition is already visible in Anthropic’s financial agents and in the broader evolution of both companies’ products. An assistant waits for a banker to ask a question. An agent can potentially gather the necessary information, perform the analysis, populate the spreadsheet, draft the memorandum, build the presentation and route the output for human approval.
That changes the economics of AI.
The unit of automation is no longer necessarily a task. It can become a workflow.
It also changes risk. A chatbot that incorrectly summarizes a filing creates one kind of problem. An autonomous system that retrieves the wrong information, modifies a financial model and propagates the mistake into a client presentation creates another. Financial institutions therefore need evaluation systems, permission structures, audit trails, human review and controls proportional to the autonomy they give these systems.
Morgan Stanley’s experience is instructive. Its adoption strategy has relied heavily on evaluations that test AI performance against real-world financial use cases, including daily regression testing. That suggests financial AI governance will increasingly revolve around measurable performance rather than broad assurances that a model is “safe” or “accurate.”
What It Means for Wealthtech and Wealth Management
The implications extend well beyond Wall Street investment banking.
Wealth management may ultimately be one of the biggest beneficiaries because advisory firms combine information-intensive work with expensive human labor and highly individualized service.
The Morgan Stanley example already demonstrates the possibilities. AI can retrieve internal research and policies, summarize meetings, prepare follow-ups and help advisors tailor information to individual clients.
Now imagine those capabilities connected to portfolio data, CRM records, financial-planning systems, tax information, research, client correspondence and custodial platforms.
The advisor’s AI could prepare for a client meeting by identifying major portfolio movements, changes in a household’s circumstances, maturing bonds, concentrated positions, tax-loss opportunities and outstanding planning tasks. Afterward it could summarize the conversation, draft follow-up correspondence, update CRM records and initiate approved workflows.
That does not necessarily eliminate the advisor. It changes what the advisor spends time doing.
For wealthtech companies, however, there is a strategic complication. OpenAI and Anthropic are no longer merely vendors supplying intelligence to financial software. Their platforms are increasingly becoming application layers themselves.
That means established wealthtech vendors must decide whether to compete with them, build on them or become specialized data and workflow providers within their ecosystems.
A financial application whose primary value proposition is summarizing documents or generating generic commentary may face significant pressure. A platform possessing proprietary data, regulated workflows, client relationships, specialized algorithms or deep integrations may instead become more valuable because AI agents need exactly those capabilities.
The likely wealthtech winners therefore will not simply “add AI.” They will expose the proprietary context and actions that make AI useful.
How Financial Professionals Are Responding
The response from institutions suggests enthusiasm tempered by the peculiar caution of a regulated industry.
Corporate finance has already become an important proving ground. CFO.com reported this spring that Claude was appearing in forecasting, reporting, underwriting, private-equity diligence, ERP workflows and audit preparation. Anthropic itself says one member of its corporate finance and strategy team saves 10 to 20 hours per week using Claude in work related to financial analysis and board reporting.
OpenAI is simultaneously building a substantial institutional customer base. Its earlier finance initiatives cited customers and partners including BBVA, Fidelity International, MUFG, Commonwealth Bank, Balyasny Asset Management and Hg. Hg’s head of AI said ChatGPT had accelerated research and due diligence while freeing investment professionals to spend more time on judgment, debate and conviction.
That may be the most useful way to interpret executive sentiment.
Financial institutions are not behaving as though AI is about to replace every banker, analyst or advisor. They are behaving as though employees using AI may increasingly outperform employees who do not.
The concern is therefore shifting from whether to permit generative AI toward how quickly institutions can deploy it without compromising confidentiality, accuracy, compliance or supervisory controls.
Is There Room for Both?
Absolutely—and probably for more than two.
Financial services is too large and heterogeneous for one AI provider to dominate every function. Investment banking, equity research, asset management, insurance underwriting, commercial banking, accounting, compliance, wealth management and retail banking have different data, regulatory and workflow requirements.
Large financial institutions also have powerful reasons to avoid excessive dependence on a single model provider.
A multivendor strategy creates negotiating leverage, reduces operational concentration risk and allows firms to select models based on specific workloads. Claude might perform better on one research workflow while ChatGPT performs better on another. Those advantages may reverse with the next generation of models.
The similarities between their data ecosystems make this especially apparent. FactSet, PitchBook, S&P Global, LSEG and other providers increasingly appear across multiple AI environments. Financial information companies do not necessarily need to choose a single model winner; they can distribute their information into whichever AI systems their customers use.
Financial institutions may eventually behave similarly.
The future could therefore look less like Bloomberg versus Refinitiv and more like the cloud-computing market, where large enterprises deliberately maintain relationships with several major providers.
From Chatbots to Financial Operating Systems
The most important thing about ChatGPT for Financial Services is not that OpenAI has released another version of ChatGPT. Nor is the most important thing about Claude for Financial Services that Anthropic built a particularly capable financial chatbot.
Both companies are trying to occupy a new layer of financial technology.
That layer sits between data and people, capable of understanding natural-language requests, retrieving information, reasoning over it, manipulating software and producing finished work.
If that layer becomes reliable enough, it could change the architecture of financial technology itself.
For decades, financial professionals have learned the interfaces of their software. They learned which database contained which information, which menu generated which report and which spreadsheet had to be updated before another system could proceed.
AI potentially reverses that relationship. The professional describes the desired outcome, and the AI navigates the applications and information necessary to produce it.
ChatGPT for Financial Services and Claude for Financial Services are still early manifestations of that idea. Neither removes the need for verification, human judgment, fiduciary responsibility, compliance oversight or financial expertise. Indeed, greater AI autonomy makes those things more important.
But the direction of travel is becoming unmistakable.
The competition between OpenAI and Anthropic in finance is no longer primarily a contest to build the chatbot with the best answers. It is a contest to become the intelligence layer through which financial work gets done.
For bankers, analysts, advisors, executives and wealthtech companies, that distinction is enormous. The winner—or, more likely, winners—will not simply sell financial institutions another piece of software.
They may help determine what financial work itself looks like.
Researched by DWN Staff
