Half the apps an advice firm pays for now do work Claude does in one window. Anthropic has been shipping into financial services on a roughly monthly cadence since mid-2025. Here is what survives the category collapse.
Half the apps an advice firm pays for now do work Claude does in one window. AI financial advice is collapsing into a single interface, with everything else commoditised underneath.
That is the shift the financial services industry is processing this quarter. It is moving faster than most software vendors planned for, and the firms I am talking with have noticed.
This piece is the field guide.
The dates are not marketing fluff. Claude for Financial Services launched in July 2025 with native integrations to Morningstar, S&P Global, FactSet, Databricks, and Snowflake. Claude Cowork hit research preview in January 2026, Windows parity on 10 February, and went generally available on 9 April. On 24 February, Anthropic shipped five dedicated finance plugins covering financial analysis, investment banking, equity research, private equity, and wealth management. Last month the company added an Excel add-in, real-time data connectors, and pre-built Agent Skills including cash flow models and initiating coverage reports. Tomorrow, 5 May, Anthropic livestreams The Briefing: Financial Services for executives leading AI transformation at the world's largest institutions.
This is a tech company shipping a category-collapsing product on a roughly monthly cadence. Not into one workflow. Into the entire industry layer.
The integrations list reads like a who's-who of the regulated data stack: Morningstar, S&P Global, FactSet, Databricks, Snowflake, Box, Daloopa, Palantir, PitchBook, MSCI, LSEG. Accenture is the deployment partner of record. The Greek banking sector has already been built on Claude through the Piraeus Bank partnership. Anthropic is not selling a chatbot. It is selling the operating layer that previously sat across a dozen separate vendors.
The X (Twitter) conversation is all over it. Chris Lee, a builder with a large following, posted a screenshot of a private dashboard streaming ten years of his transactions and stock holdings into Claude with the line "Claude is now my wealth advisor." Rohan Paul on the Cowork update: "Claude can act as a specialized financial assistant that can jump between apps like Excel and PowerPoint while keeping all your data in sync." Matt Shumer ships claude-investor, an open-source investment analyst agent. @AItechscarlett: "AI can now build financial models like Goldman Sachs analysts. 12 Claude prompts that replace $150K a year of investment banking work." Carl Vellotti documents a household budget run end-to-end through Claude Code, SimpleFIN, and Actual Budget for ten dollars a year.
These are not theoretical demos. They are the same workflow that used to require five subscriptions, now in one window. The speed gap is the clue. Adoption on social is moving in days. Procurement in advice firms is moving in quarters. The decision is being made by individual users on individual desks, well ahead of any official sign-off.
For a decade, knowledge work software sold a moat made of category lines. A note-taker for capture. A drafting tool for documents. A research platform for product comparison. A CRM for relationship history. A document store for compliance. A signature tool for execution. Each tool charged a monthly fee. Each promised the integration with the others would be solved soon.
Cowork dissolves the category lines. The model becomes the integration. The user does not switch tools; the model does. The artefacts of the work, file note through SoA, sit in shared files the model reads and writes natively, with the underlying apps reduced to commodities or removed entirely.
A piece on Medium last month put the consequence bluntly: "How Claude Just Murdered the Old Software Industry." The number cited was $830 billion of SaaS market cap exposed to category collapse. Pick your own discount on that figure. The direction is not in dispute.
Legal tech is already living it. Cowork reads contracts against the firm's own terms templates, flags deviations, explains the legal implications of non-standard clauses, and drafts redline suggestions. Half of contract review software has been rendered into a feature, not a category. The same is happening to consulting research, equity research note drafting, due diligence summarisation, audit prep workpapers, and a long list of categories that look defensible until you watch a single Cowork session do all of them in an afternoon.
Financial advice is the next industry to walk through that door. The only debate is whether it takes six months or eighteen.
This pattern has played out three times already. Bookkeeping software did not die in the cloud transition because it owned the audit trail. The CRM did not die in the mobile transition because it owned the relationship system of record. Compliance tooling did not die in the first AI summarisation wave because it owned the obligation map.
The pattern is consistent. Software that owns a regulated layer survives. Software that produces a draft, a summary, or a formatted document does not.
What is different this time is the speed. The cloud transition gave on-prem vendors years to react. The mobile transition gave desktop vendors a similar buffer. The Cowork transition is moving in weeks. The reason is the same reason a household can run a budget through Claude Code and an MCP server for ten dollars: the integration cost has gone to zero. Once a model can drive the apps directly, the apps stop selling the integration; they sell the underlying data, the regulated layer, or nothing. The middle category dies first, and the middle category is most of the market.
Apply that to firms running Claude in their advice workflows in 2026. The regulated layer is the trail of evidence ASIC asks for at audit. The version of the SoA that was approved. The reviewer who signed off. The source data the model used. The change log. The cross-border transfer record. The retention period for client data. The proof, on demand, that this output came from this prompt run against this version of the model under this human's accountability.
Cowork drafts. Cowork drafts beautifully. It does not, by itself, prove that what was drafted is what was sent, who reviewed it, or where the data lived during the work. That gap is not a feature Anthropic is going to ship. It cannot, without becoming the licensee. The proof layer is structurally outside what a model company will own.
The product that owns that proof lives. The product that does not own it becomes a feature inside Cowork by Christmas.
If you run a financial advice firm and want to spend the next 30 minutes well, do this.
Open a spreadsheet. List every monthly subscription the firm pays for that touches advice work. For each, write three things: what artefact it produces, what evidence layer it owns, and what would break if you cancelled it tomorrow.
Most firms find five or six rows where the answer to "what evidence layer it owns" is "none, the adviser still runs the audit trail in their inbox." Those rows should not be renewed in 2026.
The remaining rows usually fall into one of three buckets. The CRM that holds the client record. The compliance tool that holds the approval trail. The advice operating system that holds the SoA evidence chain end-to-end. Anything outside those three buckets is at risk of being a feature inside someone else's window by mid-year.
This is the part where I should be honest about where AdviseWell sits. We chose the third bucket on purpose. Drafting tools are interesting. They are not what survives the next eighteen months as a standalone product. The bridge between Claude Cowork and the regulated record is the part that does not commoditise. It is also the part that has to be sovereign, audit-trailed, and ASIC-aligned. We built that, deliberately, on the read of the market that this piece is about. If we are wrong, we will have built a very expensive proof layer for nothing. If we are right, the firms that move now arrive at the new desktop with a record that survives audit, while their competitors arrive with eight cancelled subscriptions and an inbox full of drafts.
The honest answer to "will my software vendor still be here in 18 months" is that some will. The ones that will are the ones that own the layer Anthropic cannot reach without a permissioned bridge. Pick those. Cancel the rest. Do it before the renewal cycle picks for you.
The associates at the firms I am talking with have already made the call. They are using Claude for half of what eight tools used to do. The principals are six months downstream of that, reading the announcements and squinting at the renewal calendar.
The 18-month frame matters because that is one budget cycle. By the next one, half the line items will read differently. The firms that lead this cycle will have rewritten their stack on purpose. The firms that follow will have it rewritten for them by Anthropic's release notes.
In a year, the desktop has fewer windows. Many fewer. The question is whose name is on the one that is left, and what proof layer is sitting behind it.
See how advice firms are boosting capacity without compromising compliance or trust. Book a demo to see AdviseWell in action.
.png)
