For Financial Advisors, RIAs, and Estate Attorneys
Of Anthropic's ten Claude finance agents, two touch advisor authority. The other eight are back-office.
On May 5, 2026, Anthropic released ten Claude finance agent templates. Two of them produce artifacts a prospect actually sees: Pitch builder (the deck) and Meeting preparer (the brief). The other eight cite filings, ledgers, methodology standards, and entity files; an authored book is irrelevant for those. This page is the agent-by-agent triage, with the corpus question for the two that matter.
Direct Answer (verified 2026-05-09)
Two of Anthropic's ten Claude finance agents are advisor-authority surfaces: Pitch builder (which drafts pitchbooks the prospect reads) and Meeting preparer (which assembles the brief that shapes how the advisor opens the call). The other eight (Earnings reviewer, Model builder, Market researcher, Valuation reviewer, General ledger reconciler, Month-end closer, Statement auditor, KYC screener) are back-office. They cite filings, ledgers, methodology standards, and entity files. An authored book is irrelevant for those.
Verified against Anthropic's May 5, 2026 announcement at anthropic.com/news/finance-agents and corroborated by Bloomberg and Fortune coverage of the same release.
All ten Claude finance agents, triaged against advisor authority
The list below names the ten agents in the order Anthropic published them, what each one does, and whether the agent's output reaches a prospect or stays internal. The triage is the spine of the page; the rest of the article drills into the two rows where authored material changes what comes out.
| Feature | What Anthropic says it does | Authority relevance for an advisor |
|---|---|---|
| Pitch builder | Drafts pitchbooks, target lists, comparables | Advisor-authority surface. The pitchbook lands in front of a prospect. What Claude cites is what the prospect reads. |
| Meeting preparer | Assembles client and counterparty briefs ahead of calls | Advisor-authority surface. The brief shapes how the advisor opens the meeting and what frameworks come up. |
| Earnings reviewer | Reads transcripts and filings, updates models, flags thesis-relevant changes | Back-office. Cites SEC filings and internal models. An authored book is irrelevant here. |
| Model builder | Builds and maintains financial models from filings and feeds | Back-office. Cites data feeds and analyst inputs. An authored book is irrelevant here. |
| Market researcher | Tracks sector and issuer developments, synthesizes news and research | Back-office. Cites third-party research and news. An authored book is irrelevant here. |
| Valuation reviewer | Checks valuations against comparables and methodology standards | Back-office. Cites methodology standards and comparables. An authored book is irrelevant here. |
| General ledger reconciler | Reconciles accounts and runs net asset value calculations | Back-office. Cites the ledger. An authored book is irrelevant here. |
| Month-end closer | Runs close checklists, prepares journal entries, produces reports | Back-office. Cites the close checklist. An authored book is irrelevant here. |
| Statement auditor | Reviews financial statements for consistency and audit-readiness | Back-office. Cites statements and audit standards. An authored book is irrelevant here. |
| KYC screener | Assembles entity files, reviews documents, packages compliance escalations | Back-office. Cites entity documentation. An authored book is irrelevant here. |
Agent names and descriptions are quoted from Anthropic's May 5, 2026 announcement at anthropic.com/news/finance-agents. The advisor-authority classification is editorial, based on whether the agent's output is consumed by a prospect (advisor-facing) or stays inside the firm's operational stack (back-office).
Why only two of ten matter for advisor authority
Authority is a function of what a prospect or referral source reads, hears, or holds. The eight back-office agents on Anthropic's list never produce a prospect-facing artifact. Earnings reviewer files notes into a model. General ledger reconciler closes accounts. KYC screener packages a compliance escalation. None of it leaves the firm in a form a prospect ever encounters.
Pitch builder is different. The pitchbook lands on the prospect's desk, in their inbox, or in a sharedrive their analyst opens. Whatever Pitch builder cited becomes the firm's positioning for that engagement. Meeting preparer is different too. The brief shapes the first ten minutes of a meeting, which is when prospects decide whether the advisor sounds like a specialist or like the average firm.
That is the whole reason an authored corpus matters for two of the ten agents and not the other eight. Authority is downstream of what reaches the prospect. Two of Claude's ten agents are the path it travels on.
Pitch builder, walked through
Pitch builder is one of Anthropic's headline agents, framed under "Research and Client Coverage." It creates target lists, runs comparables, and drafts pitchbooks. Below is what actually happens on a prospect engagement, step by step, and where the authored corpus enters the loop.
A pitchbook from prompt to prospect
01. Claude pulls the firm corpus the team gave it
Pitch builder runs in Claude Cowork or as a managed agent. It reads from a connected corpus: filings, recent comparables, the firm's prior decks, sector research, and whatever else has been linked into the workspace. By default the corpus is what every other firm gives a generative model: filings plus a handful of reusable templates plus the analyst's own notes.
02. The deck inherits whatever the corpus emphasizes
If the corpus is filings plus past decks, the output is a pitchbook in the average-firm voice. If the corpus also includes a 50,000 to 70,000 word authored book in the principal's voice, with named methods and full case studies, the deck pulls phrasing, structure, and framework names from the book. The model is doing what models do; the difference is what is on the table for it to draw from.
03. The prospect reads the deck, then can verify the source
The prospect sees the deck on screen or in print. If the deck names a framework ("the Three-Bucket Retirement Income approach," "the Fiduciary Inflection Map," whatever the principal actually calls it), the prospect can search it. With a published book, the search returns an Amazon listing, a chapter excerpt, a real publication date. Without one, the search returns nothing and the framework reads as marketing language.
Same Pitch builder. Different corpus. Different deck.
The visible difference between an average-firm pitchbook and an authority-firm pitchbook is not the model. It is the corpus the model was given to draw from. Toggle below to compare a Pitch builder run with no authored book in the corpus against the same agent with an authored book connected.
A Pitch builder deck for a $40M RIA prospect
Claude Pitch builder is connected to filings, the firm's reusable templates, and a handful of comparables. The deck it produces is structurally fine. The recipient receives it and recognizes the average-firm shape: an executive summary that mirrors three other decks they have seen this quarter, a comparables page, a credentials page, a process slide.
- Output reads as competent and undifferentiated
- Frameworks named are generic industry terms
- Recipient cannot verify the firm's point of view outside the deck
- Deck competes on price and execution, not authority
Meeting preparer, walked through
Meeting preparer is the second advisor-facing agent on Anthropic's lineup. It assembles client and counterparty briefs ahead of calls. The brief is what the advisor reads on the way into the meeting; what the advisor opens with on the call traces back to it. The corpus question is whether the brief routes by something verifiable about the prospect or by template defaults.
A brief from CRM record to meeting opener
01. Claude assembles the brief from the client record
Meeting preparer compiles a brief for an upcoming call: who the prospect is, what was discussed last time, what should come up first, what objections to be ready for. The corpus it draws from is the CRM, prior emails, calendar context, plus whatever firm-level material has been linked in.
02. The brief routes by which chapters the prospect engaged with
If the firm shipped an authored book and tracks which prospects received which chapters, Meeting preparer can route accordingly. A reader who downloaded the estate-planning chapter gets a brief that opens on estate planning and skips the retirement-income recap. A reader who has not engaged with anything gets a brief that opens with offering a copy. Without a book, every brief opens on whichever default the agent's prompt suggests.
03. The opening of the meeting matches the brief
The advisor walks into the meeting (in person, or onto the call) and opens with the framing the brief surfaced. With an authored corpus the framing is specific to the chapter the prospect already read, the named method they already know, and the question they already half-answered for themselves. Without one, the opening is generic discovery and the first ten minutes are a warm-up the prospect has already done with three other firms.
“A client that I closed the deal with last Friday bought my book from Amazon before he even came in and met with me.”
Lee Welfel, Financial Advisor
What an authored corpus actually is, for the two agents that need one
A corpus that meaningfully changes Pitch builder and Meeting preparer output is not a website, a one-pager, or a recycled bio. It is the artifact below. This is what Paperback Expert produces in roughly six months for financial advisors, attorneys, and RIA founders.
What ships when the engagement closes
- Twelve recorded interviews of about an hour each, one per chapter
- A 50,000 to 70,000 word manuscript that compresses 12 hours of audio without losing the principal's voice
- An ISBN, an Amazon listing, paperback plus Kindle plus audiobook editions
- A written marketing plan that tells the agent stack who gets a copy and when
- Named methods the principal actually uses in client meetings, not invented for the book
- Full case studies the agent can cite paragraph by paragraph
What this looks like a year from now
Anthropic's May 2026 release will not be the last Claude finance agent lineup. The company has already integrated Microsoft 365 (Excel, PowerPoint, Word, Outlook coming soon) and added data partnerships with Moody's, Verisk, Third Bridge, Fiscal AI, Dun & Bradstreet, Experian, GLG, Guidepoint, and IBISWorld. The agent count will grow. The market researcher will get smarter. New advisor-facing agents will arrive that we do not have names for yet.
The corpus question survives all of that. A book the principal authored, a compliance team reviewed, and a publisher printed is the same artifact next year that it is this year. The Claude release that matters most for advisor authority in 2027 will inherit whatever the principal's book put on the page in 2026.
That is the reason the bet is on the corpus, not the vendor. Two of Claude's current ten agents make this urgent. The next ten will too.
“We went from 1 employee to 40 and scaled from $0 to $300M AUM.”
Ready to build the corpus your Claude agents will cite?
A 30-minute intro call with Michael DeLon. We map which Claude finance agents you plan to run (Pitch builder, Meeting preparer, or both) and what an authored book in your voice would look like as the corpus underneath them. About 1 hour of author time per week, roughly 6 months to a published artifact.
Book a 30-min intro call →Want a corpus the two advisor-facing Claude agents can cite?
Book a 30-minute intro call. We will walk you through the Speak to Write process and what a book dense enough to ground Pitch builder and Meeting preparer in your voice actually looks like for your practice.
Keep going
Related guides for advisors and RIA founders
AI agents and the authority book, the broader frame
The corpus question across every agent class, not only Claude.
AI tools for financial advisors in 2026
The 2026 list and the corpus question every other roundup skips.
The authority book client acquisition framework
Place, pre-read, pre-sell, close. Four jobs, one artifact.
Frequently asked questions
Anthropic released ten Claude finance agents on May 5, 2026. Which of them touch advisor authority?
Two of the ten: Pitch builder and Meeting preparer. Pitch builder produces an artifact the prospect reads (the pitchbook itself). Meeting preparer shapes how the advisor opens the meeting. Both inherit whatever the firm's corpus emphasizes, which is where an authored book changes the output. The other eight (Earnings reviewer, Model builder, Market researcher, Valuation reviewer, General ledger reconciler, Month-end closer, Statement auditor, KYC screener) are back-office: they cite filings, ledgers, methodology standards, and entity files. An authored book is irrelevant for those.
Can Claude finance agents replace an authored book by ingesting the firm's website and bios?
Pitch builder and Meeting preparer can pull from a website and bios, and the output will reflect that. The problem is density and verifiability. A bio is two paragraphs. A website has a few feature pages. A 50,000 to 70,000 word book is structurally different: connected chapters, full case studies, named frameworks built up across the manuscript. An agent grounded only on a website produces output that reads as a website. An agent grounded on an authored book produces output a prospect can verify on Amazon before the meeting.
Where do Claude finance agents actually run?
Anthropic ships the ten agents in three forms. As plugins in Claude Cowork (the team workspace) and Claude Code (the developer environment), and as cookbooks for Claude Managed Agents on the financial services marketplace. A wealth management firm picks the form factor that matches its stack: a research desk might run Pitch builder as a Cowork plugin; an operations team might wire Statement auditor as a managed agent. The agent's output behavior does not change with the form factor; the corpus it cites does.
Anthropic added Microsoft 365 integration the same day. Does that change the authority story?
It tightens it. With Claude add-ins for Excel, PowerPoint, Word, and Outlook, context carries automatically between applications. Pitch builder drafts a deck in PowerPoint, the analyst opens an Excel model, the principal edits the cover memo in Word, and the same Claude session retains context across all three. That makes the corpus question more important, not less. Whatever the principal's authored material says, every downstream artifact in the Microsoft 365 stack inherits.
If we are an RIA or estate-planning firm, do we need to wait for Claude finance agents to be useful before commissioning a book?
No. The book is the input layer regardless of which agents you run. If you start the Speak to Write process now, in roughly six months you have a published artifact. By then Claude's agent lineup will have iterated past the May 2026 release. The agents will change, the corpus will not. The book is the input that survives the next two or three Claude releases, the next CRM-with-AI launch, and the next agent vendor that becomes table stakes.
What about the eight back-office Claude agents — does an authored book matter for those at all?
Not directly, and the page should not pretend otherwise. KYC screener cites entity documentation. General ledger reconciler cites the ledger. Earnings reviewer cites filings. Those agents are doing operational work that has nothing to do with the principal's authored point of view. The honest framing is: of the ten Claude finance agents, two are corpus-sensitive in a way an authored book changes (Pitch builder, Meeting preparer), and eight are not. A firm that runs all ten still gets only two-of-ten leverage from a book; that is enough, because those two are the ones the prospect actually sees.
Is connecting an authored book to a Claude agent a compliance issue?
It is a known-artifact story. The principal authored the manuscript, a compliance team reviewed it, a publisher printed it. When Pitch builder cites a paragraph in a deck, the supervisory path is short: the deck traces to a chapter in a published book the firm already approved. Compare that to a Pitch builder grounded only on filings and templates, where the agent's framing language traces to model defaults. Neither path replaces compliance review of agent outputs; the authored-corpus path simplifies it.
Why does Paperback Expert write about Claude finance agents at all?
Because the people we work with (financial advisors, RIA founders, estate-planning attorneys) are watching agents arrive and asking the right question: what does the agent quote when it speaks for me. We have published 275+ books since 2013 and the team is in-house, 29 people. The corpus production process (Speak to Write, about an hour a week of author time, roughly six months end-to-end) is the same one financial-services authors use whether they end up running Claude, a competitor, or no agent at all. Two of Claude's ten agents make the corpus question urgent in 2026.