A primary-year funding banking analyst at a bulge bracket financial institution within the US prices $170k–$190k all-in. They spend most of their first 12 months formatting pitch books, constructing the identical DCF they constructed final month, reconciling accounts that may want reconciling once more in 30 days, and writing variance commentary that explains the previous to individuals who already lived by it.
The ratio of judgment to repetition skews closely towards repetition, and that ratio has not modified in many years. In 2026 it’s beginning to change.
AI isn’t good sufficient to switch monetary judgement (but). However for the repetitive half of the job: the formatting, the primary drafts, the matching, the narrating, AI is now quick, correct and built-in sufficient to be genuinely helpful on the identical afternoon you set it up.
On this article, we’ll have a look at 4 sensible finance workflows the place Claude already reveals robust promise at this time: funding banking supplies, monetary modeling help, month-end reconciliation, and variance evaluation. We’ll additionally have a look at the place it nonetheless wants human overview earlier than anybody ought to belief it in a critical workflow.
How Finance Groups Use Claude for Funding Banking Work
Funding banking runs on paperwork. CIMs, teasers, course of letters, purchaser lists, merger fashions, pitch decks. The work is actual and repetitive: an analyst constructing a one-pager for a deal teaser spends hours formatting, sourcing information, and structuring the identical 4 quadrants they constructed final week for a unique firm.
Anthropic launched a devoted Funding Banking plugin for Claude Cowork on February 24, 2026. It’s open supply, free to put in, and offers Claude 7 slash instructions backed by 9 underlying abilities throughout three workflow classes: deal supplies, displays, and transaction help. Fast terminology word because it comes up all through this information: abilities are the area data modules that activate routinely when related; instructions are the slash instructions you invoke explicitly. Every command calls a number of underlying abilities.
What it comprises
Deal supplies: CIM drafting, teaser era, course of letters, purchaser lists, and information pack extraction from present paperwork. Displays: strip profiles and pitch deck inhabitants utilizing your agency’s branded PowerPoint templates. Transaction help: merger mannequin building and a deal tracker for stay milestones and motion objects.
Putting in it
The plugin requires Claude Cowork (desktop app, Enterprise plan or above) or Claude Code (Professional Plan or above.) Set up the financial-analysis core plugin first, it gives the shared modeling instruments and all MCP information connectors that the IB plugin is determined by. Then add investment-banking on prime.
Through Claude Code:
claude plugin market add anthropics/financial-services-plugins
claude plugin set up financial-analysis@financial-services-plugins
claude plugin set up investment-banking@financial-services-plugins
Through Cowork desktop: Settings → Plugins → Add market from GitHub → enter https://github.com/anthropics/financial-services-plugins → set up financial-analysis, then investment-banking.
/one-pager [Company Name] Generates a single PowerPoint slide with 4 quadrants: Overview, Enterprise, Financials, and Possession. Respects your present template’s margins and branding. That is the strip profile that populates pitch books and purchaser lists.
/cim [Company Name] Produces a full Confidential Data Memorandum: government abstract, enterprise overview, monetary evaluation, and market positioning sections. Claude drafts the construction and content material; your group fills in proprietary information and tightens the narrative.
Remainder of the instructions so that you can strive your self:
/teaser [Company Name] Generates an nameless one-page firm teaser for early-stage deal advertising. Identical core construction because the CIM however stripped of figuring out data.
/buyer-list [Company Name] Assembles a strategic and monetary purchaser universe. Claude categorizes potential acquirers by sort, sizes the match, and constructions the output for straightforward overview and prioritization.
/merger-model [Acquirer acquiring Target] Builds an accretion/dilution M&A evaluation. Output consists of sources and makes use of schedule, professional forma financials, and sensitivity evaluation on buy value and synergies.
/process-letter [Deal Description] Produces bid directions and course of correspondence for a stay transaction.
/deal-tracker Tracks energetic offers, milestones, and motion objects. A structured challenge administration view for stay mandates.
How you can get probably the most out of it
The plugin ships with generic methodology. The true worth comes if you customise the ability recordsdata to your agency: drop in your terminology, reference your branded PowerPoint template within the ability recordsdata, alter the CIM construction to your home format. After that, each CIM draft, each one-pager, each purchaser record comes out in your voice.
Claude carries full context between Excel and PowerPoint in a single session. An analyst can run /merger-model, replace assumptions in Excel, then ask Claude to construct the abstract slide in PowerPoint with out switching instruments or shedding context. This cross-app workflow is in analysis preview for paid plans as of February 2026.
Sincere caveat
These instructions produce first drafts, not closing deliverables. The CIM wants your agency’s proprietary market intelligence. The client record wants your banker’s community data. The merger mannequin wants human verification of each assumption earlier than it goes to a shopper. Use these as the start line, not the completed product.
Utilizing Claude for Comparable Firm Evaluation, DCF Fashions and Valuation Outputs
Uncooked prompting whereas constructing monetary fashions produces output that appears right and isn’t. An analyst at a monetary modeling consultancy ran this check in January 2026: similar immediate to Claude for Excel and Excel’s Agent Mode. Claude’s mannequin had a cleaner structure and higher styling. It additionally discounted money flows utilizing a debt-to-equity ratio as an alternative of WACC, set the fairness danger premium at 120% as an alternative of 5-6%, and used a unique discounting technique for the terminal worth. It seemed investment-committee-ready and was arithmetically damaged.
That failure mode has a repair, and it’s the financial-analysis plugin.
Putting in it
The financial-analysis plugin can be the muse for the IB plugin from part. For those who put in that already, you may have this too. If not:
claude plugin market add anthropics/financial-services-plugins
claude plugin set up financial-analysis@financial-services-plugins
As soon as energetic, you get two instructions plus MCP connectors to each main monetary information supplier.
/comps [Company Name]
Runs a comparable firm evaluation. Claude selects the peer group, pulls present buying and selling multiples from linked information sources, builds the comps desk, and outputs a formatted Excel workbook with industry-standard construction. The peer choice is the one factor you overview and alter – that judgment can’t be automated. The whole lot else: pulled, calculated, formatted.
/dcf [Company Name]
Builds a full DCF. The plugin’s methodology layer is what makes this totally different from a uncooked immediate: it pulls the present authorities yield curve from LSEG to set the risk-free charge, retrieves historic fairness costs and beta to anchor the price of fairness, and checks for inner consistency earlier than outputting. The inputs are market-driven and traceable, not assumed.
What you continue to confirm each time: WACC inputs (fairness danger premium, beta, value of debt), that the discounting is constant throughout projected money flows and terminal worth, and that FCF is pulling from the appropriate line objects. The plugin prevents the plain failures. It doesn’t remove the necessity for a human to learn the mannequin. Wall Avenue Prep’s 2026 testing discovered that Claude hallucinated historic monetary information and each AI instrument scored zero on circularity dealing with: each dangers that persist no matter plugin.
Utilizing Claude in Excel with out slash instructions
The plugin instructions produce new fashions. Claude in Excel additionally works on fashions you have already got, and that is the place it earns time day-after-day.
An analyst inheriting a 47-tab mannequin constructed by somebody who left the agency asks: “Clarify this whole spreadsheet to somebody seeing it for the primary time.” Claude traces each dependency chain and cites the precise cells. What used to take days of reverse-engineering takes an hour.
Situation evaluation runs conversationally. “What occurs if we delay all Q2 hires by one quarter?” Claude updates each affected cell, preserves the formulation, and reveals the precise runway affect. You discover with out touching the mannequin construction. System debugging works the identical means: as an alternative of searching by cells, you get a direct clarification of which cell is feeding the error, what format it expects, and the place the mismatch originates.
MCP connectors
You probably have energetic information entitlements with S&P International, LSEG, Daloopa, PitchBook, Moody’s, or FactSet and have configured them in your Claude settings, they’re stay in Excel routinely. “Pull [Company]’s LTM income, EBITDA, capex, and internet debt from Daloopa” populates the cells straight. “Get the present 10-year authorities yield from LSEG” updates the risk-free charge stay. The handbook export-format-paste step disappears.
The place to begin
Mannequin audit first. Add an present mannequin and ask Claude to elucidate its construction, map the important thing assumptions, and flag method errors. That works at this time with no plugin required and no danger of dangerous mannequin output. As soon as you’re comfy with how Claude reads your fashions, transfer to situation evaluation. Use /comps and /dcf final, and plan to confirm the monetary logic earlier than something goes to a shopper.
Utilizing Claude for Month-Finish Reconciliation
Account reconciliation sounds easy and destroys days. Each shut cycle, an accountant exports the GL steadiness, pulls the financial institution assertion or subledger element, manually matches transactions, investigates exceptions, paperwork the reconciling objects, and builds a workpaper for audit. Then AR. Then AP. Then intercompany. Then prepaids. By the point the working account is finished, it’s day three of shut.
Anthropic’s finance plugin (totally different from monetary evaluation plugin) ships with a structured reconciliation ability that understands the methodology and applies it persistently. It’s a separate plugin from the financial-analysis plugin utilized in sections 1 and a couple of, and lives in a unique repository.
Putting in the finance plugin
claude plugin market add anthropics/knowledge-work-plugins
claude plugin set up finance@knowledge-work-plugins
Or by way of Cowork desktop: Settings → Plugins → Add market → https://github.com/anthropics/knowledge-work-plugins → set up finance.
As soon as put in, Claude has entry to 6 abilities: journal-entry-prep, reconciliation, close-management, financial-statements, variance-analysis, and audit-support. Every has a corresponding slash command.
Operating your first reconciliation
Drop your GL export and financial institution assertion into the Cowork challenge. Then run:
/reconciliation money 2026-02
Claude compares each side, calculates the distinction, and builds the workpaper. It categorizes every reconciling merchandise: timing variations that may clear subsequent interval, objects that want a journal entry, and exceptions that want investigation. It assigns getting old buckets and flags something over your materiality threshold.
Be aware: For AR subledger reconciliation, use:
/reconciliation accounts-receivable 2026-02
The compounding curve
Month 1: Claude applies the generic methodology. Roughly 60% of things match routinely. You resolve the exceptions in the identical Cowork session: sort out the sample in plain language: “this vendor at all times settles two days after bill date,” “this intercompany cost posts to value middle 402 however must be 408,” “this financial institution payment has no GL equal and must be flagged as a brand new journal entry.” Claude incorporates these explanations into the workpaper and carries the patterns into the subsequent session.
Month 2: Claude applies what it discovered. It handles 85% or extra of matches by itself. The exception record shrinks, and the objects it flags are genuinely uncommon.
Month 3: The reconciliation takes half the time it did in Month 1.
These numbers come from a single practitioner’s account (David Dors, Constructing Revenue, February 2026), not a managed benchmark. Deal with them as directional. The compounding sample is actual no matter precise percentages, each sample you train Claude in Month 1 carries ahead.
With ERP connectors
In case your group has linked NetSuite, SAP, or one other ERP by way of MCP, Claude pulls GL balances and subledger element routinely. With out connectors, you paste information or add recordsdata. The reconciliation works both means.
The sincere limitation
The finance plugin runs inside Cowork, which requires Claude Desktop to be open in your machine. In a single day batch reconciliations, high-volume AP matching, and ERP-native reconciliation throughout a whole bunch of accounts want server-side infrastructure, not a desktop app. For that scale, purpose-built platforms are the appropriate instruments. They encode three-way matching logic, pay as you go amortization guidelines, and intercompany netting at a depth a general-purpose agent doesn’t.
What Claude’s plugin handles properly is the analyst-driven shut workflow: one accountant, a handful of key accounts, a month-to-month cadence the place the time financial savings compound. That’s most finance groups.
Each FP&A group spends hours every shut cycle writing variance commentary. The true problem isn’t quantity, it’s coherence throughout aggregation ranges. A vendor-level change flows right into a GL account, rolls into a price middle, and surfaces on the P&L line. The commentary at every stage must be constant and inform the identical story upward. Sustaining that consistency manually, throughout 4 enterprise items and two product traces, is the place time truly goes.
AI helps with the drafting layer of that drawback, not the reason layer. Claude can generate structured first-draft commentary from a verified information desk, labeling variances, flagging materials actions, sustaining constant tone throughout sections. What it can not do is clarify why a quantity moved with out being advised.
The rationale behind a variance lives in your ERP, your CRM, your headcount system, and the judgment of the analyst who lived by the quarter. Claude produces coherent narrative from the info you feed it. The richer the context you present: prior commentary, GL element, value middle breakdowns, recognized one-time objects, the extra helpful the draft.
Variance commentary continues to be price doing with AI. The drafting step is the one which consumes disproportionate time relative to its analytical worth, and that’s precisely the place Claude delivers.
Utilizing the finance plugin
You probably have the finance plugin put in, run:
/variance-analysis opex 2026-02 vs price range
The plugin decomposes the variance into drivers, builds a waterfall chart, and produces commentary structured by class. For income variances, it breaks out value and quantity results. For OPEX, it disaggregates by division and account. The waterfall goes straight into your reporting bundle.
What the analyst truly critiques
AI-generated variance commentary has one particular failure mode: it narrates what the info says with out understanding what the info means. A 12% income miss within the West area could be a single account that closed late, a structural pipeline drawback, or a pricing resolution that may reverse in Q2. Claude doesn’t know which one. The analyst does. That judgment is the one factor that can not be automated on this workflow.
The place purpose-built instruments have an edge
For groups with enterprise FP&A platforms, objective constructed instruments do variance detection plus narrative era as a linked workflow pulling actuals out of your ERP, operating the calculation, and drafting commentary in a single step. If you’re already paying for one in every of these platforms, use them for this. They’re designed for it.
Claude’s benefit is for groups not able to undertake a full FP&A platform: the finance group that runs on Excel, has entry to Claude by a broader enterprise settlement, and needs to chop commentary time this shut cycle with no new software program implementation.
The place to Begin
Choose one workflow. Not all.
In case your group does deal work, set up the IB plugin and run /one-pager on a stay firm this week. If you’re in FP&A, take final month’s variance commentary, paste it into Claude with the present numbers, and see what comes again. If you’re in accounting, run one financial institution reconciliation by Cowork this shut cycle and examine the time.
Cheers.
