Use the AI Excel add-in for financial modeling to build models faster, with fewer errors. Work in a sidebar that reads sheets, explains formulas, fixes errors, and writes new tabs. Connect live market data and research, then run pre-built Agent Skills for comps, DCFs, and earnings analysis in minutes. Financial teams live in Excel. But cleaning data, fixing formulas, and building models from scratch takes hours. Claude’s new Excel sidebar speeds up this work. It reads and edits your workbook, explains its steps, and keeps your structure and links intact. It also connects to live market data and research sources, and it ships with finance-ready Agent Skills for comps, DCFs, due diligence, earnings analysis, and coverage reports. Claude Sonnet 4.5 leads financial task benchmarks like Vals AI’s Finance Agent at 55.3% accuracy. With this stack, analysts can move from raw files to decision-ready output much faster.

How the AI Excel add-in for financial modeling works

Claude sits in an Excel sidebar. You keep your spreadsheet in view while Claude reads it, answers questions, and proposes changes. You see every change before it lands. Claude shows which cells it referenced, and you can click to jump to those cells. This gives you speed plus an audit trail.

What Claude can do inside Excel

– Read multiple sheets and named ranges to map model logic – Explain how a section calculates a metric, citing cell references – Debug broken formulas, circular references, and inconsistent ranges – Preserve structure while filling templates with new data and drivers – Create new tabs for schedules, scenarios, and sensitivity tables – Rewrite error-prone manual steps as robust formulas or scripts Because the sidebar preserves dependencies, your links, named ranges, and checks survive edits. That reduces rework and rebuild risk.

Set up in minutes

– Install the beta add-in and open the sidebar in Excel – Grant Claude permission to read the current workbook – Connect Microsoft 365 to let Claude search permitted files and emails – (Optional) Add connectors for live data feeds and research – Load finance Agent Skills to standardize your workflows The add-in is in research preview for Max, Enterprise, and Teams users, with a staged rollout. Early users help shape the feature set and ergonomics.

Build core models faster with Agent Skills

Agent Skills are reusable, auditable playbooks. They package instructions, scripts, and helpful prompts so you can run a complete task in a few steps. For finance, six Skills help you ship high-quality work faster.

Comparable company analysis in a few clicks

– Ask Claude to create a comps tab with your chosen peer list – Pull valuation multiples and key operating metrics via connected data sources – Standardize time periods, currency, and calendarization rules – Add outlier detection and row-level comments – Refresh the whole set later with one command Tip: Store your “house” definitions (EV formula, LTM convention, IFRS/US GAAP adjustments) in a hidden Inputs tab. The Skill will reference these rules every time, so comps stay consistent across teams.

Discounted cash flow you can trust

– Generate a clean DCF tab with driver-based revenue and margin builds – Auto-calc WACC with transparent inputs for beta, risk-free rate, and ERP – Create bull/base/bear toggles and scenario switches – Build sensitivity tables for WACC and terminal growth – Produce a summary bridge from EBIT to UFCF with tax and NWC logic Because the DCF Skill writes formulas and explains them, you can follow each step. Ask Claude to justify every assumption and cite the exact cells. You can then adjust any driver and watch the sensitivities update.

Earnings analysis and coverage notes

– Use Aiera to pull real-time transcripts and investor event summaries – Extract key KPIs, guidance changes, and management commentary – Build a clean “Quarterly KPIs” tab and highlight deltas vs. prior quarter – Draft a one-page earnings note and a longer initiating coverage outline – Insert quotes with time stamps and link back to source calls This takes what used to be a half-day and compresses it into under an hour, while keeping source links and a clear audit trail.

Due diligence data packs that actually tie out

– Connect Egnyte to search permitted data rooms and investment files – Use the Due Diligence Skill to parse PDFs and spreadsheets into a master Excel pack – Extract customer cohorts, contract terms, churn, and pricing ladders – Build clean revenue bridges and retention analyses – Cross-check totals against management reports and Chronograph data (for PE) Ask Claude to flag missing exhibits and open items. Keep a running checklist in a QA tab so no diligence point gets lost.

Connect to live information the smart way

Claude’s connectors bring trusted data into your workflow without copy-paste risk. You keep governed access and data lineage.

High-value connectors for finance teams

– LSEG: live pricing, FX, rates, macro, and analyst estimates – Moody’s: credit ratings, research, and data on 600M+ companies – Aiera: earnings call transcripts and investor events, with summaries – Third Bridge (via Aiera): expert interviews and industry intelligence – MT Newswires: fast, global multi-asset news – Chronograph: private equity portfolio metrics and fund-level data – Egnyte: secure search across internal models and documents With Model Context Protocol (MCP), Claude can read only what you permit. It keeps your access controls intact. For setup and prompt tips, check the product docs.

Prompt patterns that work

Try short prompts that state task, scope, and output format. Examples: – “Review ‘Model_vFinal.xlsx’ > Sheet ‘RevBuild’. Identify broken references and inconsistent ranges. Propose fixes and show the cell-level diff before applying.” – “Create a comps tab for [tickers]. Pull LTM revenue, EBITDA, EV, EV/Revenue, EV/EBITDA from LSEG. Standardize to USD. Exclude outliers beyond 2 standard deviations. Add notes for each exclusion.” – “Build a 5-year DCF using base/bull/bear scenarios. Explain WACC inputs with sources. Output a summary table with implied EV and per-share value.” – “Summarize the latest earnings call for [company]. Extract 10 metrics, guidance changes, and 5 management quotes. Add a KPI table with q/q and y/y deltas.” Keep prompts declarative and precise. When you need transparency, ask for a reference list of all cells used and all sources accessed.

Make your work auditable and accurate

Speed is useful only if your numbers are right. Build light, reliable controls into every deliverable.

Use a clear change log

– Ask Claude to maintain a “Changelog” sheet with timestamp, author, and summary – Require a cell diff for each batch of changes – Keep short rationale notes linked to key assumptions This helps reviewers scan updates and approve faster.

Adopt simple model hygiene

– Use named ranges for core drivers and costs – Keep hardcodes on a single Inputs tab, never buried in formulas – Color-code inputs vs. calculations vs. outputs – Add cross-foot checks and a balance sheet tie-out row – Reconcile metrics to source data after each refresh Ask Claude to generate and validate these checks. If a check fails, the sidebar can explain why and propose a fix.

Protect data privacy and governance

– Connect only approved data sources with least-privilege access – Keep sensitive tabs in protected sheets – Avoid pulling PII unless you truly need it – Log all external calls to data providers Claude shows which sources it touched, so you keep compliance clean.

Measure the payoff

To prove value, measure speed and quality gains week by week.

Simple metrics to track

– Time to first draft of a comps set – Time from raw transcription to earnings note and KPI table – Number of formula errors caught by Claude vs. manual review – Refresh time for DCF scenarios and sensitivity tables – Model reuse rate across new deals or coverage names Teams report large time savings on data prep and formula debugging. In adjacent workflows, engineers using Claude-powered agents have seen 8–10+ hours saved per week. Expect similar wins in spreadsheet-heavy finance tasks when you standardize Skills and connectors.

A day-in-the-life using the AI Excel add-in for financial modeling

8:30 AM: You open last quarter’s model. Claude scans the workbook and flags two broken references. It proposes fixes and shows the cell-level diff. You approve and move on. 9:00 AM: You run the Comps Skill. LSEG feeds live prices and estimates. Claude builds a clean comps tab, highlights an outlier on EV/EBITDA, and explains the exclusion. You refresh with one click. 10:00 AM: The CFO wants a DCF with tighter NWC assumptions. You open the DCF tab. Claude adds a scenario toggle and new assumptions in Inputs. It updates sensitivities and explains the impact on implied value per share. 11:00 AM: The company is on an earnings call. Aiera streams the transcript. Claude captures 10 KPIs, guidance changes, and five quotes with time stamps. It drafts a one-page summary and a longer note. 1:30 PM: You connect Egnyte to a diligence room. The Due Diligence Skill extracts customer cohorts, contract terms, and churn. Claude builds a retention bridge and flags missing exhibits. 3:00 PM: You export select charts and bullet points to PowerPoint. Claude helps draft three slide headlines and a takeaway page. You do a final numbers check using built-in cross-foot tests. 4:00 PM: You save. The Changelog shows every change, source, and assumption. Review is fast because the audit trail is clear.

Practical tips for smoother adoption

Start with a pilot model

Pick one live model and define success metrics (time saved, errors fixed, refresh speed). Share wins and lessons learned. Expand to a second team once the pattern works.

Standardize your inputs

Create a single Inputs tab template with: – Currency and calendar settings – Definitions for EV, NTM/LTM logic, and adjustments – WACC inputs and sources – Scenario labels and toggles Claude will reuse these standards in every model it builds or edits.

Codify your “house style” in Skills

Add your formatting rules, rounding conventions, and “do/don’t” lists into the Skill folder. Now every analyst produces consistent output, and onboarding speeds up.

Keep humans in the loop

AI helps you go fast, but you own the numbers. Always review sources, re-run critical tie-outs, and sanity-check outputs. Claude shows its work so you can verify it quickly.

Why this approach works right now

– The Excel sidebar reduces tab-hopping and copy-paste errors – Transparent changes build trust during review – Live connectors keep your model current without manual pulls – Agent Skills bring repeatable quality to core finance tasks – Benchmarked reasoning (like Sonnet 4.5’s strong Finance Agent score) cuts guesswork This is not a gimmick. It is a faster, safer way to do the work you already do every day. In short, the AI Excel add-in for financial modeling helps you move from raw data to solid outputs with speed, clarity, and control. Use connectors for live data, run Agent Skills for repeatable tasks, and lean on the sidebar for clean explanations and safe edits. Keep a tight audit trail, standardize your inputs, and measure your time saved. With a careful setup, you will model faster and review with more confidence. (Source: https://www.anthropic.com/news/advancing-claude-for-financial-services?utm_source=perplexity) For more news: Click Here

FAQ