This article is part of the Claude Cowork for Financial Analysts series. Here we focus on financial modelling support — specifically the documentation, commentary, and assumption rationale work that consumes significant analyst time but receives little credit in research quality assessments.
Financial models are only as defensible as the assumptions behind them. When a portfolio manager challenges a revenue forecast or an auditor reviews your model documentation, the question isn't "what did you assume" — it's "why did you assume that and what evidence supports it." Claude Cowork produces exactly this: structured assumption narratives that trace back to specific management statements and filing disclosures, generated in minutes rather than hours.
Where Claude Cowork Fits in the Financial Modelling Workflow
Claude Cowork for financial modelling isn't about building models — it's about the documentation and narrative layer that sits above the model. The actual financial model lives in Excel, where Claude for Excel handles formula documentation and structural analysis. Cowork handles the analytical reasoning and narrative generation that requires reading across multiple source documents simultaneously.
Assumption Rationale Narratives
For each key model input, generate a 2-3 sentence narrative explaining the assumption, grounded in specific management statements and filing disclosures.
Quarterly Variance Commentary
Compare actual results against model expectations and generate structured variance commentary explaining the key drivers of any deviation.
Model Update Documentation
After each earnings release, generate a structured change log documenting what assumptions changed, by how much, and why — for internal records and client communication.
Sensitivity Analysis Narratives
For key model assumptions with uncertainty, generate structured commentary on the bull and bear case scenarios and what evidence from filings supports each.
The Cowork Model Documentation Workflow
The standard Cowork modelling support workflow runs across three document types loaded simultaneously: your current model assumptions sheet (or a summary of key line items), the most recent earnings transcript, and the relevant sections of the most recent annual or quarterly filing. With these three sources in the canvas, Cowork can produce fully grounded assumption narratives — not invented rationales, but narratives that trace directly to specific evidence in the loaded documents.
Assumption Rationale Generation
The most common use case is quarterly assumption update documentation. After each earnings release, you update your model assumptions based on new guidance and disclosed metrics. Cowork then generates the written rationale for each updated assumption, citing the specific management statement or financial disclosure that drove the change.
"I've loaded my updated model assumptions sheet, the earnings transcript, and the press release in this canvas. For the following key assumptions — [FY26 Revenue Growth Rate, Gross Margin, Operating Expense Ratio, CapEx as % Revenue] — write a 2-3 sentence assumption rationale for each. Each rationale must: (1) state the assumption value and direction vs prior period, (2) cite the specific management statement or financial disclosure that drives this assumption (with exact quote), (3) note any key risk to this assumption. Use precise, factual language."
Quarterly Variance Commentary
When actuals deviate from model expectations, explaining the variance is both analytically important and time-consuming. Load your prior model assumptions alongside the actual results press release and transcript, and ask Cowork to produce structured variance commentary for the key line items that deviated. The output gives you the analytical explanation of the variance with the supporting evidence — and takes 8 minutes instead of 45.
"I've loaded our prior quarter model assumptions alongside the actual results from the earnings press release. Produce variance commentary for [Revenue, Gross Margin, Operating Income] covering: (1) the actual result vs our prior model assumption (numeric), (2) the primary management explanation for the variance from the transcript, (3) whether this variance was structural or one-time in nature based on management's commentary, (4) how we should incorporate this into forward assumptions. Be specific and cite the transcript directly."
Sensitivity Scenario Documentation
Financial models typically have 2–3 scenarios: base, bull, and bear. Documenting the assumption basis for each scenario — and particularly the bear case — is often done informally or not at all. Cowork can produce structured scenario documentation by loading the key risk factors from the annual filing alongside your scenario assumptions, and generating the narrative that connects the risks to the downside scenario inputs.
"I've loaded our bear case model assumptions and the Risk Factors section from the annual 10-K in this canvas. For each of our bear case revenue and margin assumptions, write a 2-sentence narrative connecting the assumption to specific disclosed risks from the 10-K. Then identify any material risks in the filing that are not captured in our bear case assumptions. This output will be used as internal model documentation."
The Cowork + Claude for Excel Integration for Modellers
The most effective configuration for financial modellers combines Claude Cowork (for document reasoning and narrative generation) with Claude for Excel (for formula documentation and model structure navigation). The workflow: Cowork generates the assumption rationale text from the loaded filings and transcripts; Claude for Excel helps embed that text into the model's documentation tabs and comment cells; and the resulting model is both analytically grounded and fully documented.
For teams running this combination, the typical time saving across a quarterly model update cycle is approximately 90–120 minutes per covered company — roughly 35–40% of the total time currently spent on the documentation layer of the modelling workflow. Across a coverage universe of 10 companies, that's a saving of 15–20 hours per quarter.
If your firm uses a specific financial modelling platform — Visible Alpha, FactSet Fundamentals, or proprietary internal systems — our MCP server development team builds direct integrations that allow Cowork to read model data and output documentation directly into your existing model architecture.
Model Change Log Automation
Every time you update a model after earnings, there's an implicit change log: what assumptions changed, in which direction, and why. Most analysts maintain this informally — or not at all — because writing it up takes time they don't have. Cowork automates it completely.
The setup: maintain a standard model assumptions snapshot before each earnings call (a simple export of your key assumption rows). After the earnings call, load the pre-earnings snapshot alongside the new assumptions and the transcript. Ask Cowork to produce a structured change log. The output documents every assumption revision, the magnitude of the change, and the specific evidence from the call that drove it — ready for client communication or internal records in under 5 minutes.
"I've loaded my pre-earnings model assumption snapshot, my post-earnings updated assumptions, and the earnings transcript. Produce a structured model change log covering: [1] Every assumption that changed between the two snapshots (table: metric, prior assumption, new assumption, % change), [2] For each changed assumption, the specific management statement from the transcript that drove the change, [3] Any assumptions that should have changed based on management's guidance but currently have not been updated — flag these explicitly. Format as a client-ready change log."
More from the Financial Analyst Series
Your Model Documentation Is an Untapped Efficiency Opportunity.
Most analysts spend 15–20% of their modelling time on documentation they know isn't as good as it should be. Cowork fixes that — and the setup pays back in the first earnings cycle.