Claude Code for Automating SWOT Analysis in Financial Strategy

SWOT analysis is one of the most requested deliverables in financial strategy work, and one of the most frequently botched. A CFO asks for a strengths, weaknesses, opportunities, and threats assessment ahead of the annual planning cycle, and what comes back is usually a recycled slide deck: generic strengths like “strong brand,” vague threats like “market volatility,” and no link whatsoever to the actual numbers in the forecast. The analysis gets presented, nodded at, and quietly ignored because it does not change any decision.

The deeper problem is structural. A rigorous SWOT for financial strategy requires synthesizing inputs that live in different places and different formats: the three-statement model, the debt covenant schedule, the segment profitability report, competitor filings, macro assumptions, and the risk register. Pulling those together manually takes days, and by the time the analysis is drafted, the assumptions have moved. Analysts end up working from memory and last quarter’s slides rather than the current model, which is exactly how a strategic document ends up disconnected from the numbers it is supposed to inform.

Claude Code changes the economics of this task. Because it operates directly on files in your working directory, it can read the model exports, the covenant schedule, and the reference documents you point it at, then produce a SWOT that cites specific figures and flags where the financial data contradicts the narrative. The work shifts from drafting to reviewing, and the review is where a controller or FP&A lead actually adds value. The two prompts below show how to structure that workflow so the output is decision-grade rather than decorative.

Setting up the working directory before you prompt

Claude Code is only as good as the files it can see. Before running either prompt, assemble a single project folder containing your current forecast export, the covenant and liquidity summary, segment-level margin data, and any competitor or market research you want considered. Convert everything to markdown or plain text where possible; tables survive the conversion far better than PDFs, and Claude Code will read a well-formed markdown table more reliably than a scanned board pack.

The first prompt below is the diagnostic pass. It asks Claude Code to extract the financial facts that will anchor the SWOT, then build the analysis on top of them. Note the instruction to ask clarifying questions before executing. This matters in financial work: if the model does not know whether the planning horizon is twelve months or three years, the threats section will be wrong, and you will not find out until the review meeting.

I want to build a financially grounded SWOT analysis for [COMPANY / BUSINESS UNIT] covering the [PLANNING HORIZON, e.g. FY2027] planning cycle so that the strategy team can defend every point with a number from our own model rather than generic market commentary.

First, read these files completely before responding:
[forecast_export.md] — current three-statement forecast, base case, by quarter
[covenant_summary.md] — debt covenants, headroom, liquidity position, maturity schedule
[segment_margins.md] — revenue and contribution margin by segment and region, last 8 quarters
[risk_register.md] — existing risk items, owners, and mitigation status

Here is a reference for what I want to achieve:
[Upload a prior SWOT or strategy memo as markdown that you consider high quality]

Here’s what makes this reference work:
– Every strength and weakness ties to a specific metric or trend, not an adjective
– Opportunities are sized in revenue or margin terms with a stated assumption
– Threats include a quantified exposure range, not just a category
– No item appears in both strengths and opportunities; the categories are kept distinct
– Tone is direct and unhedged; no filler phrases like “in today’s dynamic environment”

Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: Four-section SWOT, 4-6 bullets per section, each bullet 1-2 sentences with the supporting figure cited
Recipient’s reaction: The CFO should be able to challenge any bullet and find the number already attached to it
Does NOT sound like: A consulting template, a marketing brochure, or a list of industry truisms
Success means: Zero bullets that cannot be traced to a file in this directory, and at least three items the leadership team has not previously discussed

My context file contains my standards, constraints, audience. Read it fully before starting.
DO NOT start executing yet. Ask clarifying questions first.

Give me your execution plan (5 steps max) before you begin.

Once the diagnostic SWOT exists, the second prompt turns it into the artifact that actually gets used: a strategy memo that pairs each SWOT item with a financial implication and a recommended action. This is where most SWOT exercises die, because the leap from “we have a weakness in working capital” to “here is what we do about it and what it costs” is the part nobody has time to write. Claude Code can draft that bridge quickly, and because it has already read the model, the cost and benefit figures it proposes will be consistent with your forecast rather than invented.

Keeping the human in the loop

Two guardrails are worth building into the workflow. First, require Claude Code to mark any claim it could not source from the provided files, so you can see exactly where it is extrapolating. Second, run the output past the segment owners before it reaches the executive committee; a model can identify that a business unit’s margins are deteriorating, but only the operator knows whether that is structural or a one-off. The prompt below includes both requirements.

I want to convert the SWOT we just produced into a financial strategy memo so that each item carries a quantified implication and a funded action the leadership team can approve or reject.

First, read these files completely before responding:
[swot_draft.md] — the SWOT output from the previous step
[forecast_export.md] — current three-statement forecast, base case
[capital_allocation.md] — approved budget envelopes and discretionary spend limits
[context.md] — my standards, audience, and formatting constraints

Here is a reference for what I want to achieve:
[Upload a strategy memo or investment paper as markdown that you consider decision-grade]

Here’s what makes this reference work:
– Each recommendation states the action, the owner, the cost, and the expected financial effect
– Quantified items show the assumption behind the number in a single clause
– Risks are paired with a leading indicator that would signal them early
– The memo ends with a clear list of decisions requested, not a summary paragraph

Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: Strategy memo, 900-1200 words, with a decisions-requested section at the end
Recipient’s reaction: The CFO should be able to approve, defer, or reject each action in the meeting without further analysis
Does NOT sound like: A status update, a restatement of the SWOT, or a document that ends without asking for anything
Success means: Every action is costed within an approved budget envelope, and every unsourced claim is explicitly flagged as an assumption

My context file contains my standards, constraints, audience. Read it fully before starting.
DO NOT start executing yet. Ask clarifying questions first.

Give me your execution plan (5 steps max) before you begin.

The practical tip that matters most: version the inputs, not just the outputs. Keep your forecast export and covenant summary in the project folder with dated filenames, and rerun both prompts whenever the model is reforecast. The SWOT stops being an annual ritual and becomes a living document that stays synchronized with the numbers, which is the entire point. Try starting with a single business unit rather than the consolidated group; the tighter scope produces sharper, more defensible output, and it gives you a calibration run before you apply the workflow to the full portfolio.

Published on 17 November 2026 on growwithgpt.com