The environmental, social, and governance (ESG) reporting landscape has become a compliance minefield. For CFOs and controllers, the friction is no longer about whether to report—it is about how to survive the sheer volume of data collection, narrative drafting, and assurance-readiness checks demanded by frameworks like CSRD, ISSB, and GRI. Your team is likely spending hundreds of hours manually extracting emissions figures from utility bills, cross-referencing diversity metrics from HR systems, and rewriting boilerplate risk narratives that auditors will inevitably redline. The cost of getting it wrong is not just a footnote; it is regulatory scrutiny and reputational damage.
Anthropic’s Claude Code offers a pragmatic solution to this chaos. Unlike generic chatbots that generate plausible-sounding text, Claude Code operates as an agentic coding and document processing tool. It can read entire file directories, parse structured and unstructured data, apply your organization’s specific reporting standards, and generate draft disclosures that are internally consistent. More importantly, it does not hallucinate numbers—when configured correctly, it pulls directly from your source files and flags gaps. This shifts your team’s role from data entry to review, cutting weeks off the reporting cycle and reducing the risk of material misstatement.
The key to unlocking this efficiency is not the tool itself, but how you instruct it. A vague prompt like “write my ESG report” will yield generic fluff. A structured, context-rich prompt—what we call an anatomy-of-a-prompt—turns Claude Code into a junior analyst that never sleeps. Below, I provide two copy-paste-ready templates you can adapt immediately: one for the initial data gathering and gap analysis, and one for drafting the actual narrative disclosure.
Bridging the Gap: From Raw Data to Structured Intelligence
Before you draft a single sentence, you need a complete inventory of what data exists, where it lives, and what is missing. Most ESG failures in year one are data failures, not narrative failures. The first prompt below forces Claude Code to act as a meticulous auditor, reading your file tree, cross-referencing against a disclosure checklist, and producing a gap report. This is the highest-leverage step because it prevents you from writing a narrative that later collapses under missing evidence.
First, read these files completely before responding:
[esg_data_folder/] — contains all raw utility bills, HR diversity reports, safety incident logs, and board meeting minutes for FY2025
[prior_year_report.pdf] — the FY2024 ESG report we published, including all metrics and narrative sections
[csrd_requirements.md] — the specific disclosure requirements from the Corporate Sustainability Reporting Directive that apply to our entity size and sector
Here is a reference for what I want to achieve:
[Upload a sample gap analysis report from a Big 4 auditor, or describe the structure: a table listing each disclosure requirement, the data source, the file path where the evidence lives, and a status column marked “Complete”, “Partial”, or “Missing”]
Here’s what makes this reference work:
– It separates data availability from data quality (a file can exist but be stale)
– It assigns a confidence score (High/Medium/Low) to each data point based on whether it is primary source data or a derived estimate
– It explicitly lists the owner of each data gap (e.g., “Operations Director” for missing Scope 1 fugitive emissions)
– It flags any metric where the prior year’s methodology changed, so we can note restatements
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: A structured markdown report, maximum 15 pages, organized by ESRS topical standards (E1, E2, S1, G1)
Recipient’s reaction: The CFO should be able to read this in 30 minutes and immediately assign owners to the “Missing” items without asking clarifying questions
Does NOT sound like: A generic sustainability essay, vague corporate fluff, or a request for more budget
Success means: We can identify the critical path to a complete data set within 48 hours, and we know exactly which disclosures we might need to omit or explain as “in development”
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.
This prompt works because it forces Claude Code to be an auditor, not a writer. The success brief is explicit about the recipient (CFO) and the measurable outcome (48-hour critical path). Notice the instruction to ask clarifying questions first—this prevents Claude from making dangerous assumptions about your file structure. In practice, you will run this prompt, receive three to five clarifying questions, answer them, and then receive a gap report that becomes your project plan for the next two weeks.
From Gap Analysis to Draft Narrative: Generating the Disclosure
Once you have the gap report and have filled the missing data, the second prompt takes over. This is where you transform tables of numbers into a coherent, audit-ready narrative. The danger here is that Claude will write in a generic “corporate sustainability” voice that sounds like every other greenwashing document. The second prompt below anchors the output to your specific prior-year report, your tone, and the exact regulatory structure. It also demands that the narrative reference specific file paths as evidence, so your assurance team can trace every claim back to a source.
First, read these files completely before responding:
[gap_analysis_report.md] — the output from our prior gap analysis, listing all complete data points and their source file paths
[fy2025_verified_data/] — contains the final, validated data files (emissions calculations, HR metrics, governance policies) with version control dates
[prior_year_report.pdf] — our FY2024 report, which contains the baseline narrative tone and the methodology notes we must carry forward
[assurance_provider_guidance.md] — specific instructions from our external auditors on what language is acceptable vs. what triggers additional review
Here is a reference for what I want to achieve:
[Upload a sample CSRD-aligned narrative from a peer company, or describe the structure: each section starts with a “Summary of Approach” (2-3 sentences), followed by “Performance Against Prior Year” (with specific numbers), followed by “Forward-Looking Targets” (with clear timelines)]
Here’s what makes this reference work:
– It uses precise, quantitative language (“reduced Scope 1 emissions by 12.4%”) instead of vague claims (“we are committed to reducing emissions”)
– It explicitly ties each narrative claim to a specific data table or file, using bracketed references like [see E1-2_data.xlsx]
– It acknowledges limitations honestly (“data on Scope 3 Category 15 is estimated with a margin of error of ±8%”)
– It mirrors the exact numbering and headings from the ESRS framework to make audit mapping trivial
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: A draft narrative of approximately 40-50 pages, covering ESRS E1 (climate), E2 (pollution), S1 (own workforce), and G1 (governance), with each section clearly delimited
Recipient’s reaction: Our external auditors should be able to map every sentence to a data point or a policy document without asking us for clarification; our CFO should feel confident that the tone is factual, not promotional
Does NOT sound like: Marketing copy, press releases, or the “sustainability as a journey” clichés that dominate non-compliant reports
Success means: We can send this draft to our assurance provider and receive a “comments limited to minor wording” response within 10 business days, with zero requests for additional data
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.
This second prompt is where the real time savings occur. A typical controller might spend three weeks drafting these sections. Claude Code will produce a defensible first draft in under an hour, provided you have done the data work upstream. The critical difference from a standard GPT prompt is the “Does NOT sound like” clause—this is your guardrail against generic ESG speak. The “Recipient’s reaction” clause ensures Claude writes for the auditor, not for a sustainability award committee.
My practical tip for implementation: start with a single material topic, such as ESRS E1 (climate), using the two prompts above. Run the gap analysis, fill the data gaps for that one topic, then generate the narrative. Review the output with your internal sustainability lead before scaling to all topics. Do not attempt to automate the entire report in week one. Also, be explicit about your materiality assessment—Claude needs to know which topics are “material” for your industry, or it will over-report on irrelevant metrics and under-report on what matters.
Next, try pairing these prompts with your existing project management tool. Export your gap analysis into a Jira or Asana board, assign owners to the “Missing” items, and use Claude Code to generate follow-up reminders based on the audit trail. The goal is to make ESG reporting as routine as your quarterly financial close. With this structured prompting approach, you can cut your reporting cycle from four months to six weeks, and your assurance providers will thank you for the traceability.
Published on 25 August 2026 on growwithgpt.com