Operations & execution
Rewrite a Process Document with ChatGPT
Summary
Cihan's view: Upload your existing SOP as a DOCX or PDF to ChatGPT. Ask it to rewrite the document with numbered steps, clear headings, a when-to-escalate section, and plain language. Review the draft, then prompt for specific fixes like shortening paragraphs or adding a quick-reference table.
Who this is for and the question
You have a standard operating procedure that people skip or misread because it is written in dense paragraphs, uses internal jargon, or buries the escalation path. You want a version that a new hire can follow without asking questions, but you do not have time to rewrite it from scratch.
Question: How do you use ChatGPT to turn an existing process document into a clear, scannable procedure with numbered steps and a visible escalation guide?
Fictional input
A twelve-page PDF titled “Customer Refund Approval Workflow v3.2” that covers:
- Eligibility rules across three product tiers
- Required evidence (receipt, screenshot, support ticket ID)
- Approval thresholds by refund amount (under 50 USD, 50 to 200 USD, over 200 USD)
- Current turnaround targets (48 hours for tier 1, 72 hours for tier 2)
- A buried escalation matrix in paragraph four of section six
- Fifteen defined terms in a glossary appendix
The file is 1.3 MB and uses DOCX-compatible formatting.
Steps
-
Open ChatGPT at chat.openai.com and sign in with a plan that supports file uploads (Free, Plus, Pro, or Enterprise).
-
Attach the file by clicking the paperclip icon or dragging the PDF into the chat composer. Wait for the file pill to appear showing “Customer Refund Approval Workflow v3.2.pdf”.
-
Paste the following prompt and send:
Rewrite this SOP as a clear, scannable procedure for front-line support agents. Requirements: - Use numbered steps for the main workflow - Add a "Quick reference" table at the top with eligibility, thresholds, and turnaround targets - Create a separate "When to escalate" section with explicit triggers and contacts - Replace jargon with plain language (for example, "initiate" to "start", "adjudicate" to "decide") - Keep all specific rules, amounts, and timelines from the source - Flag any rules that appear contradictory or ambiguous - Output in Markdown with headings for each major section -
Review the first draft. Check that the quick-reference table captures all three tiers, the thresholds match the source, and the escalation section lists specific triggers (for example, “refund over 200 USD” or “missing evidence after 24 hours”).
-
Iterate with targeted follow-ups as needed:
- “Shorten the eligibility rules to one sentence per tier.”
- “Add a decision flowchart description in text form for the approval thresholds.”
- “Move the glossary to an appendix link and keep only the five terms agents actually need daily.”
- “Convert the turnaround targets into a two-column table: tier / target hours.”
-
Copy the final Markdown into your internal wiki or document system. Replace the old PDF link with the new version.
Illustrative output (example)
Quick reference
Tier Eligibility Auto-approve limit Turnaround Basic Purchase within 30 days, digital only 50 USD 48 hours Pro Purchase within 60 days, any product 200 USD 48 hours Enterprise Purchase within 90 days, any product 500 USD 24 hours Workflow
- Verify the request matches tier eligibility.
- Confirm required evidence is attached.
- Compare refund amount to the auto-approve limit for that tier.
- If within limit, approve and notify finance.
- If over limit, route to the tier lead with the evidence pack.
- Log the decision in the refund tracker with the ticket ID.
When to escalate
- Refund amount exceeds the tier auto-approve limit
- Evidence is missing and not provided within 24 hours of request
- Customer threatens legal action or social media escalation
- Request falls outside stated purchase windows
- Contact: Tier Lead (Slack #refund-escalation) or Refund Manager (email refunds-lead@company.com)
This output is a documentation-based illustration. Your actual result will reflect the specific rules in your uploaded file.
Acceptance checks
- The quick-reference table includes every tier, eligibility rule, threshold, and turnaround target from the source.
- The numbered workflow covers the complete happy path without gaps.
- The escalation section lists specific, measurable triggers (amounts, timeframes, conditions) not vague phrases.
- All monetary amounts, time windows, and product names match the source document exactly.
- No new rules, tiers, or contacts appear that were not in the original.
- The Markdown renders cleanly in your wiki with working tables and headings.
Troubleshooting
| Symptom | Likely cause | Fix |
|---|---|---|
| ChatGPT says “file too large” | PDF exceeds 512 MB or 2 million tokens | Export a reduced PDF or split into sections; upload one section at a time |
| Output misses a rule from the source | Model truncated the context or overlooked a page | Prompt: “Re-read section 4 of the PDF and confirm the 72-hour target for tier 2 is included.” |
| Escalation triggers are generic | Prompt did not ask for measurable conditions | Add: “Use exact amounts and hours from the document. No vague words like ‘large’ or ‘extended’.” |
| Tables render as plain text in wiki | Wiki expects HTML or a specific Markdown dialect | Ask: “Output the tables in GitHub-flavored Markdown with pipe syntax.” |
| Free plan hits daily upload limit | Three file uploads used today | Wait for the limit to reset or upgrade to a paid plan |
Closing
Want one practical AI workflow in your inbox each week? Subscribe at https://www.cihanuygur.com/#signup