Hands-on: turn a long policy into a one-pager people will read

The task

Long documents get written once and read by almost nobody. A frequent, high-value job for ChatGPT is turning one into a short, plain-English version people will actually skim, while staying faithful to what the original says. You're going to do exactly that: produce a one-page summary and a three-question FAQ from a policy document, then refine both.

We'll use the Fernway Group hybrid and remote-working policy so we're working from the same material. Download the Fernway remote-working policy. Prefer your own material? A staff handbook section, a process doc or a long email chain works just as well, but keep to the Phase 0 privacy rules and strip anything confidential first. Set aside about 15 minutes with ChatGPT open.

The prompts you'll use

Here are the three prompts the exercise is built around, so you can see the shape before you start.

You're helping me make a dense policy readable for busy staff. From the policy below, write a one-page plain-English summary: a two-sentence overview, then short headed sections with bullet points for the things staff most need to know (days in the office, core hours, how to request a change, equipment, and the data rules). Keep it faithful to the policy; don't add rules that aren't there. Policy: [paste].

Now write a short FAQ of the three questions staff are most likely to ask about this policy, with a two-to-three sentence answer to each. Answer only from the policy; if it doesn't say, write "the policy doesn't specify" rather than guessing.

Good. Two changes: put the two-days-in-the-office point and the 10:00 to 16:00 core hours right at the top where people will see them, and make the whole thing shorter, aim for something that fits on one screen.

Step by step

1. Open a new chat. Fresh conversation, so nothing bleeds in from earlier.

2. Send Prompt 1 with the policy pasted in. You'll get a headed, bulleted one-pager. Don't judge it yet; just read it through once.

3. Check it against the source. This is the heart of the exercise. Open the policy beside the summary and confirm the specifics: is it two days a week in the office, are core hours 10:00 to 16:00, is the data rule about not pasting confidential information into unapproved AI tools actually reflected? Look hard for anything the summary states more loosely, or more strictly, than the policy does.

4. Send Prompt 2 to add the FAQ. Then check those answers the same way. The "if it doesn't say, say so" instruction is deliberate: see whether ChatGPT respects it or invents a plausible answer to a question the policy doesn't actually cover.

5. Refine with Prompt 3, then one or two nudges of your own: "add a line on who to contact for equipment", "make the tone a bit warmer". One change per message.

6. Do the final human check. Read the whole thing once more against the policy. Fix any figure that's drifted, delete anything invented, and make sure nothing important was dropped in the name of brevity.

7. Save what you made. Copy the one-pager and FAQ somewhere real: a note, a draft, a document. You've produced something usable.

Your checklist

• [ ] I started a fresh chat for this task.
• [ ] My first prompt named the reader, the format and a "stay faithful" rule.
• [ ] I checked the key figures (office days, core hours, data rule) against the original.
• [ ] I tested whether the FAQ admitted what the policy doesn't say, instead of guessing.
• [ ] I refined at least twice, one change at a time.
• [ ] I did a final human check and corrected at least one thing.
• [ ] I ended with a one-pager and FAQ I'd actually circulate.

How to tell you did it well

• It's faithful, not just fluent. The summary matches the policy on every specific: no rule tightened, loosened or invented. A readable summary that's subtly wrong is worse than the original.
• You caught something. Most people find at least one drift: a "minimum of two days" softened to "a couple of days", or an FAQ answer confidently made up. Catching it means you were reading critically.
• The FAQ was honest about gaps. Where the policy is silent, your version says so rather than inventing.
• You steered it. The final version is shorter and clearer than the first because you nudged it there, not because the first draft happened to land.

If some of that didn't happen, start a fresh chat and run it again; repetition is how this becomes second nature.

Try it now

Common mistakes

• Judging the summary by how it reads, not whether it's right. A polished one-pager can misstate a rule. Readability is worthless if the facts have drifted.
• Refining everything in one giant message. One nudge at a time is clearer and gives better results.
• Letting brevity drop something important. "Shorter" can quietly delete a rule that mattered. Check nothing essential was lost.
• Trusting a confident FAQ answer to a question the source never addressed. This is the over-trust trap in miniature: ChatGPT will happily answer beyond the document unless told not to, and a made-up policy answer read as fact can cause real problems. Verify every answer against the source, and treat "the policy doesn't specify" as the correct, honest response, not a failure.

Keeping current

The exercise is durable, and it gets more convenient over time. You can increasingly upload the document as a file rather than pasting it. When the ways of feeding ChatGPT a document change, check OpenAI's ChatGPT release notes. Accurate as of 13 July 2026.