ع
Start Topics Teams Reference What's new Saved
playbook

Rewrite a policy into plain English

Turn a jargon-heavy policy into something a new hire understands on the first read — every rule kept, a one-line summary up top, and a human confirming nothing legal or factual shifted.

medium ~45 min
when to reach for this

Your handbook is full of policies nobody reads because nobody can read them — "the Company reserves the right, at its sole discretion, to…" runs for three paragraphs before saying anything useful. The fix isn't to dumb it down or quietly drop rules; it's to keep every rule intact while cutting the jargon, and to add a one-line summary up top so a new hire gets the gist in five seconds. The non-negotiable: a human confirms nothing legal or factual changed in the rewrite.

gather this first
  • The current policy — paste it straight into the chat in Claude Desktop, or drop in the file (remote-work-policy.md or the relevant handbook section).
  • Who it's for and what they keep getting wrong about it — the questions HR answers over and over are exactly what the rewrite should pre-empt.
  • Any hard constraints: clauses that are legally required, must stay verbatim, or were written by counsel and can't be paraphrased.
the workflow

  1. Have Claude map every rule before rewriting a word

    Start by extracting the actual rules as a checklist. This is what lets you prove later that nothing was lost — you can't verify "every rule kept" against a wall of prose, but you can against a list.

    you ask
    Read remote-work-policy.md. Before rewriting anything, list every distinct rule, requirement, and right it contains as a plain checklist — one line each — and flag any clause that reads like required legal language that should stay verbatim. Don't rewrite yet.

    what you get back A numbered checklist of every rule ("1. Employees may work remotely up to 3 days/week. 2. Manager approval required for full-remote…") plus a short list of clauses marked "likely legal — keep verbatim" — your reference for the verification step.

    This rule-checklist is your verification anchor. Save it; you'll diff the rewrite against it at the end.

  2. Rewrite for first-read comprehension, rules untouched

    Now do the plain-English pass, but bind it explicitly to the checklist so meaning can't drift. Plain English means shorter sentences and everyday words — not fewer rules.

    you ask
    Now rewrite this policy in plain English a brand-new hire understands on the first read: short sentences, everyday words, 'you' and 'we', active voice, no Latin or legalese. Keep every rule from the checklist exactly as binding — change the wording, never the meaning — and leave the clauses I marked 'keep verbatim' unchanged. Use headings and bullets where it helps.

    what you get back A readable policy at roughly an 8th-grade reading level, organized with clear headings, where the marked legal clauses survive word-for-word and every checklist rule is still present and just as binding.

  3. Add the one-line summary and a quick-reference

    The summary up top is what makes the policy actually get read. A new hire scans one line, gets the gist, and reads the detail only where it applies to them.

    you ask
    Add a one-line 'the short version' summary at the very top, then a 3–5 bullet quick-reference of the rules that affect day-to-day life (what you can do, what needs approval, who to ask). The full policy stays below for the details.

    what you get back A summary line a new hire grasps in five seconds ("Work remote up to 3 days a week; full-remote needs your manager's OK") and a short quick-reference, with the complete policy underneath for when the details matter.

  4. Diff the rewrite against the original rules

    This is the verification gate. Have Claude prove the rewrite is faithful by checking it line-by-line against the rule checklist — and show its work so you can audit it, not just trust it.

    you ask
    Now compare the rewrite against the rule checklist from step 1, line by line. For each original rule, show: is it still present, and did its meaning change at all? Flag anything that got softer, stronger, dropped, or newly added. Show your work as a table — I need to verify nothing legal or factual shifted.

    what you get back A rule-by-rule table — "Rule 3: present, meaning unchanged" / "Rule 7: now reads 'should' instead of 'must' — meaning weakened, fix" — so you can see exactly where to look instead of re-reading both versions cold.

    Read this table yourself, and send the result past whoever owns the policy. Claude's diff narrows where to verify; it doesn't replace the human sign-off.

make it your own
  • A whole handbook: run policies one at a time, but keep a shared style-rules.md (reading level, person, banned words) so every rewritten section sounds like one voice — the same idea as the brand-voice system, applied to HR docs.
  • Bilingual policy, same diff discipline: once the English plain-language version is signed off, produce the Arabic — but here translate faithfully, don't creatively adapt: a policy's rules have to mean exactly the same thing in both languages. Run the same step-4 rule-by-rule diff on the Arabic version, and have a fluent person (ideally the policy owner or counsel) confirm no rule got softer, stronger, or dropped in the crossing. Meaning can shift in a translation just as easily as in a rewrite.
  • Make it repeatable (Power Track): save the map → rewrite → diff flow as a /plain custom command (see the Playbook's Capabilities tab) so every policy refresh runs the verification step automatically, not just when someone remembers. Custom commands are the opt-in Power Track — on Desktop you keep the three prompts in a note and run them in order each time, which preserves the all-important diff step just as well.
watch out for
  • Plain English can quietly change meaning — "must" becomes "should", an exception gets dropped, a deadline softens. The line-by-line diff exists precisely to catch this; never skip it to save time.
  • Claude is not a lawyer. Anything with legal weight — contracts, leave entitlements, disciplinary rules, anything regulated — goes past a real lawyer or the policy owner before it ships. Mark legally-required clauses to keep verbatim.
  • Claude drafts the rewrite; a human owns whether it's still accurate and binding. The summary and quick-reference are a convenience, not the authority — the full policy stays the source of truth.

you'll end up with A policy a new hire actually understands on the first read — every rule intact, a one-line summary up top, and a rule-by-rule diff plus human sign-off proving nothing legal or factual shifted.

Questions people ask

What do I need to prepare before starting?
The current policy as a file (e.g. `remote-work-policy.md`), a note on who reads it and what they keep getting wrong, and a list of any clauses that are legally required to stay verbatim or were written by counsel. That context is what lets Claude produce a faithful rewrite, not just a shorter one.
How long does this take?
About 45 minutes for a single policy: a few minutes to extract the rule checklist, 10-15 minutes for the plain-English rewrite, and then the diff step which is the most important and shouldn't be skipped. Human sign-off time is on top of that and depends on your review process.
Can plain-English rewrites accidentally change the legal meaning?
Yes — that is exactly the risk this workflow is designed to catch. "Must" quietly becoming "should", a deadline softening, or an exception getting dropped are all real failure modes. The rule-by-rule diff in step 4 is specifically there to surface these shifts; never skip it to save time, and always have the policy owner or legal counsel sign off.
Can I adapt this for a whole handbook at once?
Run policies one at a time, but keep a shared `style-rules.md` specifying reading level, person (you/we), and any banned words — so every rewritten section sounds like one consistent voice. The 'Whole handbook' variation in this playbook describes this approach.