Write and update policies

Write and update policies

Writing a policy is rarely a blank-page job. You have existing standards to stay consistent with, regulations to honor, and reviewers who will ask what changed and why. This walkthrough shows you how to draft a policy that’s grounded in your real source material, refine it without retyping, and save the finished version back where it becomes a citable part of your knowledge base.

We’ll use a running example: a Healthcare Remote Work Policy. By the end you’ll have a reviewed, downloadable policy document — and that document will itself be a searchable source the next time someone asks “what’s our remote-work rule for clinical staff?”

What you’ll need

• The policies, templates, or standards your new policy should stay consistent with (PDF, DOCX, TXT, XLSX, Markdown, or HTML).
• Optional: the web addresses of any public regulatory pages you want Chat to draw on.
• Permission to create a Knowledge Collection in your workspace. Most members can; if you can’t, ask your admin.

Why ground a policy in sources?

Chat answers and drafts from exactly what you give it. When you point it at a curated Knowledge Collection, the draft reflects your standards and definitions — not generic boilerplate — and every supporting fact comes with a citation you can check.

Step 1 — Create a collection for your source material

Start by gathering the documents your policy needs to be consistent with.

1. Open Knowledge from the sidebar and choose New Collection.
2. Name it something recognizable — Healthcare Remote Work Policies — and add a short description like “Source policies, standards, and regulations for remote-work drafting.”
3. Save. You now have an empty collection ready for documents.

The collection starts as Personal (visible only to you) while you set it up. You’ll decide who else can see it in Step 6.

Step 2 — Add your existing policies and standards

Fill the collection with the material the draft should respect.

1. In your collection, choose Add Source → Upload.
2. Drag in your documents — your current remote-work policy, your HR handbook, data-handling guidelines, anything the new policy must align with. Supported formats include PDF, DOC/DOCX, XLS/XLSX, PPT/PPTX, CSV, TXT, Markdown, HTML, JSON, XML, and YAML.
3. Watch each file move through uploading → scanning → ready. The scan checks every file for malware before it joins your knowledge base — nothing is added until it’s confirmed clean.

Limit	Value
File size	Up to 100 MB per file
How many	Add as many as you like — upload them in batches so you can watch each one finish

See Upload a file for the full upload reference.

Optional: crawl regulatory pages from the web

If a relevant regulation or standard lives on a public web page rather than a file, you can bring it in directly:

1. Choose Add Source → Web Crawl.
2. Paste the page URL — or several. Chat fetches the public pages and adds their content to the collection, showing progress as it goes.
3. The pages become searchable, citable sources just like uploaded files.

Use public, official pages

A web crawl only reaches genuinely public pages — private or internal-only addresses are blocked automatically for safety. Point it at official regulator or standards-body pages so your draft cites authoritative sources, and re-crawl when those pages change.

Step 3 — Point a conversation at your sources

Now connect a conversation to the material you just gathered.

1. Open a new conversation.
2. From the Knowledge dropdown above the message box, select Healthcare Remote Work Policies. This tells Chat to draw on that collection — and cite it.

Or use a pre-configured assistant

If your team drafts policies regularly, link this collection to an Assistant so every conversation with that assistant draws from it automatically — no dropdown needed each time. An assistant can also carry a consistent house style and structure, so you don’t restate them in every request. See Choosing an assistant.

Step 4 — Request the draft, with context

A policy draft is only as good as the context you give it. Name the topic, the department or audience, and the jurisdiction — the specifics that make a policy fit-for-purpose rather than generic.

Type a request that spells out the structure you expect and the context that shapes it:

Draft a Remote Work Policy for our clinical and administrative staff in California. Use our standard policy structure: Purpose, Scope, Definitions, Policy, Procedure, Roles & Responsibilities, Exceptions, Audit, and References. Stay consistent with our existing handbook and data-handling guidelines, and reflect relevant healthcare privacy requirements. Keep the tone formal and the language clear.

Send it. As Chat works, you’ll see it move through plain-language steps — Understanding your intent, Gathering information, Building context, Composing response — so you can see it’s reading your documents, not improvising.

The draft comes back as a complete document following your structure, with inline citations: numbered markers like [1] and a Sources Used card listing each document the draft drew from. Click any citation to see the exact snippet and open the source. That’s how you know which parts trace back to your real standards.

Naming the jurisdiction matters

A remote-work policy for staff in California is not the same as one for staff in another state or country. Stating the jurisdiction (and department) up front lets Chat tailor the draft — and tells reviewers exactly what scope it was written for.

Step 5 — Iterate until it’s right

A first draft is a starting point. Refine it without retyping your request, using the action buttons under any reply:

Button	Use it when…
Refine this	The draft is close but needs polish — tighter wording, smoother flow. Chat improves its previous answer while keeping the same intent.
Try different approach	A section isn’t framed the way you need. Chat reworks it from a fresh angle instead of tweaking the existing version.
Explain more	A section feels thin and you want it expanded with more detail from your sources.
Copy	Grab the current draft to paste into your document editor.

You can also steer in plain language, exactly as you would with a colleague:

• “Make the Policy section stricter on device encryption.”
• “Add an Exception for staff who travel for work.”
• “Expand Roles & Responsibilities to name the data-privacy officer.”
• “Keep everything else; only rewrite the Procedure section to add an approval step.”

Each reply builds on the conversation, so Chat applies your change to the draft you’re working on and keeps the rest intact. Keep going until every section reads the way your reviewers need it to. For more on the drafting-and-refining rhythm, see Draft emails and documents.

Step 6 — Finalize, then make it citable

Once the policy is reviewed and approved, close the loop so the finished version becomes part of your knowledge base — not a file that disappears into someone’s downloads folder.

1. Export the finished policy. From the conversation menu, choose Export as PDF or Export as Markdown to download a clean, formatted copy. See Export a conversation.
2. Re-upload it as the authoritative source. Back in your Healthcare Remote Work Policies collection, choose Add Source → Upload and add the approved document. Remove or replace the older version it supersedes, so the collection only ever holds the current, approved policy.
3. The finished policy is now a citable source. The next time anyone asks “what’s our remote-work rule for clinical staff in California?”, Chat answers from this very document and cites it.

Keep the source set clean

Because Chat answers from exactly what’s in the collection, an old draft left alongside the approved version can be quoted back to you. When a policy is finalized, retire the documents it replaces. Your collection stays the single source of truth.

Step 7 — Share it with the right people

With the approved policy in place, open the collection to the people who need it.

1. Open your Healthcare Remote Work Policies collection and find its Access or Share settings.
2. Choose who gets access — a specific person or group, or everyone in your workspace (Tenant-wide), so anyone can ask.
3. Save.

Readers can ask, owners can change

People you share read access with can ask questions and see citations, but they can’t add, edit, or remove documents. The approved policy stays in the hands of whoever owns the collection — exactly what you want for official material.

Returning later for a targeted revision

Policies don’t stay still. When a regulation changes or a section needs an update, you don’t start over.

1. Reopen the same conversation thread from the sidebar — Chat keeps the full history and the context you built.

2. Ask for the specific change you need:

   Update the Procedure section to require manager sign-off before a remote-work arrangement starts. Leave every other section unchanged.

3. Refine with the action buttons if needed, then export and re-upload the new version (Step 6) so the collection always reflects the latest approved policy.

Working in the same thread keeps your edits narrow and traceable — you change one section, not the whole document — and the conversation itself becomes a record of how the policy evolved.

The outcome

You’ve turned policy-writing from a blank-page chore into a grounded, repeatable flow:

• The draft reflects your real standards and regulations, with citations you can check.
• You shaped it with one-click refinements and plain-language edits — no retyping.
• The finished policy is downloaded, re-uploaded, and citable, so it answers future questions about itself.
• Revisions happen in the same thread, narrow and traceable, so reviewers can always see what changed.

The same pattern works for any policy or standard — HR, compliance, security, quality, safety. Gather your sources, draft with context, refine, and fold the result back into your knowledge base.

Where to go next

• Draft emails and documents — the full drafting-and-refining reference.
• Upload a file — the complete guide to adding documents.
• Choosing an assistant — keep a consistent structure and voice across every policy you write.
• Build an internal knowledge base — make your finished policies ask-able by the whole team.