Claude Project: Your Persistent HIM Department AI Assistant

What This Builds

Instead of re-explaining your department's policies, communication style, and organizational context every time you open Claude, a Claude Project stores all of that permanently. Every conversation starts from shared understanding — Claude already knows your policy formats, your hospital's name, your communication standards, and which regulatory frameworks you operate under. Policy updates that took 2–3 hours now take 20–30 minutes because you never start from scratch.

Prerequisites

• Claude Pro subscription ({{tool:Claude.price}}/month) — Projects require Pro
• 3–5 key HIM policies in digital format (Word or PDF)
• At least one completed denial appeal letter (your best example)
• 1–2 examples of your executive summary format
• Basic comfort with Claude from Level 3 work

The Concept

A Claude Project is like having a new staff member who has read every policy in your department manual, studied your preferred writing style, and is available 24/7. You set it up once with your documents and instructions — and then every conversation with that project starts from that shared foundation.

Without a project: Every conversation, you explain who you are, your hospital type, your policy format, and your writing preferences — before you can ask anything useful.

With a project: You open it and say "Update our ROI policy based on the new OCR guidance" — and Claude already knows what your current ROI policy says and what format to use.

Build It Step by Step

Part 1: Set Up the Claude Project

1. Log in to claude.ai with your Pro account
2. In the left sidebar, look for Projects — click it
3. Click New Project (the + button)
4. Name it: HIM Department — [Your Hospital Name]
5. You'll see two main areas: Project Instructions (always-on context) and Knowledge (uploaded documents)

What you should see: A project workspace with an instructions panel on the left and a conversation window.

Troubleshooting: If you don't see Projects in the sidebar, your account may need to be verified as Pro — check billing settings.

Part 2: Write Your Project Instructions

Click Edit Project Instructions and paste the following, customized with your specifics:

You are the AI assistant for the Health Information Management department at [Hospital Name], a [size]-bed [hospital type, e.g., "community"] hospital in [City, State].

## Your Role
You help the HIM Director with: policy writing and updating, denial appeal letters, executive communications, staff training content, regulatory research, and CDI query templates.

## Department Context
- Staff: [number] coders, [number] CDI specialists, [number] ROI staff
- EHR: [Epic/Cerner/other]
- Coding software: [3M/Optum/other]
- Key regulatory frameworks: HIPAA, CMS IPPS, ICD-10-CM/PCS, Joint Commission, [state] state health department
- Credential standards: RHIA, RHIT, CCS

## Writing Standards
- Policy format: [Describe your standard policy sections: Purpose, Scope, Definitions, Policy, Procedure, References, Review Date]
- Executive communications: plain language, 1 page max, narrative format (no bullet points for senior leadership)
- Physician communications: collegial and specific, cite clinical evidence
- Denial appeals: professional and evidence-based, always cite specific guideline sections

## What NOT to Do
- Do not include patient names, Social Security numbers, or other PHI in any output
- Do not cite specific dollar amounts in policies — use ranges or "as determined by leadership"
- Do not make legal conclusions — flag items requiring legal review

## When in doubt
Ask a clarifying question before drafting — it's better to take 30 seconds to confirm than to draft 2 pages in the wrong direction.

Click Save.

Part 3: Upload Your Reference Documents

Click Add to Project or the document/paperclip icon in the Knowledge section. Upload:

1. Your current HIM policy manual (or the 5–10 most frequently updated policies)
2. Your best denial appeal letter (remove patient-identifying info first)
3. Your executive summary format (1–2 examples)
4. AHIMA compliant query guidelines (PDF from ahima.org)
5. Any active OIG Work Plan items affecting your service lines

What you should see: Uploaded documents appear in the Knowledge section of the project.

Important HIPAA note: Review all uploaded documents to ensure no patient PHI is included. De-identify any example documents before uploading. Claude does not use uploaded documents to train its models, but best practice is to treat all AI inputs as you would any third-party vendor with a BAA.

Part 4: Test the Project with Your First Real Task

Open a new conversation within the project. Try this test:

What to type: "Our state just extended medical record retention for minor patients from age 21 to age 26 for records created after [date]. Update our Medical Record Retention and Destruction policy to reflect this change. Flag any other sections that may be affected."

What good output looks like: Claude opens your uploaded policy, identifies the specific section covering minor patient records, drafts the updated language, and flags the Destruction Schedule table as also needing review — because it knows your policy structure.

Compare to without the project: Claude would ask you to paste the policy, explain the format, and describe what changed — before it could begin.

Real Example: Weekly Policy Update Session

Setup: Your HIM Policy Project has your full policy manual uploaded and your instructions configured.

Input: "CMS published updated guidance on hospital medical record authentication requirements. Here is the relevant section: [paste 2 paragraphs from CMS guidance]. Update our Medical Record Completion and Authentication policy accordingly."

Output: Claude reads the guidance, compares it to your existing policy (which it has in its knowledge), drafts specific language changes for the affected sections, and notes: "Section 4.3 (Physician Authentication Timeframe) and Section 6.1 (Electronic Signature Requirements) need updating. Section 7.2 appears unaffected. Recommend legal review of the authentication timeline change before finalizing."

Time saved: From 90 minutes of reading, comparing, and drafting — to 15 minutes of reviewing and approving Claude's draft.

What to Do When It Breaks

• Claude says it doesn't know the current policy → Check that the policy file uploaded correctly. Re-upload and try again. For very large policy manuals, break them into individual policy documents.
• Claude drafts in the wrong format → Add a more specific format example to your Project Instructions. Paste a sample policy opening and ending and say "always use this structure."
• Claude includes a regulatory citation you can't verify → Ask: "What is the exact source document for that citation?" If it can't produce a specific source, treat it as unverified and research it directly.
• Uploaded documents aren't being referenced → In your prompt, explicitly say: "Refer to the [Policy Name] document I've uploaded when answering this."

Variations

• Simpler version: Use a single long conversation with all context pasted in at the start (no Project needed, but you lose the context between sessions)
• Extended version: Create separate sub-projects for different workstreams: "HIM Policies," "Denial Appeals," "Executive Reports" — each with its own specialized knowledge and instructions

What to Do Next

• This week: Build the project, upload your top 5 policies, and use it for your next real policy update task
• This month: Add your denial appeal template library to the project knowledge and start using it for all new appeals
• Advanced: Connect a second project for CDI query templates and regulatory research — keep the workstreams separate so context doesn't conflict

Advanced guide for Health Information Manager professionals. Claude Pro required for Projects and document uploads.