Claude CoWork for Pharmacists
A practical guide to using Claude as your AI co-worker in your pharmacy workflow — from setup to daily use.
What is Claude CoWork?
Claude CoWork is the practice of using Claude as a consistent, knowledgeable assistant embedded in your pharmacy workflow. Rather than using AI as a novelty or occasional search tool, you configure Claude with your practice context so that every interaction produces clinically relevant, professionally formatted output that fits how you actually work.
Claude-native prompts. The prompts in this guide use Claude's native XML tag structure (,,,) for more precise, consistent output. These tags help Claude parse your intent with less ambiguity. They work in ChatGPT too, but are optimized for Claude.
For pharmacists, this is particularly powerful. Your day is a constant stream of documentation tasks: prior authorization letters that need to sound clinical and persuasive, patient counseling points that need to be accurate and understandable, MTM notes that need proper formatting, and drug information questions that demand precision. Claude handles the drafting and formatting so you can focus on the clinical decision-making that requires your expertise.
This guide covers how to set up Claude specifically for pharmacy work, the five workflows where it adds the most value, and the critical privacy boundaries you must maintain when using any AI tool in a healthcare setting.
Setting Up Claude for Pharmacy Work
Step 1: Create a Pharmacy Project. In Claude, go to Projects and create one named something like "Pharmacy Practice." This gives you persistent context across every conversation.
Step 2: Set your custom instructions. In the Project settings, add:
You are my pharmacy practice assistant. Here is my context:
<practice-profile>
- Role: [Staff pharmacist / Clinical pharmacist / Pharmacy manager] at [Practice setting: retail chain, independent, hospital, clinic]
- Specialties: [e.g., MTM, immunizations, compounding, oncology]
- State: [Your State] (for scope of practice reference)
- Common insurance plans I work with: [List your top 3-5 payers]
- EHR system: [Epic, Omnicell, QS/1, etc.]
</practice-profile>
<rules>
- Never generate content that could be mistaken for a direct clinical recommendation for a specific patient
- All clinical content should reference current guidelines and evidence-based standards
- Prior auth letters should use ICD-10 and medical necessity language
- Patient education materials should target an 8th-grade reading level
- Always remind me to verify drug information against primary sources (Lexicomp, Clinical Pharmacology, Micromedex)
</rules>Step 3: Upload reference templates. Add your preferred prior auth letter format, your MTM documentation template, any formulary summaries you reference frequently, and your pharmacy's patient education style guide if you have one.
Step 4: Work inside this Project every time. Your custom instructions load automatically, so Claude already knows your practice context.
Your Top 5 Workflows with Claude
1. Prior Authorization Letters
This is where Claude saves pharmacists the most time. A well-written PA letter can mean the difference between approval and a peer-to-peer review. Claude drafts the clinical narrative while you supply the patient-specific rationale.
<task>Draft a prior authorization letter for the following scenario (all data is de-identified).</task>
<clinical-data>
- Medication requested: Eliquis (apixaban) 5mg BID
- Diagnosis: Non-valvular atrial fibrillation (ICD-10: I48.91)
- Insurance: [Payer name], denying as non-formulary, requesting warfarin trial first
- Clinical rationale: Patient is 72 years old, has a history of GI bleed on warfarin (documented 2024), CHA2DS2-VASc score of 4, HAS-BLED score of 3
- Failed therapies: Warfarin (adverse event — GI bleed requiring hospitalization)
- Supporting guidelines: AHA/ACC/HRS 2023 AFib guidelines recommend DOACs over warfarin for most patients with non-valvular AFib
</clinical-data>
<instructions>
- Format as a formal letter to the medical director
- Include ICD-10 codes
- Cite the guideline
- Emphasize medical necessity
- Use clinical language appropriate for peer review
</instructions>
Think step-by-step about the clinical rationale and why warfarin rechallenge is contraindicated before drafting.Before Claude: 20-30 minutes per letter, often from scratch.
After Claude: 5 minutes to input the clinical details, 5 minutes to review and customize the draft.
2. Patient Counseling Scripts
When you need to explain a new medication, a complex regimen, or a drug interaction, Claude can draft counseling points at the right reading level.
<task>Create patient counseling points for a new metformin ER 500mg prescription.</task>
<context>
The patient is newly diagnosed with type 2 diabetes and has no prior experience with diabetes medications.
</context>
<instructions>
Cover:
- What the medication does (simple language)
- How to take it (with food, timing, titration schedule)
- Common side effects and what to do about them
- What to watch for that requires calling their doctor
- Lifestyle points to reinforce
Write at an 8th-grade reading level. Use short sentences.
</instructions>
<avoid>Medical jargon — if you must use a clinical term, define it in parentheses.</avoid>3. MTM Documentation
Medication Therapy Management documentation is time-intensive but critical for reimbursement and patient outcomes. Claude can structure your notes.
<task>Format an MTM encounter note using this information.</task>
<clinical-data>
- Patient: De-identified, 68-year-old female
- Encounter type: Comprehensive Medication Review (CMR)
- Medication list: lisinopril 20mg daily, atorvastatin 40mg daily, metformin 1000mg BID, amlodipine 5mg daily, omeprazole 20mg daily, aspirin 81mg daily, vitamin D3 2000 IU daily
- Drug therapy problems identified:
1. Omeprazole — on for >1 year with no documented indication, risk of long-term PPI use (Mg depletion, C. diff, fracture risk)
2. No documented statin intensity assessment — patient is diabetic, may qualify for high-intensity statin per ACC/AHA guidelines
- Recommendations made: Discussed PPI deprescribing taper with patient, recommended physician reassess statin intensity
- Patient agreed to: PPI taper, follow-up with PCP about statin
</clinical-data>
<format>Standard MTM SOAP structure. Include clinical rationale for each recommendation.</format>
Think through each drug therapy problem systematically before writing the note.4. Drug Information Research Summaries
When a prescriber or patient asks a complex drug information question, Claude can help you organize your response — though you must always verify against primary references.
A physician asks: "Can my patient take ibuprofen while on apixaban?"
<task>Draft a drug information response.</task>
<instructions>
Cover:
- The interaction mechanism
- Clinical significance and risk level
- What the evidence says (note: I will verify citations against Lexicomp)
- Practical recommendation
- Alternative analgesic options to suggest
Format as a brief clinical consult response (under 200 words). Professional tone, cite guideline sources where applicable.
</instructions>
<note>I will verify all clinical details against primary drug information databases before sending.</note>
Think step-by-step about the pharmacological mechanism before drafting the response.5. Pharmacy Operations and Administration
From staff schedules to policy drafts, Claude handles the administrative work that eats into your clinical time.
<task>Draft a pharmacy policy for handling controlled substance discrepancies.</task>
<context>This is for an independent community pharmacy in [State].</context>
<instructions>
Include:
- Purpose and scope
- Definitions (what constitutes a discrepancy)
- Immediate steps when a discrepancy is discovered
- Documentation requirements
- Reporting chain (pharmacist-in-charge, DEA if applicable)
- Investigation process
- Record retention requirements
Reference DEA regulations where applicable. Format with numbered sections and clear headers.
</instructions>
<note>This is an internal policy document, not legal advice — it should be reviewed by our compliance consultant.</note>Prompt Engineering Tips for Pharmacists
1. Always specify the clinical context. "Draft a prior auth for Eliquis" is vague. "Draft a prior auth for Eliquis for a 72-year-old with AFib who failed warfarin due to GI bleed" gives Claude what it needs to write a persuasive letter.
2. Request specific formatting. Say "Use SOAP format," "Format as a clinical consult response," or "Structure as a formulary monograph." Claude knows these formats and will use them correctly.
3. Set the reading level for patient materials. Always specify "8th-grade reading level" or "plain language" for anything a patient will read. Clinical language is appropriate for provider-facing documents.
4. Include your verification step. Add "Note: I will verify all clinical information against [Lexicomp/Micromedex/Clinical Pharmacology] before use" to any prompt involving drug information. This is both a reminder and a documentation habit.
5. Ask Claude to flag uncertainties. Add "If you are unsure about any clinical detail, flag it clearly rather than guessing" to your Project instructions. Claude will mark areas where you need to verify.
6. Use Claude to check your own writing. Paste a draft PA letter and ask: "Review this for clinical persuasiveness, proper ICD-10 coding, and completeness. What am I missing that could strengthen the medical necessity argument?"
Privacy & Compliance
HIPAA is absolute. Never paste real patient names, dates of birth, medical record numbers, Social Security numbers, or any combination of data that could identify a specific patient into Claude or any AI tool. This is not optional.
De-identify everything. When using Claude for clinical scenarios, use de-identified information: "68-year-old female with atrial fibrillation" instead of actual patient identifiers. Change ages slightly, omit dates, and never include chart numbers.
Claude is a drafting tool, not a clinical reference. Always verify drug information, dosing, interactions, and guideline citations against primary clinical databases before using them in patient care. Claude can make errors, and clinical errors have consequences.
Institutional policies apply. Check with your employer about their policy on using AI tools with any patient-related information, even de-identified data. Some health systems have specific AI use policies.
Document your verification. When you use Claude to draft a clinical document, your review and verification is what makes it clinically valid. The final document is your professional work product.
Going Further
Ready to integrate Claude deeper into your pharmacy practice? Explore these resources: