Solo dermatology practice. 1 NP, 1 office manager. ~30-40 patients/day. Bedrock-only.

Sana, here's how Bright Skin's tools talk to each other today, and where AI actually earns its keep.

Three sections below. Pantry is your raw clinical and billing data. Prep table is the weekly briefs Claude assembles. Plate is what each provider reads on iPad before rounds and what the office manager works from on Mondays.

7
Tools mapped
4
Recipes ready to build
4
Quick wins flagged
~10
Hours/week back (est.)
Legend:
✓ You've got it
? Missing or not set up yet
! Quick win, money or hours on the table
Where to start: click any card with the orange ! badge first, those are the 4 unlocks Claude can earn its keep on.
Pantry
Your raw ingredients, the tools your data already lives in.
✓
Athena Health EHR
DOCX/HL7 real-time
High volume
✓
Athena Practice Management
XLSX weekly
Steady flow
!
External dermpath lab (PDF results)
PDF daily
Steady flow
✓
Patient intake forms
PDF daily
Steady flow
✓
Insurance appeal templates
DOCX weekly
Trickle
✓
Staff schedule (Google Sheets)
XLSX weekly
Trickle
✓
HIPAA compliance policy
PDF rare
Trickle
Prep Table
Where the night's mise-en-place gets done, weekly summaries Claude writes for you.
✓
Patient pipeline
Patient intake Mon 6am
In: Patient intake forms, Athena Health EHR
✓
Biopsy review pending
Lab results Daily 5:30am
In: External dermpath lab (PDF results), Athena Health EHR
✓
Billing reconciliation
Billing Fri 5pm
In: Athena Practice Management
✓
Insurance appeal status
Insurance appeals Weekly
In: Athena Practice Management, Insurance appeal templates, Athena Health EHR
Plate
What gets served to a human, the briefs and drafts you actually act on.
✓
Morning brief per provider
Written by the intake-prep specialist For: Dr. Sana Anwar, NP Aimee
🔒 Provider initials at top of brief before rounds
✓
Malignant biopsy alert
Written by Biopsy Result Summarizer For: Dr. Sana Anwar
🔒 Sana acknowledges in the audit log before patient is contacted
!
Insurance appeal letter draft
Written by the billing-appeals specialist For: Maya (office manager), Dr. Sana Anwar
🔒 Maya reviews + Sana co-signs before send
✓
Weekly clinic P&L
Written by (read-only summary) For: Dr. Sana Anwar
🔒 Read-only, not auto-distributed

The mental model in one picture. Your data moves left to right: raw sources feed weekly silver platters, platters feed the briefs and drafts Claude writes for you, and those land in the channels you actually use (Slack, your inbox, your iPad). The AI layer sits on top of the data prep, never the other way around. Hover any line to trace it. Click a node to open the details.

Pantry (raw data)
Prep (silver platters)
Plate (briefs you read)
Where you read it
Opportunity / suggested

If a Pantry box has no line leaving it, that data is sitting unused, those are usually the loudest quick wins. Dashed boxes on the right are channels we suggest adding.

Bright Skin Dermatology, these are the briefs and drafts your team can stop building manually. Each one shows the ingredients, the time saved, and the back-and-forth Claude does on your behalf.

~30 min/day per provider

Sana reads the day on iPad before her first patient at 8

Provider reads a one-page morning brief on iPad before rounds, with initials and MRN only.

Athena EHR → Intake forms → Patient pipeline summary → Biopsy review queue → Intake-Prep Specialist → Morning brief
Before: Maya prints the list at 7:30am, Sana skims, often misses the buried biopsy flag.
After: Brief is on Sana's iPad at 6am. Malignant flags at the top. Zero misses.
Click for the full walkthrough →
Hours, plus zero missed-flag risk

Malignant biopsies page Sana within 60 seconds, with a draft patient call script

When a biopsy comes back malignant, Sana is paged within 60 seconds with the summary, recommendation, draft call script, and pre-auth check.

Dermpath PDF result → Athena chart context → Biopsy review queue → Telegram alert + call script
Before: Faxed PDFs filed by Maya, Sana sees results 1-3 days later.
After: Malignant results paged immediately. Patient called same morning. No missed flags.
Click for the full walkthrough →
4-6 weeks per appeal

Dupixent denied? Appeal drafted Monday, sent Tuesday, weeks shaved off

When a claim is denied, draft the appeal in a controlled cross-domain bridge between clinical and billing. Save weeks per case.

Denied claim → Appeal template → Encounter notes (clinical) → Drafted 7-section appeal
Before: Manual appeal: 6 weeks turnaround, often missed entirely on smaller denials.
After: Drafted Monday, sent Tuesday. 4-6 weeks of revenue recovery accelerated.
Click for the full walkthrough →
~1 hr/wk plus better visibility

Sana reads the clinic's week with coffee, no Athena hunt

Read-only weekly P&L brief, numbers only, no PHI. Owner reads on iPad with coffee.

Athena PM ERAs → Weekly billing summary → P&L brief
Before: Sana logs into Athena Monday afternoon if she remembers.
After: One-page brief on iPad Monday 6am. Numbers visible at a glance.
Click for the full walkthrough →

Bright Skin Dermatology, here's the order to build this in. Each step builds on the last, total: About 1-2 weeks for Bedrock + BAA. Then a focused weekend for steps 2-6 (or ~1 day with @claude-code-guide handoff).. If a step looks like a cliff, hand it to @claude-code-guide.

  1. 0
    Pre-step (skip if you already have Claude Code)

    Get Claude Code on your machine

    What

    Install the Claude Code CLI. It's the tool that runs everything below. If you've never opened Terminal, that's fine, this is the only command-line moment until you're set up.

    Install
    npm install -g @anthropic-ai/claude-code
claude --version  # should print a version number
    You'll know it's working when: typing claude --version in Terminal prints a version number instead of an error.

    Setup time: 5 minutes.

  2. 1

    Get the HIPAA-compliant Claude turned on (BAA + Bedrock)

    What

    AWS Bedrock 3P Claude config + BAA signed with Anthropic. The community deep-dive has the IT-committee FAQ and a sample compliance memo.

    Why this first

    PHI cannot leave the tenant. Every step that follows depends on this being done first. The data_map is unusable in production until this is signed.

    Install
    1. Spin up AWS account (or use existing).
2. Enable Bedrock 3P Claude in us-east-1.
3. Sign BAA with Anthropic (15 min legal).
4. Configure Claude Code to use Bedrock model in claude config.
    Before: PHI flowing through public anthropic.com. Compliance hazard.
    After: All model calls happen inside the practice's AWS tenant. BAA on file. CCO can sign off.
    You'll know it's working when: Run 'claude config get model' from Terminal, output includes 'bedrock'.

    Setup time: 1-2 weeks (legal + IT)

  3. 2

    Make Claude read your messy files (PDFs, scanned intake, faxed labs)

    Requires: Step 1 done.
    What

    SessionStart hook converts PDF / DOCX / EML in data/raw_dropzone/ into markdown inside data/converted/. Path-scoped so PHI never leaves the clinical/ tree.

    Why this first

    Removes the 'Claude can't read this PDF' wall for biopsy results, intake, and appeal templates.

    Install
    Drop SessionStart entry into .claude/settings.json. brew install pandoc poppler. pip install xlsx2csv.
    Before: Faxed lab PDFs filed by hand. Intake retyped. Templates copy-pasted.
    After: Drop a PDF in data/raw_dropzone/biopsy/, restart session, see the markdown in data/converted/biopsy/.
    You'll know it's working when: Drop today's intake PDF in raw_dropzone, restart Claude, see the .md file appear with the patient's chief complaint already extracted.

    Setup time: 10 minutes

  4. 3

    Build your morning briefs (one per care area)

    Requires: Step 2 done.
    What

    Write four ingest skills (Athena encounters, Athena billing, biopsy summarizer, intake forms). Schedule via cron. Output silver_platters/ files daily.

    Why this first

    The silver platters are the 80%. Once they exist, every specialist below has clean PHI-scoped data to read.

    Install
    # Skills under .claude/skills/:
#  ingest_athena_encounters
#  ingest_athena_billing
#  summarize_biopsy_pdf
#  ingest_intake_forms
# Cron: 0 5 * * * cd ~/brightskin && claude /weekly_silver_platters
    Before: Sana hunts through Athena chart by chart. Maya retypes intake. Faxes pile up.
    After: Daily 5:30am, four silver platters land in silver_platters/. Initials and MRN only. PHI walls hold.
    You'll know it's working when: 5:31am, ls silver_platters/ shows four files dated today, biopsy_review_pending.md has malignants flagged at the top.

    Setup time: 1-2 days with @claude-code-guide handoff (recommended for non-technical clinics)

  5. 4

    Hire the care chief-of-staff (and three specialist agents)

    Requires: Step 3 done.
    What

    Define four subagents under .claude/agents/. Care Coordinator routes Sana's free-text questions. Intake-Prep Specialist scoped to patient_pipeline. Biopsy Summarizer scoped to biopsy_review. Billing-Appeals Specialist is the controlled cross-domain bridge.

    Why this first

    Each specialist is path-scoped to one platter. Cross-domain (clinical + billing) is allowed ONLY for the appeals specialist, logged on every read.

    Install
    # .claude/agents/:
#  care-coordinator.md (model: opus, tools: [Read, Glob, Task])
#  intake-prep-specialist.md (sonnet, scoped clinical/intake/**)
#  biopsy-summarizer.md (sonnet, scoped clinical/biopsy/**)
#  billing-appeals-specialist.md (sonnet, controlled bridge: clinical/** + billing/**, logged)
    Before: Generalist agent reads everything. Compliance hazard.
    After: Each specialist sees only their slice. Appeals specialist crosses domains only when invoked, logs every read.
    You'll know it's working when: Ask Care Coordinator 'pull KT's history' from a billing context, it refuses (out of scope). Ask Billing-Appeals to draft an appeal, it pulls clinical context and logs it.

    Setup time: 2-3 hours total

  6. 5

    Add receipts, sign-off nudges, and the PHI guardrail

    Requires: Step 4 done.
    What

    PostToolUse hook to outputs/audit_log.md. Stop hook nudges Sana if a brief is unsigned. rules/phi_scoping.md is always-on for clinical/** and intake/** paths.

    Why this first

    PHI scoping is the technical backstop for the BAA. Audit log is the trust layer. CCO can audit any read at any time.

    Install
    Templates in business_os_hooks.md. Paste 3 hook entries into .claude/settings.json. chmod +x hooks. Drop rules/phi_scoping.md in.
    Before: AI did something? You don't know what. CCO has no audit trail.
    After: Every action recorded. Stop hook nudges if a draft is unsigned. PHI scope locks at the OS level.
    You'll know it's working when: Open audit_log.md after a session, every read and write is logged with timestamp and the rule it ran under.

    Setup time: 30 minutes

  7. 6

    Turn weekly motions into one-keystroke plays

    Requires: Steps 3 and 4 done.
    What

    Write four slash commands: /morning_brief, /draft_appeal, /weekly_pnl, /biopsy_alert.

    Why this first

    Slash commands turn Sana's recurring motions into one-keystroke triggers.

    Install
    # .claude/commands/:
#  morning_brief.md
#  draft_appeal.md (takes claim id arg)
#  weekly_pnl.md
#  biopsy_alert.md (manual fallback if cron fails)
    Before: Each interaction is a fresh prompting session.
    After: Sana types /morning_brief, brief lands in 30 seconds. Maya types /draft_appeal CL-3401, appeal lands in 5 minutes.
    You'll know it's working when: /morning_brief returns a clean brief. /draft_appeal CL-3401 returns the 7-section letter.

    Setup time: 30 minutes for all four

  8. You've shipped your back-of-house.

    Next: hand @claude-code-guide a one-line ask for your next skill, or join the Early AI Dopters community to see how other operators evolved theirs.