I listened to your voice memo. Here's what I heard, where I think the real leverage is, and
a few things you didn't bring up that might matter more than the stuff you did.
We're picking one thing to build today.
01
Your business, in one paragraph.
So you know I actually listened.
You run RMDA, a mobile pediatric dental anesthesiology practice. Your team
drives to partner dental offices and puts kids under so the dentist can do their work. The
dental offices are your customers; the kids are your patients but only for the day. You don't
have a waiting room or a front desk. Your "office" is whichever practice you're at.
Your spine is Curve Dental - the cloud software where every chart, schedule,
and bill lives. Most of what your team does all day is moving information into and out of Curve.
02
Your 8-step workflow.
Mapped from your audio. Tell me where I got it wrong.
01
Office books a day
Partner dental office reserves a date with you. Logged in Curve.
02
Email office for patient list (2-4 weeks out)
Your team emails: "Send us the patient list." Required fields: name, email, phone, DOB, insurance type (Medicaid yes/no), parent contact.
03
Create chart for each patient in Curve
Manual data entry from the office's spreadsheet into Curve, one patient at a time.
Bottleneck #2 - manual data entry
04
Chase parents for medical paperwork
Curve sends an automatic intake email. Most parents ignore it. Your team sends manual follow-ups until forms are done.
Bottleneck #1 - low response rate
05
PCP outreach (if needed)
For complex cases, the anesthesiologist calls the kid's primary care doctor for medical info. You said this stays human - I agree.
06
Procedure day
You and the team show up at the partner office and deliver anesthesia.
07
Chase office for post-op docs
Treatment plan, medical necessity form, any other case docs. Anesthesiologist uploads the anesthesia record to Curve.
Document chase #2
08
Bill and reconcile
Medicaid: click through Curve to enter codes + units (1 pre-op + 1 first-15-min + N additional-15-min). Curve sends to insurance directly. Cash-pay: bill the parent at fee-for-service rate.
Then watch the Medicaid portal to make sure payments actually land.
Bottleneck #3 - billing click-through
03
What you said is slowing you down.
Your words, ranked.
Bottleneck 1
Parent paperwork collection
Parents don't fill out the intake forms. Your team chases them with manual follow-up emails until they do.
"A lot of patients don't respond. So we're sending follow-up emails until they fill out all of our paperwork."
Bottleneck 2
Patient info into Curve
Office sends you a spreadsheet. Your team retypes every patient into Curve, one at a time, by hand.
"I'm hoping that it can go into our software on the cloud and get everything it needs."
Bottleneck 3
Medicaid billing click-through
Per case: pre-op code, first-15-minute code, then additional-15-minute codes per increment after. Click, click, click.
"It's really simple. We only use four codes... but it's a little confusing."
Bottleneck 4
Cold outreach (paying $1K/mo)
You're paying a contractor a thousand dollars a month to make cold calls all day to dental offices.
"We're pretty busy right now... but that's down pipeline, something we might want to consider."
04
The HIPAA reality check.
Read this before we touch real patient data.
Important
Patient data is regulated. You can't paste real patient names into ChatGPT or Claude on
the public website. To run AI on your real patient data, you need a BAA
(Business Associate Agreement - the HIPAA contract a vendor signs saying they'll handle
your patient data correctly). Anthropic offers a BAA for paid API customers.
Today we demo with fake patients only. Real ones come AFTER the BAA is signed.
That's not a Claude Code limitation - it's the same rule that applies to every other vendor
that touches your charts.
05
What I want to show you.
Pick the two or three that grab you most. We don't have to do all four.
A
"Read my voice memo, summarize my business"
Already done. This whole page is the demo. I sent your audio to Claude. It transcribed it,
mapped your business into a structured doc, drafted a meeting plan, and generated this page.
Total of my attention: maybe 3 minutes.
A 6-minute voice memo became a polished business map in the time it takes to brew coffee.
B
"Take a chaotic patient list and clean it up"
I'll drop in a fake patient list with missing fields.
Read this patient list. Tell me which rows are missing data. Fix what you can. Output a clean version ready for Curve import.
This is exactly your Step 3 pain (chart creation in Curve). Watch the manual retyping disappear.
C
"Draft my parent follow-up email chain"
Three-email sequence in two languages, in one shot.
Write a 3-email follow-up sequence for parents who haven't filled out medical intake paperwork for their kid's dental sedation. Friendly, not pushy. Mention what we need and why. Spanish version too.
This is your Step 4 pain (chasing parents). Your team writes these by hand right now. Claude drafts; humans still hit send.
D
"Calculate Medicaid billing units from a case"
Pure logic, no PHI involved. Just arithmetic on time codes.
I had a Medicaid case that ran 1 hour 22 minutes. We bill 1 pre-op code, 1 first-15-minute code, then additional-15-minute codes for every increment after. How do I bill this?
This is your Step 8 pain (billing click-through). Output is paste-ready for Curve.
06
10 things you haven't named yet.
You've got a list of obvious pain points. These are the ones that didn't come up in your
voice memo and might be where the real money is. Tell me which of these are actually a
thing for you.
01
Insurance prior authorization automation
Pediatric dental anesthesia + Medicaid = constant prior-auth fights (getting approval
BEFORE the procedure). Template the PA letters, auto-pull medical justification, draft
the appeal when denied.
02
Claim denial + appeal engine
Every denied claim has a reason code on the EOB (Explanation of Benefits - the doc the
insurer sends back). Claude reads the EOB, classifies the denial, drafts the appeal.
Recovers revenue you're already leaving on the table.
03
Office profitability dashboard
Do you know which dental offices actually make you money? Probably not, in detail. Pull
case counts, billed amounts, paid amounts per partner office. Cut the bottom 20%, double
down on the top 20%.
04
Office onboarding playbook
When your cold-call guy lands a "yes," what happens next? An automated sequence: contract
draft, scheduling intake, first-day prep. Today this lives in your team's heads.
05
SOP builder for the team
Every workflow you described lives in your team's heads. Claude can interview each team
member, turn it into written SOPs, and onboard the next hire 10x faster.
06
Pre-procedure parent comms
Parents are terrified the night before sedating their kid. AI-drafted reassurance email,
FAQ, day-of prep checklist = fewer day-of cancellations. Direct revenue protection.
07
Post-op clinical check-in
"Is your child OK tonight" follow-up text the next day. Tracks recovery issues, flags
anything serious for the anesthesiologist, shows parents you care.
08
Personal brand / content engine
You're the face of RMDA. LinkedIn for referring dentists. Instagram for parents.
Conference talks. Claude drafts based on your case notes and voice.
09
Compliance + risk monitoring
State dental anesthesia regulations change. Claude monitors the relevant boards and
summarizes anything that affects you. Cheap insurance against expensive surprises.
10
Cold outreach (replaces the $1K/mo guy)
Personalized first-touch emails to dental offices in your region, qualified by office
size, specialty mix, Medicaid acceptance. Not yet - but later, when you want pipeline.
07
Now zoom out for a second.
We're going to start with one small pilot, because that's how anything actually ships.
But before we converge, I want you to see how big this could get. Because the
question isn't "can AI clean up a patient list." It's "what would RMDA look like if
AI was on payroll across your whole org?"
Part 1 - Roles AI can already play on your team
Think of these as virtual teammates, not features.
The Intake Coordinator
Cleans patient lists from offices. Drafts and sends parent paperwork. Chases follow-ups. Validates that every chart has what it needs before procedure day.
Replaces: $45-60K/yr role
The Insurance Specialist
Drafts prior authorizations. Submits Medicaid claims. Reads denial EOBs (the "Explanation of Benefits" insurers send back), classifies the reason, drafts the appeal automatically.
Replaces: $55-70K/yr role + recovers denied revenue
The Sales Development Rep
Personalized first-touch emails to dental offices in your region, qualified by office size and Medicaid acceptance. Follows up. Books discovery calls. Replaces your $1K/mo cold-caller AND scales 5x.
Replaces: $12K/yr contractor + opens new revenue
The Operations Analyst
Daily dashboard: cases per office, revenue per office, denial rates, days-to-payment. Flags when an office's profitability drops. Tells you which 20% of partners drive 80% of revenue.
Replaces: $50-65K/yr role or $1K/mo consultant
The Patient Experience Manager
Pre-procedure parent comms (the "I'm scared to sedate my kid" anxiety prep). Day-of reminders. Post-op check-in texts. Recovery surveys. Flags clinical concerns to the anesthesiologist.
Replaces: $40-55K/yr role + reduces day-of cancellations
The Marketing Manager
LinkedIn for referring dentists. Instagram for parents. Email newsletters to partner offices. Conference talk drafts. All in your voice, drafted from your case notes.
Replaces: $24-120K/yr (in-house or agency)
The Compliance Officer
Monitors state dental anesthesia regulations. Tracks BAA status with every vendor. Audits your charts for documentation gaps before insurance does. Pulls together responses to records requests.
Replaces: $50K/yr role or $5K/yr consultant
The HR + Training Coordinator
Job descriptions. Resume screening. Onboarding playbooks. Turning your team's tribal knowledge into written SOPs so the next hire ramps in days, not months.
Replaces: $45-60K/yr role
The Bookkeeper
P&L per office, per anesthesiologist, per case. Expense categorization. Tax-prep packets. Reconciliation against your bank.
Replaces: $3.6-12K/yr service
The Clinical Research Assistant
Pulls dosing references for unusual cases. Summarizes a kid's PCP records before a phone call so the call is 5 minutes instead of 30. Tracks your case outcomes for case studies and CE credit.
Replaces: hours of your time/wk
Part 2 - What this is worth in dollars
Rough ranges, not promises. We'd refine these once I know your real numbers.
Direct cost savings
$80K - $200K / yr
Roles partially or fully absorbed. Some you have today, some you'd need to hire as you grow. Either way, you don't add the headcount.
Recovered revenue
$30K - $150K / yr
Denied Medicaid claims that get successfully appealed. Reduced day-of cancellations. Fewer write-offs from incomplete documentation.
Growth from outreach
$100K - $500K+ / yr
Each new partner office = real annualized revenue. AI-powered outreach scales beyond what one cold-caller can do, with better targeting.
Your time recovered
10-20 hrs / wk
Yours and your team's. Hours that go back into clinical work, business development, or just not working at 4am.
Realistic 12-month total impact
$200K - $700K+ / yr
Lower end if you stay current size. Upper end if you scale offices in parallel.
Part 3 - What's possible TODAY vs. SOON vs. EVENTUALLY
The state of AI in 2026 is wilder than most people realize.
Today (Q2 2026)
Already shippable
PDF form parsing (medical necessity, treatment plans)
Spreadsheet cleanup + validation
Multilingual email drafting (English + Spanish)
Computer-use agents that click around web apps like a human
Voice agents that handle outbound calls
Vision models that read X-rays and clinical photos
Long-context models that hold an entire patient history
Anthropic BAA available for HIPAA workflows
Next 6-12 months
Coming soon
HIPAA-compliant inbound voice agents (parents call, AI handles)
Persistent memory: agent remembers every prior interaction
Autonomous workflows that run for days unsupervised
Direct Curve integration (if you get into their partner program)
Real-time clinical decision support during procedures
Multi-agent teams (intake agent passes to billing agent passes to follow-up agent)
Video-call assistants that take notes during your discovery calls
1-3 years out
Where this goes
AI co-pilot whispering to the anesthesiologist mid-case
Predictive scheduling that cuts no-shows by 30%+
Fully autonomous office onboarding (lead to first case)
White-label spinoff: package YOUR system, sell to other anesthesia groups
AI that reads a kid's chart + parent intake and recommends sedation protocol
Custom-trained models that sound exactly like you in writing
Part 4 - What RMDA could actually look like
A picture of you in 2029.
"RMDA does 4x the case volume with the same human team you have today. You hired
one more anesthesiologist. You didn't hire a single new admin."
A dental office books a day. Your AI intake agent emails them, gets the patient list,
cleans it, creates the charts in Curve via the partner API, sends parent paperwork in
their preferred language, follows up until forms are 100% complete, and texts you a
green-light summary the night before the case. You walk in, do anesthesia, walk out.
Your AI billing agent files claims, catches denials, and appeals them. Your AI office-relations
agent monitors which partners are slipping and prompts you to call them. Your AI marketing
agent ships a LinkedIn post twice a week in your voice and pulls in 2-3 inbound office
inquiries per month. Your dashboard tells you which decision to make next. You're not
drowning in admin. You're doing the medicine and choosing what to grow.
But here's the thing
None of that gets built by trying to do it all at once. Every team that's pulled
this off started with ONE workflow, got it boring and reliable, then layered in the
next one. That's why the next section converges back to "pick one." The vision is
the destination. The pilot is the first step.
08
What I dug up about Curve's API.
I went looking so we'd know before this meeting. Here's the actual picture.
Good news
Curve has a real API
Confirmed. They use OAuth 2.0 for authentication (the same login standard
banks use), they list webhooks as a feature, and they already integrate with partners
across payments, voice charting, AI x-rays, and patient engagement.
Catch
It's partner-gated, not self-serve
There's no public developer portal. No documented SDK on their GitHub. No
"sign up and get an API key" page. Access goes through their Integration
Partner program - you contact them, they approve, you build.
Opportunity
Nobody is doing patient intake
I checked their entire partner list. Zero partners are focused on patient intake or
paperwork automation. Curve has their own "Smart Forms" feature, but no third party is
layering AI on top of it. That's the gap.
Practical move
Two-stage approach
Stage 1 (now): Build everything that lives BEFORE and AFTER Curve - patient
list cleanup, email drafting, billing prep. Your team still hand-enters into Curve.
Stage 2 (after we apply): Email Curve's partner team. If they grant API
access, Claude writes directly into Curve and the hand-entry goes away.
What this means for today
We don't have to wait on Curve to ship value. The intake-clean-up, email drafting, and
billing-prep work all run outside Curve and feed into it. Stage 1 is buildable starting
next week. Stage 2 is a parallel track - I'll send you the partner contact info
(info@curvedental.com / 888-910-4376) and we apply.
09
Pick one thing.
Not five. Not three. One. We build it on fake data, get it working, then plug in real patients once your BAA is signed.
My recommendation
Patient list to Curve-ready import
Office emails you a patient spreadsheet. Claude validates every row, flags missing fields,
reformats it into a clean import file ready for Curve. Hits your Stage 2 + 3 pain directly.
Highest-frequency repetitive task in your business. Mostly works without needing Curve API
access on day one.
Backup picks if this doesn't grab you:
- "Parent paperwork follow-up engine" - if you want something more visible to your clients
- "Medicaid billing assistant" - if you want the fastest direct-dollar payoff
10
Questions to answer before you leave.
If you don't know one, your homework is to find out.
What email system does your team use?Gmail, Outlook, something else? Determines how fast we plug Claude into the email side.
Where do completed parent forms come back?PDF email attachments? E-sign service like DocuSign? Curve patient portal? Paper?
Who on your team is the most likely day-to-day user?You're not going to run this yourself. Who's the internal champion?
Have you signed any BAAs (HIPAA vendor contracts) before?Tells me how comfortable you already are with that paperwork.
What's your monthly tooling budget appetite?You're already paying $1K/mo for a cold-caller. There's room.
11
Things I'm NOT saying we can do today.
Just so we're being honest with each other.
"Claude will drive Curve for you next week." Curve's API exists but it's partner-gated. Direct integration depends on getting approved into their partner program. Stage 2, not Stage 1.
"It works on real patient data tomorrow." BAA first. No exceptions.
"AI handles the PCP phone calls." You called this one yourself - that's human work.
"We replace your cold-caller this month." You're busy now. Pipeline isn't urgent. Later.
"You've been thinking about Claude as cool tech. Start thinking about it as a teammate who
never sleeps and types fast. Today we pick one job for that teammate."