Recall automation that pays for itself at one $400–$800 implant patient a month

Your recall list is worth more than your ad budget

Dental practices put 4–7% of revenue into marketing, and 74% recently increased that budget, most of it aimed at new-patient acquisition at $150–$300 per general-dentistry patient and $400–$800 for implant and cosmetic cases. Meanwhile the cheapest production in the building sits inside the practice management system: overdue hygiene recall, accepted-but-unscheduled treatment plans, and patients who quietly lapsed eighteen months ago. Most offices work those lists off a printed PMS report on slow afternoons.

That gap is the entire case for automation in dentistry. Re-engaging a patient who already accepted a treatment plan costs a fraction of acquiring a new one at $400–$800, and with only 25–33% of practices using AI anywhere in operations, the office across town probably has not built this either.

One caution first: 75% of enterprises rolled back their customer-facing AI agents by May 2026, citing data exposure (31%) and hallucination (22%) (Sinch, n=2,500+). In a HIPAA environment those failure modes are not embarrassing, they are reportable. Hence the design rules here: doctor-approved templates, human escalation, audit trails. We unpacked the pattern in why companies are rolling back AI agents.

Workflows we build for dental practices

Six dental-specific systems. Most are deterministic n8n workflows, not free-running agents, a distinction that drives cost and risk.

• Hygiene recall sequences. A timed SMS → email → voice sequence with a live booking link, confirmations written back to the schedule, via Open Dental's API directly, or Dentrix and Eaglesoft through a sync layer such as NexHealth.

• Unscheduled treatment-plan follow-up. The highest-value list in any PMS: patients who accepted treatment and never booked, segmented so an implant plan gets a different cadence than a single filling. The doctor approves every template.

• Missed-call text-back and phone triage. A missed call at lunch or after 5pm triggers an immediate text with a booking link; an optional AI receptionist layer triages emergency-versus-routine and pages the on-call dentist, costs itemized in the AI receptionist cost breakdown.

• Broken-appointment backfill. A cancellation fires a same-day SMS offer to patients flagged for short-notice availability, instead of leaving the chair empty.

• Lapsed-patient reactivation. Eighteen-month-plus inactive patients, segmented by last treatment and insurance status, with hard opt-outs and quiet hours built in.

• Post-visit reviews and referrals. Review requests fire on completed appointments, never after a visit the front desk flags as difficult.

What you get

Every dental engagement ships with:

1. A PMS integration audit naming your exact sync path, Open Dental direct API, or Dentrix, Eaglesoft, and Curve Dental via middleware

2. Signed BAAs with every vendor that touches patient data, before any data flows

3. A TCPA consent inventory, who can receive what, plus capture flows for the gaps

4. Recall, reactivation, and treatment-plan sequences built in n8n with doctor-approved templates

5. Missed-call text-back on your existing numbers via Twilio or Telnyx, coexisting with Weave or Solutionreach if you already run them

6. Claude-drafted, human-approved message variants behind a model adapter, so provider swaps are configuration, not rebuilds

7. Escalation rules: any clinical question routes to staff, never to the model

8. An exportable audit log of every automated touch, opt-outs honored on every channel, quiet hours enforced

9. A monthly dashboard: recall coverage rate, reactivation bookings, production scheduled from sequences

10. Front-desk training and a runbook, your team operates this, not us

11. A 30-day post-launch defect window

Explicitly out of scope: anything diagnosis-adjacent, claims adjudication, and clinical messaging. The system books appointments and answers logistics; dentistry stays with the dentist.

The math on one recovered implant patient

Acquiring one implant or cosmetic patient costs $400–$800 in marketing spend. The reactivation list in your PMS holds patients who already accepted treatment, reaching them costs cents per message. If automation recovers a single implant-grade patient per month, it has replaced $400–$800 of acquisition spend plus the full production value of the case, the number your own fee schedule puts on it, measured against the $2,000–$6,000/month market retainer band below.

The break-even question is not whether AI is impressive. It is whether recall books more dentistry daily than it did on slow afternoons. Want this math run on your own lists? Get a scope and quote.

How an engagement runs

1. Scope call, 45 minutes. PMS, phone system, comms stack, consent status, largest dormant lists. Output: a workflow shortlist and the integration path.

2. Compliance and data audit, week 1. BAA chain mapped and signed, TCPA consent inventory, PMS access tested against synthetic patient records.

3. Build, weeks 2–3. Recall and missed-call recovery first, fastest payback. The doctor signs off on every template.

4. Supervised launch, weeks 3–4. Staff approve each outbound message for the first two weeks before full automation.

5. Handover and 30-day watch. Runbook, dashboard, and a day-30 review against the recall-coverage baseline from week 1.

What this work costs in the market

Market anchors with sources, not our rates, your quote gets built against your scope at /contact. Agency retainers for dental marketing automation cluster at $2,000–$6,000/month (NetPartners, single-source, directional). For component-level pricing, how much an AI agent costs itemizes everything from model fees to monitoring.

What moves a dental quote inside that band:

• PMS integration path, Open Dental's direct API is the short road; Dentrix or Eaglesoft adds a middleware subscription

• Location count, each site brings separate phone lines, schedules, and consent records

• Whether AI voice answering is in scope, or text-back alone

• Compliance scope, the length of the BAA chain and your state's rules

• Who operates it after day 30, your front desk with our runbook, or a managed arrangement

Why Entropy & Co

Three claims you can check:

1. BAAs before data, every time. No patient record flows through any system until a signed BAA covers it. There is no such thing as “HIPAA-certified software”, compliance is the configuration plus the paper chain, and we hand you the file.

2. Model-agnostic by construction. Message drafting sits behind a model adapter; swapping Claude for another provider is a config change you can verify in the repo.

3. Clinical questions never get an AI answer. Scope guardrails route anything beyond scheduling and logistics to your staff, and every automated touch lands in an exportable audit log.

FAQ

Does this work with Dentrix, Eaglesoft, or Open Dental?

Yes, through different paths. Open Dental exposes a direct API we integrate against. Dentrix, Eaglesoft, and Curve Dental connect through a sync layer such as NexHealth. The scoping call names your exact path, and what it costs, before you commit to anything.

Is this HIPAA compliant?

Compliance is built, not bought. We sign a BAA with you, require one from every vendor in the chain, restrict data flows to the minimum fields each workflow needs, and log every access. You receive the complete BAA file and a data-flow map as deliverables.

What about TCPA rules for recall texts?

Appointment and recall reminders to existing patients carry different consent requirements than marketing messages, and the system treats them differently. Scoping includes a consent inventory, capture flows for missing consent, automated opt-out handling, and quiet-hours enforcement on every outbound channel.

We already pay for Weave or Solutionreach. Why add this?

Those platforms are channels with basic reminders. The gap is the logic between your PMS and the channel: treatment-plan segmentation, case-value cadences, backfill triggers, escalation rules. We orchestrate the tools you already pay for first, and only recommend replacement when the math says so.

What stops the AI from saying something clinical?

Scope restriction, not model trust. Outbound messages come from doctor-approved templates only. Inbound questions beyond scheduling route straight to staff. The Sinch rollback data points at data exposure and hallucination, both are answered by narrowing what the system is allowed to touch.

Get a scope and quote

One 45-minute call maps your PMS, consent status, and three largest dormant lists, then you get a quote against that scope, not a rate card. Get a scope and quote.

Related: AI automation, the parent service behind this page; paid ads when new-patient acquisition is the right spend; and AI automation for med spas if you run an aesthetics line.