PRD — AI Marketing Engine for Dental Practices (Pilot)
Overview
Status: Initial draft for review · Owner: Josh (build) / Jim (business) · Date: 2026-08-07 Objective: Launch a paid pilot with ~10 dentists from Dr. Zahedi's implant-investor group. Revenue first; everything else is sequenced behind that.
1. Summary
An AI-powered marketing service for dental practices, sold as a low-priced introductory subscription. Given a practice name, the system can:
- Research — comprehensive market research around the practice's location.
- Create — landing pages, ad copy, and creative in whatever format the target ad platform requires.
- Run — independently launch and manage advertising campaigns.
- Optimize — monitor live campaigns and propose intelligent adjustments based on metrics.
Layered on top is the GEO/AI-visibility story (practices invisible in ChatGPT/Perplexity/Google AI answers) — this is the sales hook and the differentiator, even though the day-to-day engine is research → creative → campaigns → optimization.
One operator serves many practices because AI does the volume work; humans approve anything that publishes or spends money.
2. Strategic context
- This is the marketing arm of the CareFirst Phase 1 plan — the quickest path to real revenue and the trust-building engine with the dental network. It deliberately stays on the marketing-services side of the line: no patient PHI, no BAAs, no Care Wallet dependencies. Nothing here conflicts with the long-term records vision; it funds it and builds the provider relationships it needs. - Cross-checked against the MVP business plan: the flat-subscription model is exactly the compliant safe harbor that document demands (fair-market-value marketing services, never per-patient fees), so this pivot is the legally cleanest of the options on the table.
3. Market & customers
- Total market: ~500 Chinese-American dentists across North America. Typical profile: outdated website, weak local SEO, no AI-search presence, poor Google Maps ranking. Underserved by mainstream agencies; Jim has a warm network advantage via the weekly study group he has run since the pandemic.
- Pilot cohort: ~10 dentists being recruited by Dr. Zahedi as investors in his new implant company. They buy implants at reduced rates and want implant case volume — which means their marketing goal and the implant company's revenue goal are the same goal: more implant patients. Campaigns should be implant-led. - Buyer psychology: skeptical and price-sensitive. They will not pay agency prices ($1.5k–5k/mo) on faith. The offer must be cheap, concrete, and demonstrable on their own phone.
4. The offer & pricing
Model: flat monthly subscription, paid directly by each dentist. Never per-patient, never per-case, never tied to implant purchase volume (keeps it clean under CA B&P §650 / anti-kickback framing, and independent of the Zahedi investment deal).
Pricing direction (open — see §11): as low as possible while still valuable. A drastically reduced feature set is acceptable to hit the price point. Working hypothesis for discussion:
| Intro tier (pilot) | Future tiers (not now) | |
|---|---|---|
| Price | ~$299–499/mo, 3-month pilot commitment | $1k+/mo full service |
| Ad spend | Dentist's own ad accounts; spend billed to them directly by the platform | Managed media budgets |
| Includes | AI-visibility audit + baseline, technical GEO fixes, 1 implant-focused campaign (landing page + ads) on one platform, monthly optimization + report | Multi-platform, bilingual content program, Chinese platforms, review engine, front-desk/recall automation |
Key structural choice: ad spend never flows through us. The dentist connects their own Google Ads account; we manage it. This keeps our price low, avoids fronting cash, and keeps us a service provider rather than a media buyer.
Zahedi: not part of the go-to-market — Jim is pitching the dentists directly rather than bundling with Dr. Zahedi's investment meetings. Pricing/invoicing is independent of implant purchases regardless.
5. Product requirements
F1 — Market research engine
Input: practice name (+ address). Output: a research dossier —
- Local market: demographics, language mix, competing practices within radius, their ratings/review volume/services.
- Search landscape: what patients in this area actually ask (implant-focused query set, English + Chinese), keyword cost estimates.
- AI-visibility audit: run real queries against ChatGPT, Perplexity, and Google AI for "best implant dentist near [location]"-class questions; record who gets cited and whether the practice appears. Must be real captured results — fabricated audit data is banned, full stop.
- Baseline snapshot stored so month-3 reporting can show movement.
F2 — Creative generation
- Landing pages: fast, mobile-first, implant-focused, per-practice (template
- practice config → deployed page). Bilingual where the market warrants. - Ad copy/creative in the format the target platform requires (Google Search/Performance Max first; Meta later; WeChat/Xiaohongshu manual-first).
- Answer-first GEO content targeting real patient questions, plus technical fixes (schema markup, AI-crawler permissions, Bing Places, Google Business Profile hygiene). - Nothing publishes without human approval (see §6).
F3 — Campaign execution
- Connect to the dentist's ad account (Google Ads via MCC manager account); create campaigns, ad groups, ads, budgets from approved creative. - Budget caps set by the dentist at onboarding; the system can never raise a budget on its own.
F4 — Monitoring & optimization
- Pull campaign metrics daily; detect problems (spend without conversions, rising CPC, disapproved ads). - Propose adjustments with reasoning ("pause keyword X, shift $Y to ad group Z — here's why"); operator approves in one click; approved changes execute automatically. - Track AI-visibility over time (re-run the audit query set monthly) — this is how the 2–4 month GEO promise gets proven.
F5 — Reporting
- Monthly per-practice report: spend, leads/calls, visibility movement vs. baseline, what we did, what's next. Simple PDF/email; no client dashboard in the pilot.
Explicitly out of pilot scope
Patient-facing automation (front-desk conversion, recall/follow-up — touches PHI, requires BAAs; sequence later as an upsell tier), PMS integrations, payments automation (invoice manually), client self-serve dashboard, automated Chinese-platform posting, anything under the contested CareFirst/Care Wallet brands.
6. Autonomy & safety grading
Every system action has a grade:
| Grade | Action class | Examples |
|---|---|---|
| Auto | Safe, reversible, internal | Research, audits, metric pulls, drafts, report generation |
| Approve | Publishes or spends | Launching/editing campaigns, deploying landing pages, budget shifts, GBP edits |
| Human-written | Medical claims, patient-facing clinical content | Anything asserting treatment outcomes, safety, suitability |
Operator (pilot): Josh — monitors agent output and approves day-to-day actions. Spend-related approvals additionally require sign-off from someone affiliated with the practice (dentist or office manager); exact per-practice contact TBD.
Non-negotiable honesty rules (product requirements, not aspirations): GEO results take 2–4 months and pricing/expectations say so; never promise AI engines will recommend a practice; no fabricated testimonials; no fabricated audit data.
7. Sales motion
- The own-phone demo: the dentist asks ChatGPT/Perplexity "best implant dentist near me" on their phone and watches competitors appear while they don't. 2. The free audit as lead magnet: the F1 audit report, generated per-prospect before the meeting, with real screenshots. 3. Channels: Jim's direct outreach to the dentists + his weekly study group (not bundled with Dr. Zahedi's investor recruitment meetings). 4. Close: intro-tier subscription, 3-month pilot commitment, expectations letter stating the 2–4 month horizon and what is and isn't promised.
8. Compliance guardrails
- Flat FMV fees only; independence from implant volume documented in the service agreement. - Dental advertising rules apply to the content we generate (CA and other states restrict superlatives, guarantee language, testimonial handling) — the human-written grade covers this. - No PHI: lead handoff stops at "prospective patient contact info goes to the practice." If a form collects health details later, that's the trigger to revisit BAAs. - Pilot runs under the brand Alonvia; do not create "use in commerce" on the contested CareFirst/Care Wallet marks via paid ads.
9. Pilot success metrics
- ≥7 of 10 pitched dentists sign at the intro price (revenue = the metric; ~$3–5k MRR at hypothesis pricing). - Every client: audit + baseline + landing page + one live campaign within 30 days of signing. - By day 90: measurable AI-visibility movement for ≥half of clients; per-client lead counts reported monthly; ≥70% renew past the pilot commitment. - Operator load: one person can run all 10 accounts in <10 hrs/week (proves the leverage thesis).
10. Engineering: what Josh can start now
Build-now (invariant under every open decision):
- Audit engine (F1) — first, this week. It's the sales weapon for the pitch meetings and the baseline for reporting. Real queries, captured citations, competitor comparison, clean one-page output. Highest leverage per hour of work.
- Practice-config schema. One structured record per practice (name, locations, languages, procedures, competitors, query set, ad account refs, budget caps). Everything else keys off it.
- Google Ads API lead-time items. Apply for a developer token and set up the MCC manager account now — approval takes time and blocks F3 regardless of any business decision.
- Approval-queue harness. The draft → human-approve → execute spine (a simple queue UI is fine). Every publishing feature depends on it; it also embodies the safety model.
- Landing-page template + deploy pipeline. Config-driven, brand-neutral, one implant-focused template.
- Reporting/baseline store. Persist audit snapshots and campaign metrics from day one so month-3 proof exists.
Decision-gated (don't build yet): billing automation (pricing open), Meta/WeChat/Xiaohongshu integrations (platform priority open), client dashboard, review-management tooling, anything patient-facing, any branded public site (brand: Alonvia).
11. Open questions
Blocking-ish (needed before/at the weekend pitch):
- Exact intro price and commitment term. Hypothesis: $299–499/mo, 3-month commitment, no setup fee. Pending sign-off with Jim.
- First ad platform. Recommendation: Google (Search + Maps/GBP) only for the pilot. Pending sign-off with Jim.
Important, not blocking:
- Geography of the 10 dentists — confirmed multi-state (not CA-only); exact state distribution still TBD. (Affects ad targeting, state ad rules, and whether "local research" is 10 separate markets.)
- Bilingual requirements — confirmed P0 across the pilot; exact % of practices / which markets need Chinese-language pages/ads still TBD.
- Monthly ad-spend expectation per dentist — still open; holding off on an estimate until there's a read on interest levels (implant CPCs run $15–50+).
- Contract paper: leaning toward drafting the service agreement + expectations letter with AI rather than outside counsel. Open: how pressing is a counsel review before signatures, given the FMV/anti-kickback framing in §8?
- What happens to leads — call tracking? Forwarding number? Needs more design work before deciding.
12. Resolved since initial draft (2026-08-07)
- Brand name: Alonvia (was open question 1).
- Campaign focus: all campaigns implant-led, confirmed (was open question 5).
- Zahedi bundle: not a consideration — Jim is pitching dentists directly, no bundle/co-brand/cut (was open question 8).
- Pilot operator: Josh, monitoring agent output; spend approvals also need sign-off from someone affiliated with the practice (was open question 7).
Source context: CareFirst-MVP-Business-Plan.md (this folder). That document's blocking conditions were written for the Care Wallet MVP; the ones that carry over here — flat fees, neutral branding, honest promises, staying behind the PHI line — are already incorporated above. Its other concerns do not block this pilot.