Recall and no-show, handled before the front desk has to.
Two of the highest-ROI levers in any practice, running as sequences that talk to patients in your voice and update your PMS or EHR when they reply.
See details →Healthcare
AI workflows for independent practices and clinics.
BAAs, audit logging, PHI handling, PMS or EHR integration. These are part of the build from day one, not week three. The team spent years inside patient communication and clinical ops, so the compliance vocabulary is muscle memory and the conversation with your IT or compliance officer goes fast.
What this means in practice.
Three things a non-healthcare consultant usually doesn't do, and that we treat as standard:
Sign a BAA before kickoff.
The Business Associate Agreement gets signed before discovery generates any artifact that could touch PHI. Downstream vendor BAAs (Twilio, LLM provider, audit log storage) get set up during discovery, not promised in a sales call and chased after launch.
Strip identifiers before the LLM call.
For most workflows, the LLM never sees a patient identifier. Where PHI-to-LLM is genuinely necessary, the data flow gets documented, reviewed by your compliance officer, and limited to BAA-covered endpoints. Written down, not waved at on a slide.
Log every contact for two years.
Every patient contact attempt. Every response. Every status change in your PMS or EHR. Every LLM call, with prompt and response stored and redacted if PHI is present. Two-year retention by default. When an auditor asks, the log is right there.
Who this is for.
Independent practices and small clinics.
Two to ten providers. One office manager, one or two front-desk staff per location. You can name the workflows bleeding revenue (recall, no-shows, intake); adding another headcount isn't the right answer to any of them.
Specialty practices growing into systems.
Dental, dermatology, orthopedics, PT, mental health, ophthalmology. The PMS or EHR is doing what it does well. The gap is everything happening around it.
Practice-owned, not PE-owned.
Decisions happen in days, not quarters. You can sign a BAA without routing it through three layers of legal. Off-boarding documentation matters because you might sell the practice yourself someday, not because corporate has a policy.
The healthcare builds.
Each one is a fixed-price engagement with a named outcome. Click through for what's in, what's out, and how the delivery actually runs.
Eight pages of intake, one page of clinical signal.
A new patient submits the form. Ten minutes later your clinical team has a one-page summary in the shape they actually use, with red-flag items called out at the top.
See details →The fax tray stops being a risk.
Faxes and PDFs land in your inbound channel and get read, classified, and routed to the right provider in the right EHR location. Urgent items get flagged at the top.
See details →The morning briefing the practice manager actually reads.
One screen at 6am, tuned to the calls a practice manager has to make today. Schedule density, no-show risk by provider, AR aging, anything trending wrong.
See details →What we won't sell you.
- Clinical decision support. Anything that influences a treatment decision needs FDA or CE pathways and clinical evidence we are not the right people to produce. The work here is communication, operations, and admin. The clinical call stays with the clinician.
- An LLM-only "AI receptionist." A bare LLM at the front line of patient communication, with no deterministic rule layer underneath it, is a compliance risk and a reputation risk. We use LLMs for the parts they're actually good at and keep rules where they belong.
- A rip-and-replace of your PMS or EHR. Your PMS or EHR is doing its job. The builds sit on top of it via API and write back into it. When a vendor tells you the answer is to switch systems, that's usually the vendor's revenue need, not yours.
- Vague "AI strategy" engagements. If we can't describe the outcome in one sentence and bound the price, we won't sell it. Strategy without delivery is how a lot of healthcare AI money has disappeared in the last two years.
- Anything that puts PHI in front of an LLM without a BAA. No workarounds, no "just for now," no exceptions. If a workflow needs PHI in the prompt and no BAA-covered endpoint can handle it, the workflow doesn't ship.
BAA, compliance, and the boring questions.
Do you sign a BAA?
Yes, on every healthcare engagement. BAAs with the downstream vendors (Twilio, LLM provider, audit log storage) are set up as part of the deployment, not promised for later.
What vendors are in the data flow?
Documented in a one-page diagram you get before go-live. Typical stack: your PMS or EHR via API, n8n (self-hostable), Twilio with a BAA, Claude under Anthropic's enterprise BAA or via AWS Bedrock, and an audit log database.
Do you send any PHI to LLMs?
Only with your compliance officer's explicit sign-off on the specific flow, and only to BAA-covered endpoints. For most workflows the answer is no; patient identifiers get stripped before any LLM call.
What happens if our compliance officer says no to part of the architecture?
We adjust the design. Every healthcare build assumes one or two architectural revisions during discovery. Your compliance officer reviews before the contract gets signed, not after the build ships.
How is access controlled after handoff?
Role-based access in n8n. Secrets in your chosen vault (1Password, AWS Secrets Manager, whatever you already run). Audit log retention for two years minimum. Off-boarding documentation included so the system is yours, not ours.
What's logged for audit purposes?
Every patient contact attempt, every response, every status change in your PMS or EHR, every LLM call (prompt and response, redacted if PHI is present). Two-year retention by default; longer on request.
Two ways to start.
Take the 5-minute audit to see which build fits. Or skip straight to a discovery call if you already know.