Your medical cost
doesn’t start at the claim.
It starts months earlier, in a visit where the diagnosis didn’t arrive on time. That’s where we work, with the physicians already in your network.
65.2% of what a Mexican insurer pays out is hospitalisation, not consultation.AMIS · 2024
Your medical cost doesn’t start at the claim. It starts at the visit.
Every link below comes from an independent source. Chained together, they explain why the spend keeps growing and where it can be interrupted without touching your network or your systems.
- $122,278M
pesos were paid out by Mexican insurers in medical claims during 2024, a figure that grew 106% in five years.
AMIS, 2024 - 65.2%
of that payout is hospitalisation, not consultation. The money isn’t where everyone looks: it’s in the episode that escalated.
AMIS, 2024 - 23%
of patients who escalated to intensive care or died had a missed or delayed diagnosis. Escalation has a clinical cause, not an actuarial one.
JAMA, 2024 - 55%
is diagnostic accuracy by the sixth hour of a shift. The diagnosis arrives late because the physician tires and because the history isn’t in the room.
Our own measurement · El Salvador
That’s where DeepClinic works: it holds accuracy at 96.1% inside the provider’s visit, with the patient’s full history in front of them. And it is priced against avoided cost, measured in the same record that produced it.
Accuracy is our own measurement over 2 million consultations in El Salvador, against the historical baseline of that same system. The effect on admissions is not yet measured in our data: it is exactly what we propose measuring on your book before any contract.
The cost is decided
at four moments.
Triage
Classifies reason and urgency before the appointment.
Clinical copilot
The full history in front of the physician, and the note writes itself.
Follow-up
Reminds about medication and tests between visits.
Coding
ICD-10 with justification cited from the note itself.
What you see today as loose claims, here reads as episodes.
- Cost per episode
- Escalation rate
- Variability across physicians
- Protocol adherence
- Out-of-pattern cases
- Duplicate testing
All of it on the providers you already contract. No implementation on your side.
Triage decides where the patient goes.
We decide what happens once they’re there.
We don’t compete: we start where triage ends.
That payers buy this category is already proven: Infermedica reports 27% fewer in-person visits and $7–8 saved per interaction in Latin America. Our unit isn’t the interaction — it’s the episode. Admission figure: market average cost.
Move the three factors. What comes out is your avoidable cost.
The arithmetic has three steps and none of them are ours: what you pay out, the share that goes to hospitalisation, and the share of that hospitalisation which starts with a diagnosis that arrived late.
The Mexican market aggregate was $122,278M in 2024 (AMIS).
65.2% is the Mexican market average (AMIS, 2024).
This is the hypothesis. JAMA (2024) found a missed or delayed diagnosis in 23% of patients who escalated to intensive care or died; the default below is deliberately conservative against that figure.
This is a hypothesis, not a promise. The effect on admissions is not yet measured in our data: the only number that counts is yours, and it comes from running a year of your claims against your own baseline. That analysis costs nothing and commits you to nothing.
It documents while the physician sees the patient.
Much better, honestly. I'm sleeping well and the headaches have barely come back since we changed the dose.
Where the needle actually moves.
65% of what a Mexican insurer pays out is hospitalisation (AMIS, 2024), and hospitalisation is triggered by the diagnosis that arrives late.
The scribe is where it starts. Not where it ends.
The note is the first step. Behind it sits the whole episode: the pre-visit that arrives with history, the follow-up that doesn’t depend on anyone remembering, the charge that gets closed and the data that finally gets measured. One auditable journey, from first contact to the claim.
Before. The patient arrives ready.
They register, book and get reminders from their own digital front door. Before the visit they complete symptoms, reason for consultation and history, so the doctor arrives prepared.
During. The doctor just practices.
The clinical note is written while the physician sees the patient. Diagnosis, instructions, treatment and plan are documented and coded before the door closes, so the claim comes from the same record that documented the visit.
After. Care doesn't stop.
The plan continues with reminders, post-visit tasks, symptom surveys and alerts when the patient needs to come back. The visit is closed administratively and everything feeds the dashboards.
The clinical brain behind every module.
The modules aren't separate products: they're doors into the same model. DeepClinic One is our own clinical model, inside a layer that orchestrates patient context, evidence and the guardrails of each specialty.
Clinical reasoning
It doesn't complete text: it evaluates. Behind every suggestion sits a structured evaluation engine, not a next-word prediction.
Evidence with provenance
Every answer carries where it came from — literature, clinical guidelines and our own corpus. No source, no answer.
Longitudinal context
A single record over HL7 FHIR. What the model knows about today's visit includes every visit before it.
A connected health network. Not islands of software.
Patients move through the practice, the lab, the pharmacy and their insurer — and today every step starts from zero. DeepClinic connects that chain on a single record, so information follows the patient instead of staying behind at each door.
One record
The patient is the same at the lab, at the pharmacy and at the practice. No duplicates, no re-entering the same data.
One standard
Everything speaks HL7 FHIR. Whatever enters through any door of the network comes out structured and usable.
One audit trail
Every access and every change lands in an auditable log, whether it came from the doctor, the patient or a connected system.
One identity
The doctor authenticates once. Their permissions and scope apply the same across the whole network.
Your entire provider network, in one layer.
Agents that schedule, document, follow up and govern every visit on the system the provider already uses.
Document and code
Clinical note (SOAP) and diagnosis code (ICD-10) on the provider’s system, audit-ready.
See it in action, in a real visit.
We meet every rule so you never have to think about it.
Every note, every decision and every access lives in an auditable log. Ready for a HIPAA audit, for your medical review or for a check across your provider network, with no extra work on your side.
We test DeepClinic in real clinical scenarios so it stays accurate, reliable and clinically sound.
Our medical team reviews and refines the results. Automation alone isn't enough.
Every record stays secure, private and under your control, always.
What any claims director would ask.
Answered here instead of in the meeting, because these are the ones that decide whether this is a purchase or just another presentation.
How do I know the saving is yours and not the cycle's?
Because it isn't measured against a market average. The baseline is built from your history and your codes, and providers are activated in waves: each is its own before-and-after control, and those not yet live are the control for the period. That way we never compare the physician who adopts against the one who doesn't — they are never alike. If the effect doesn't show up, there is no saving to charge for.
Why would the provider accept this?
Because we don't add work, we remove it. The note, the diagnosis code and the prescription come out while they talk to the patient, and the claim is born from that same record, so fewer get denied. They don't change systems and install nothing. Without that incentive on the physician's side there is no adoption, and without adoption there is no saving.
Does this replace my medical audit?
No, it feeds it. Today you audit a sample and request paperwork; with DeepClinic the whole episode sits in HL7 FHIR, with the provenance of every suggestion and the signature of whoever accepted it. Your team stops reconstructing records and starts reviewing the cases that matter.
Who carries the clinical responsibility?
The physician, always. DeepClinic gathers the evidence, reasons the case and traces every suggestion; it never decides and never signs. Clinical liability does not move from where it sits today, and that is deliberate.
How long does it take and what do you need from me?
For the analysis, an export of one year of claims: no implementation, no budget. For activation, providers in your network who already see your members. Your adjudication system is never touched.
What happens to member data?
It is encrypted and isolated per provider, under Mexico's data protection law (LFPDPPP) and HIPAA when data crosses into the United States. We do not train models on one client's records to serve another, and every access lives in an auditable log.
Where every number on this page comes from.
- 01AMIS, 2024
$122,278 million pesos paid out by Mexican insurers in medical claims during 2024, up 106% in five years. 65.2% of that payout is hospitalisation.
- 02JAMA, 2024
23% of patients who escalated to intensive care or died had a missed or delayed diagnosis.
- 03WTW · Global Medical Trends, 2026
Payers’ medical cost rises 10.3% globally in 2026, the third consecutive double-digit year. Latin America accelerates from 10.5% to 11.9%.
- 04OECD
Roughly one in five dollars of health spending goes to avoidable or low-value care. The definition of avoidable hospitalisation we use is ambulatory care sensitive conditions (ACSC).
- 05Infermedica
27% reduction in in-person visits and an estimated $7–8 saving per interaction in Latin America, with digital triage. Cited as market reference, not as our own result.
- 06Our own measurement
96.1% sustained diagnostic accuracy against 55% at the sixth hour of a shift, across 2 million consultations in a country’s public health system, compared against that system’s historical baseline.
What is not yet measured: DeepClinic’s effect on the hospitalisation rate. It does not exist in our data and we do not claim it. It is precisely what we propose measuring on your book, with the methodology agreed before we start.
DeepClinic isn’t just software: it’s the expensive episode never happening. Every diagnosis that arrives on time is an admission nobody pays for and a patient who didn’t deteriorate. We’re building the AI layer healthcare runs on — from Latin America, for the world.
A better experience for clinicians, a better outcome for patients, and a healthcare system that gets more efficient, not more expensive, as it grows.
Stop paying for the episode that could have been avoided.
We start by analysing a year of your claims, with no commitment: we show you duplicate testing, repeat visits and escalation from late diagnosis across your own book.