HUMAN-GATED AUTOMATION

Your body is already sending signals. BioFlow acts on them, after you say yes.

Wearables read what is happening to a person. BioFlow decides what should happen next, then stops and holds that action at a human approval gate until someone releases it.

6 signal classes<120 ms to evaluation72 h with no uplinkSigned decision ledger
BIOFLOW · WARD 3BF-PATCH-00418
Approvals
1 waiting on you
APPROVAL REQUIRED0:20
Send the last 30 s of ECG to the on-call cardiologist?
Irregular rhythm across 4 consecutive beats. Confidence 0.91. No motion artefact.
Baseline learned for this wearerauto
HR118bpmRHYTHM FLAGGED · SENT TO GATE
LEAD II · 250 HzHR 72 bpmSpO₂ 98%GSR 2.1 µSwatching

Illustrative trace. In a live deployment this strip is the wearer's own signal, and the rules watching it are the ones you wrote.

THE PROBLEM

Automation platforms connect apps. BioFlow connects a person.

Rule engines already fire when an email arrives or a form is filled. None of them can read that a night-shift technician's heart rate has been elevated for forty minutes, or that a patient's oxygen saturation is drifting while they sleep. And none of them are built to pause and ask.

WHEN heart_rate.variability drops below baseline − 2σ for 4 min AND motion.artefact = false AND context.location = "ward-3" ASK the on-call nurse, expires in 20 s THEN page cardiology · attach 30 s ECG · open a chart note

Three of those clauses are ordinary automation. The third one (ASK) is the product. It is a first-class part of every rule, not a compliance checkbox bolted on afterwards.

WHAT IT CONNECTS

Wearables and phones on one side. The systems you already run on the other.

Reads from

ECG patch250 Hz · 10 days
WristbandPPG · IMU · GSR
Smart ringbaselines
Tactical patch72 h · DDIL
iOS and Android companion app for approvals
APPROVAL GATEA person decidestrust × urgency × reversibility

Acts on

Clinical recordFHIR R4
Plant systemsOPC-UA
Paging & commswebhook · SIP
TicketingREST
Every dispatch written to the signed ledger
THE PIPELINE

From a heartbeat to an action, in five stages

STAGE 1

Sense

Sensors sample continuously. Filtering and artefact rejection happen on the device, so a loose electrode never becomes an alarm.

≤ 4 ms per window
STAGE 2

Interpret

Signals from several devices are fused with context (location, shift, activity) into a state rather than a number.

≤ 30 ms
STAGE 3

Propose

The rule graph produces a candidate action with a confidence, an urgency and a written reason for it.

≤ 80 ms
STAGE 4

Approve

A person confirms, refuses, or lets the window expire. What expiry means is set per rule, in advance, by name.

human time: 5 s to 5 min
STAGE 5

Act

The workflow runs to completion, retries safely, and writes a signed record of who released it and why.

durable · exactly once
Read the full pipeline
DEVICES

Certified hardware, or bring your own

BioFlow is not a hardware company. It certifies a reference set of six wearables and adopts anything else that can present the same signal contract and attest its firmware to the gateway.

Compare devices
$275Bcombined addressable market across wearables, HITL AI and remote monitoring
24.9%annual growth in human-in-the-loop AI
4.1 smedian time from proposal to a human answer
100%of executed actions carry a signed audit record

Start with one ward, one shift, or one squad

A pilot runs four to sixteen weeks depending on deployment model. You bring the population and the responders. We bring devices, gateways and the first set of rules, and we measure override rate and time-to-answer from day one.

Request a pilotSee how it works