SMS cold chain alert for a biologic vaccine — the alert does not wait for the cloud.

The problem

The thermograph records everything — and the warning arrives when it is no longer useful.

The classic cold chain monitoring flow for a biologic vaccine works reasonably well — until it fails on the day that matters:

1. The datalogger uploads the reading to the vendor's cloud. The cloud generates the alert. The alert lands in the portal and in an email.
2. But the truck is on a motorway with no stable 4G, or the depot has unstable WiFi, or the vendor's cloud had an incident that morning. The SMS never arrives because the chain depends on steps that all have to be working at the same time.
3. And the cold chain manager checks the portal three hours later, when the cumulative time out of range is already unrecoverable. The batch is discarded — and in biologics that is hard cost and a delicate matter at once.

The problem is not a lack of technology — thermographs are good. The problem is that the path from the reading to the manager's eye depends on too many external pieces. The alert should start as close to the thermometer as possible.

How it fits into the IRIS system

iLEAN Connect does not replace the datalogger — it carries it to SMS without waiting for the cloud.

iLEAN Connect's graduated capture also applies to transport and to the depot: a local terminal that reads the thermograph by whatever route fits (Bluetooth with the manufacturer's datalogger, NFC, serial reader for the old chart recorder) and fires the warning through whichever channel arrives first — SMS, push, email — without depending on the vendor's cloud. iLEAN Connect is the filler that covers the gap between what the datalogger already knows and what the QP needs in order to decide, without asking you to change datalogger or logistics operator.

Connect carries the alert by SMS from the edge, the agent cross-checks it against the product data sheet, the QP signs the decision on the batch.

The three iLEAN pieces applied to the cold chain SMS alert:

• Connect (local capture + outbound channels) — a terminal in the truck or in the depot that reads the thermograph and fires the SMS through the GSM gateway, the push through the QP's app, the email through SMTP — in parallel. Multichannel by design: never a single path. If the Internet goes down, the SMS still goes out. If GSM is down too, the local log persists and is re-injected when the network returns.
• Agent — cross-checks the thermograph reading against the product data sheet (acceptable range, cumulative time, acceptance criterion specific to the vaccine), applies the rules signed by the QP (preliminary alert to the carrier at the first crossing, firm alert to the QP if the deviation persists, escalation to management if it exceeds the limit) and keeps the record alive.
• Central + Ring 2 — the extracted data passes through the validation chamber before it touches the quality system. Exportable GDP-ready record: batch, carrier, datalogger, time of first deviation, time of alert, person notified, response time, decision signed by the QP. Audit-ready with no manual reconstruction.

See the full IRIS architecture →

Before and after

Cloud-based alert vs. SMS alert from the edge

Aspect	Classic alert (cloud + portal)	With iLEAN Connect + edge SMS
Latency to the carrier	Depends on the cloud → email → app chain	SMS from the edge within the first minute
Operation without stable 4G	Fails — no warning at all	SMS goes out over GSM 2G
Operation if the vendor's cloud goes down	Fails — depends on the datalogger SLA	The edge keeps alerting locally
Parallel channels	Email + push (same path)	SMS + push + email in parallel
GDP record	Reconstructed from the portal	Generated at second zero, signed
Decision on the batch	Manual, with no product context	Cross-checked against the data sheet + QP signs

Impact estimate

Impact estimate for your plant — to be validated with your numbers.

The block below is an estimate to be validated with the specific data of your operation. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• National or regional distribution of a biologic vaccine, dataloggers from one or several vendors, own fleet plus external carriers, peripheral depots with variable connectivity.
• Connect pilot at the edge on one critical node (terminal + datalogger reader + SMS gateway + routing agent + integration with your quality system). First value expected within a few weeks: reduction of the average time between deviation and alert to the QP by ≥ 30%.
• Indicative payback between 4 and 9 months, depending on the frequency of documented deviations, the average value of the affected biologic batch and the percentage of batches discarded because of a late warning.
• Hard lever: a single biologic vaccine batch saved (an informed decision in the first minute instead of in the third hour) pays for the pilot. And it lowers the regulatory cost of the deviation record.

And the Qualified Person's fair objection

"What if the AI decides badly and lets a critical deviation through, or floods us with false alerts?" — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely reads a thermograph value and compares it against a threshold signed by your team, the best models brought the error below 1.5% ^[1]. And even so, the rules are signed by the QP — they are not ours. The decision on the batch is signed by a person; Connect leaves the alert and the evidence.

^[1] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.