© 2026 Scripta Health. All rights reserved.
Built on AWS Bedrock · HIPAA BAA Active · Trust & Security
Activate member · Learn about AWS Activate
Sentinel is the first AI pre-verification layer for pharmacy. It runs between data entry and pharmacist verification — catching the math errors, unit confusions, and compliance gaps that hide in plain sight.
Every Rx is a puzzle — 50 states of pharmacy law, dose ceilings, BUD limits, midlevel prescriber caps, amendment histories. Sentinel runs 68 independent checks on every prescription so your pharmacist walks into verification with the full picture, not a gut feeling.
Sentinel is.
You know the moment. You're verifying your 200th Rx of the day. A tech just handed you three more. The patient on line 2 wants to know why her refill isn't ready. You're trying to remember whether gabapentin is Schedule V in Kentucky. And somewhere in this stack is a decimal slip on a testosterone vial that could put someone in the hospital.
You can't be perfect 24/7. Nobody can. That's not a failure of training or attention — it's math. Your brain wasn't built to hold federal DEA rules, your state's pharmacy regulations, BUD compliance tables, FDA-approved dose ceilings, every drug interaction matrix, every sound-alike/look-alike pitfall, and every tech's data entry — all at once, all day, every day.
Sentinel is.
It reads every prescription the moment it's entered. Deterministic math catches the things math should catch — refill limits, dose math, quantity versus days supply. AI catches the things that need reasoning — dose plausibility, state-specific controlled substance rules, off-label prescribing, self-prescribing red flags. The flags appear in the fill note where you can't miss them. You read. You decide. You verify — or you don't. Sentinel never clicks Save. That's always yours.
The result isn't "AI replacing pharmacists." It's pharmacists finally getting to use the part of their brain they went to pharmacy school for — the part that makes clinical judgment calls. Everything else runs on autopilot. Your error rate drops toward zero. Your workday gets shorter. Your job becomes what it was supposed to be when you chose this profession.
Your mental bandwidth back. Your error rate near zero. Your shift ends on time.
Fewer near-misses. Lower liability exposure. Happier staff. Higher throughput.
A safety net for your outgoing prescriptions and the pharmacy relationships you depend on.
These aren't rare edge cases. They're the everyday mistakes that slip through busy workflows and reach the pharmacist's screen looking perfectly normal.
Decimal slips that turn 0.1 ml into 1 ml — a 10x testosterone overdose. Levothyroxine 50 mg instead of 50 mcg — a 1000x thyroid overdose. The errors that put people in the hospital.
Semaglutide above the FDA-approved 2.4 mg/week ceiling. Tirzepatide prescribed to a 16-year-old. GLP-1s used outside their approved indication and age range — without explicit physician justification.
Physicians self-prescribing controlled substances. NY HCG with the wrong expiration window. NY testosterone past 30 days. C-II refills entered against federal DEA rules. Plus the small stuff: days supply math, missing sig route.
Testosterone cypionate 200 mg/ml. Prescribed volume: 0.1 ml per dose. Under time pressure, the decimal slips and the entry reads 1 ml. That's 10x the prescribed dose — and it looks perfectly normal on screen.
10x overdose dispensed. The tech enters the volume as 1 ml instead of 0.1 ml. That's 200 mg per injection instead of 20 mg. The pharmacist is verifying 40 scripts in a row. The entry looks like a normal testosterone Rx. It gets verified. The patient injects 10x their prescribed dose of a hormone that can cause permanent voice changes, excess hair growth, and significant hormonal side effects.
Caught in 2 seconds. Sentinel flags the entry before the pharmacist even opens it: "Volume 1 ml at 200 mg/ml = 200 mg per dose. This is 10x a standard HRT starting dose. Verify volume with prescriber." The pharmacist sees the warning, catches the decimal error, fixes it to 0.1 ml. Patient is safe.
Sentinel doesn't rely on a single check. It stacks three independent layers — deterministic rules, AI clinical judgment, and your pharmacist's final call.
Hard math and compliance checks that run in milliseconds — days supply math, BUD caps, state-specific controlled substance laws, prescriber restrictions, typed-vs-source field comparison, and more. No AI needed, just rules that never miss.
Frontier AI evaluates what rules alone can't catch. Dose plausibility, sig coherence, concentration math, unit confusion, and clinical context that requires reasoning — not just pattern matching.
Your pharmacist always has the last word. Sentinel surfaces every finding as a flag — it never clicks Verify, never overrides clinical judgment, and never removes the human from the loop. It's a notes bot, not a gate.
Sentinel is designed to add a safety layer without introducing new risk. These are architectural commitments, not just policies.
Never touches Verify. Sentinel cannot approve or reject a prescription. That authority belongs to your pharmacist, always.
Read-only on clinical fields. Sentinel reads data entry fields but never modifies them. It can only append flags.
No patient data leaves your pharmacy. Patient name, DOB, address, and insurance stay on your machine. Only clinical fields are sent for analysis.
No workflow disruption. Sentinel runs in the background. Your team's existing process doesn't change — it just gets safer.
HIPAA-compliant infrastructure. Built on AWS Bedrock with an active Business Associate Agreement. Encrypted end-to-end.
Instant rollback. If anything goes wrong, Sentinel can be disabled immediately with no impact on your pharmacy's operations.
A single Sentinel bot checks one prescription in under 10 seconds — including screen reading, compliance checks, AI analysis, and note-writing. Need more throughput? Start another bot.
More than enough for most independent pharmacies
High-volume compounding or busy retail
Enterprise or multi-location operations
Conservative estimate based on ~10 seconds per Rx end-to-end. Measured AI latency averages ~5 seconds — real-world throughput may be higher. Bots run independently and scale linearly.
Sentinel catches the errors that hide between data entry and verification — the ones that matter most. Scribe writes. Sentinel watches. You verify.
Request Early Access© 2026 Scripta Health. All rights reserved.
Built on AWS Bedrock · HIPAA BAA Active · Trust & Security
Activate member · Learn about AWS Activate