Responsible AI
Why Human Approval Matters in AI-Assisted Medication Review
A drafting tool is only as trustworthy as the review gate that stands between it and the record.
6 min readTransparent AI support for structured pharmacist medication reviews.
Rediscover Dosage is a human-in-the-loop workspace for statin-related case intake, cited guideline support, explainable review summaries, uncertainty handling and editable professional documentation.
A synthetic, demonstration-grade product in development.
Product principles
Organized review, grounded guidance, human control.
- 01
Structured, not free-text
Case context arrives as organized fields the pharmacist can scan quickly, rather than fragments spread across notes.
- 02
Source-grounded guidance
Guidance is tied to approved source categories with citation references the pharmacist can inspect before relying on it.
- 03
Human approval, always
Drafts stay editable. Nothing is finalized until a pharmacist takes a separate, deliberate approval step.
Inside the workspace
A structured review workspace, built for pharmacist judgment.
Switch between the structured case view, the explanation view and the human approval step.
Review workspace
RD-2481 · Statin-related review
Review objective
Organize a structured review of a statin-related medication case.
Current medication
Statin therapy recorded in the intake record.
Tolerance notes
No tolerance concerns recorded at intake.
Canadian dyslipidemia guidanceOntario medication-review material
Contributing factors
- Structured fields completed
- Recorded tolerance notes
- Review history depth
Each factor links back to the source material and the recorded case fields behind it.
Pharmacist approval
The AI-assisted draft stays fully editable. Approval is a separate, deliberate step taken by the reviewing pharmacist before a report is produced.
Apply pharmacist approval
How the workflow runs
Seven steps, with the pharmacist in control of the outcome.
Each step is designed so the reviewing pharmacist sees the case, the sources, the reasoning view and the draft before anything is approved.
Step 1: Open a review case
Select a case from the platform's case library.
Step 2: Review structured information
Read the structured case fields organized for pharmacist review.
Step 3: Identify missing information
Surface fields the workflow marks as absent before any review summary is prepared.
Step 4: Review source-linked guidance
Open the guideline copilot panel and inspect the cited source material.
Step 5: Examine explanation and uncertainty
Review the factor view and the confidence or abstention state.
Step 6: Edit the AI-assisted review draft
The draft remains fully editable by the reviewing pharmacist.
Step 7: Apply human approval and preview the report
Approval is a separate, deliberate step before the report preview is produced.
Capability pillars
Three commitments that shape every screen.
Source-grounded guidance
Guidance is represented through approved source categories with citation references the reviewing pharmacist can inspect before relying on it.
Explainability and abstention
A factor view shows what shaped the synthetic output, and an explicit abstention path applies when information is insufficient.
Human approval
AI assistance does not replace professional judgment. Approval is a separate, deliberate step taken by the reviewing pharmacist.
Synthetic demonstration cases
Five scenarios shaping the review workspace.
New statin-review case
Organize a first structured review of a newly opened statin-related case.
Inadequate lipid-control review
Structure a review where the case record describes control as not yet at the documented target context.
Possible statin-intolerance / SAMS review
Structure a review where the case record contains reported tolerance concerns for pharmacist attention.
Medication-review / polypharmacy context
Structure a broader medication review where the case record contains an extended co-medication list.
Incomplete-data / abstention case
Illustrate the abstention path when a case record does not contain enough structured information.
Synthetic demonstration case — not patient data.
Technology and responsible AI
A six-layer architecture designed around human review.
The architecture below describes how structured case data, approved sources, AI-assisted drafting, explainability, human review and governance fit together.
Structured Case Layer
Neutral case fields organized for review rather than free-text notes.
Approved Source and Citation Layer
A controlled set of approved source categories with citation references attached to guidance.
AI-Assisted Draft Layer
Draft review language prepared for pharmacist editing, always labelled as an AI-assisted draft.
Explainability and Abstention Layer
Factor views and an explicit abstention path when information is insufficient.
Human Review and Approval Layer
A separate approval control operated by the reviewing pharmacist.
Governance and Audit Layer
Audit history, model card, data card and feature-status registry.
From the team
Insights
Product notes, responsible-AI thinking and workflow writing from the team building this workspace.
Responsible AI
What Makes a Guideline Copilot Source-Grounded?
Grounding is a workflow discipline, not a single model feature — it shows up in how citations are stored, displayed and checked.
7 min readResponsible AI
Designing an Honest Abstention State
Sometimes the most responsible output a review tool can produce is a clear statement that it doesn't have enough to say anything.
6 min read
Company and leadership
Built by a founding team spanning pharmacy and engineering.
Dr. Abolfazl Esmailzadeh Ghandehary
CEO
Esmaeil Faraji
COO
Mohammad Mehdi Daevaeiha
CTO
Sahar Akhavan
CSMO
Review the product direction, then walk the synthetic workflow.
The transparency page records intended use, limitations and current validation state before anything else is shown.
Transparency summary
- All cases and outputs shown here are synthetic, not patient data.
- No clinical or Canadian validation is claimed for this build.
- Every capability carries an explicit build-status label.