Technology & Governance

Evidence, assurance and academic governance

How AccuRXy grounds every AI-generated pharmacy scenario in approved, jurisdiction-specific evidence, validates it against clinical rules and keeps a full academic audit trail.

Architecture
Evidence Pack Specification v1.0
Source inventory
Ireland jurisdiction v0.1
Jurisdiction
Republic of Ireland
Status
Working draft, 25 June 2026

01. The principle

The AI is one component, not the authority

AccuRXy does not simply ask a language model to generate a scenario and hope for the best. The reference sources are not cited after the fact; they are retrieved first and used to guide and constrain what the model produces.

The AI model is one component of AccuRXy, not the authority.

AccuRXy owns the evidence, governs what is authoritative, determines the jurisdiction and maintains the academic audit trail. The AI generates content within a governed framework; it does not determine what is clinically correct.

Seven layers of AI governance

01

User experience

Scenario Design Lab, AI Tutor, Scenario Library, Academic Review

02

Scenario orchestration engine

Interprets requests, retrieves evidence, controls generation, saves results

03

Evidence retrieval layer

Search and filter, ranking and matching, jurisdiction filtering

04

Model gateway

Provider-neutral interface to GPT, Claude and Gemini

05

Knowledge and clinical rules

Documents, medicines data, protocols, taxonomies, rules

06

Validation and academic governance

Citations, consistency checks, review workflow, audit trail

07

Integration interfaces

Internal API, Partner API, FHIR where useful, MCP gateway

02. The governed workflow

How a scenario is built

Every request begins with a structured scenario specification rather than a free-text prompt, and passes through the same orchestration engine from evidence retrieval to academic review.

The structured specification collects

Learner levelJurisdictionPharmacy settingClinical topic and medicine or drug classIntended medication error typeDifficulty and learning objectivesPatient characteristics and assessment formatFormative, summative or exploratory

The engine never hands the specification directly to an AI model. It follows a governed workflow.

  1. 01

    Receive scenario specification

    The structured parameters define what kind of scenario is needed.

  2. 02

    Determine required evidence categories

    The engine identifies which source documents, medicine records and clinical rules are relevant.

  3. 03

    Retrieve approved evidence

    The Evidence Retrieval Layer fetches jurisdiction-appropriate, current, approved materials.

  4. 04

    Construct an evidence pack

    Relevant passages, structured medicine records and clinical rules are assembled into a package.

  5. 05

    Generate scenario skeleton

    The AI model produces a structured scenario using the evidence pack as its source material.

  6. 06

    Generate dialogue and assessment content

    Full content is produced within the constraints of the evidence pack.

  7. 07

    Run medicine and consistency checks

    Generated content is validated against structured medication data and internal consistency rules.

  8. 08

    Attach citations

    Every material clinical claim is mapped to a specific passage in a specific approved document.

  9. 09

    Calculate validation status

    Automated checks produce an assurance summary before the scenario enters academic review.

  10. 10

    Save as draft for academic review

    The scenario enters a review and approval workflow before publication or assessment use.

03. The knowledge platform

Three forms of evidence, one version-controlled platform

AccuRXy is not a generic document store. Each form of knowledge serves a different function in scenario generation, and each source carries the metadata that makes academic traceability possible.

A

Source documents

Original authoritative documents: clinical guidelines, regulatory guidance, SmPCs and authorised product information, practice standards, SOPs, peer-reviewed literature, university-approved educational material and medication-safety reports. Stored in encrypted object storage.

B

Structured medicines and clinical data

Active ingredients, strengths, formulations, routes, licensed indications, contraindications, interactions, monitoring requirements, controlled drug status and jurisdiction, held as structured records. A language model should not be relied upon as the definitive source of whether a strength, formulation or interaction exists.

C

AccuRXy proprietary knowledge

The commercially defensible layer that develops over time: medication-error taxonomy and error pattern library, scenario templates and difficulty models, learning-objective mappings and scoring rubrics, common misconceptions and distractor rules, clinical reasoning pathways, validated exemplars and data on which scenarios learners find difficult.

Every source document carries structured metadata

Source ID and title

Publisher and document type

Jurisdiction and clinical topic

Publication, effective and review dates

Version, superseded status, approval status

Licence, academic owner and original URL

This metadata is what creates academic traceability. AccuRXy can demonstrate not just that a scenario was informed by a source, but which version of that source was approved and current on the date the scenario was generated.

04. Retrieval

Evidence first: retrieval-augmented generation

AccuRXy prepares the evidence pack before the model is called. Jurisdiction and version are explicit filters, and the model works only within the evidence it has been given.

AccuRXy does not ask an AI model a free-text question. The evidence comes first: the retrieval service selects approved, jurisdiction-appropriate material before the model is involved at all.

01

Semantic search for conceptually relevant passages

02

Keyword search for exact terminology and medicine names

03

Metadata filters: jurisdiction, approval status, effective date, learner level

04

Source-authority weighting

05

Relevance reranking

Jurisdiction and version control are explicit filters, not afterthoughts. An Irish scenario will not silently return a superseded UK or US protocol because it is semantically similar.

The evidence pack

A curated collection of relevant source passages, structured medicine records, applicable clinical rules and required learning outcomes, each carrying its source identifiers and version details. The pack is supplied to the model, not generated by it: the model completes a defined output schema with sections for patient profile, scenario steps, decision points, correct actions, distractors, feedback and source claims.

Provider neutral by design

AccuRXy can use OpenAI, Anthropic (Claude) or Google Gemini through the same model gateway. The evidence pack is prepared by AccuRXy before the model is called, so whichever model is selected works from the same governed evidence. The academic standard is consistent regardless of AI provider.

05. Validation

From mostly accurate to demonstrably governed

Generating a scenario is not the end of the process. Six layers of validation sit between a generated draft and a scenario approved for teaching or assessment.

01

Schema validation

Every scenario must contain all mandatory sections: learning objective, patient information, clinical cues, decision points, correct answer, explanation and references. A scenario missing any mandatory section cannot progress.

02

Evidence-entailment validation

For every important clinical claim, AccuRXy checks whether the cited source passage actually supports that claim. This is claim-level citation mapping, not a bibliography appended at the end.

03

Structured-data validation

Generated content is checked against the medicines database: is the product real, does the stated strength exist, is the route valid, is the dose plausible, does the named interaction exist, is the patient age compatible.

04

Internal-consistency validation

The intended error must appear in the scenario, the correct answer must genuinely resolve it, the explanation must agree with the answer, the medicine list must stay consistent throughout, and no distractor may be accidentally clinically correct.

05

Second-pass model review

A second AI model can act as a critic, evaluating the draft against the evidence pack before human review. It improves quality but is not treated as the final academic authority.

06

Human academic review

For assessed teaching or formal CPD, final approval rests with a qualified academic or relevant professional reviewer, moving through a defined status workflow.

Claim-level citation mapping, an example

Clinical claimSourceSupporting passageResult
Proposed dose is potentially unsafeSource 104Section 4.2Supported
Drug interaction is clinically relevantSource 221Interaction tableSupported
Pharmacist should contact the prescriberSOP 018Escalation sectionSupported

Review status workflow

01 AI Draft

Generated and awaiting automated validation

02 Automatically Validated

Passed all automated checks

03 Academic Review Required

Flagged for human expert review

04 Faculty Approved

Reviewed and approved by a named academic

05 Published

Available for use in teaching or assessment

06 Review Due / Superseded

Source material updated, re-review required

06. Transparency

What academics and partners can see

Every scenario carries an Evidence and Assurance panel, and every generation saves an immutable evidence pack that records exactly how the content was produced.

Evidence and Assurance panel

Governance made visible inside every scenario

Scenario assurance status
Faculty Approved
Evidence retrieval
8 approved sources
Jurisdiction
Ireland
Sources current at generation
Yes
Medication records validated
Passed
Clinical-rule checks
12 / 12 passed
Citation support
17 / 18 claims supported
Academic reviewer
Named faculty member
Review date
25 June 2026
Scenario version
2.1
Model used
Provider, model name and version recorded
Evidence-pack version
EP-8462

An academic reviewing a decision point can click through to the specific source title, supporting passage, section or page reference, source version, date accessed and a link to the original where permitted.

The immutable evidence pack, saved at the moment of generation

Scenario specification and prompt-template version

Retrieved source IDs and passages

Structured medicine records used

Clinical rules applied

Model provider, model name and version

Generation settings and raw structured response

Validation results

Human edits

Reviewer identity and approval decision

Timestamp

If a clinical guideline is updated six months after a scenario was generated, AccuRXy can demonstrate which version was approved and current on that date. Old scenarios are never silently regenerated using newer sources; they are marked for review.

Technical data stores

StorePurpose
PostgreSQL with pgvectorUsers, organisations, scenarios, versions, sources and metadata, review records, citations, validation results, clinical rules, model run logs, audit trail
Object storage (e.g. AWS S3)Original PDFs, guideline documents, tables, images, evidence file imports, generated exports
Search index / vector storeDocument chunks, embeddings, keyword search and hybrid retrieval. pgvector for MVP, OpenSearch or Elasticsearch at scale
Rules serviceApplication code and database-driven clinical validation rules. No complex enterprise rules engine required for MVP

07. Evidence source inventory

Seven pillars of source material

A granular inventory of the reference sources, guidelines, legislation and clinical data from which AccuRXy constructs jurisdiction-appropriate evidence packs for the Irish jurisdiction.

Structured data

Queryable database records: medicines, rules, taxonomy

Document chunks

Indexed text passages for semantic and keyword retrieval

Linked / referenced

External source, licensing or access arrangement needed

01

Regulatory: Medicines Authorisation and Product Information

The anchor of the evidence pack. Every Irish pharmacy scenario involving a specific medicine must be grounded in the authorised product information for that medicine as licensed in Ireland by the HPRA or the EMA. This pillar is what gives AccuRXy jurisdiction specificity: it uses Irish product information, not UK or US sources.

SourceHeld asRole in the evidence pack
medicines.ieDocument chunksThe official Irish SmPC and PIL database, continuously updated by pharmaceutical companies under HPRA/EMA approval. Covers all brands and generics licensed and marketed in Ireland. The primary Irish jurisdiction product information source, not the UK EMC and not the BNF.
HPRA Find a Medicine databaseDocument chunksThe statutory HPRA listing of every human medicine assessed and licensed for the Irish market: SmPC, PIL, Public Assessment Report and educational materials such as pregnancy prevention programmes for isotretinoin, valproate and clozapine.
EMA product database (centrally authorised products)Document chunksAuthoritative for biologicals, orphan drugs, oncologics and other centrally authorised products. EMA SmPCs are valid across all EU member states including Ireland, used where a product is not separately authorised by HPRA.
HPRA Drug Safety NewsletterDocument chunksRegular pharmacovigilance updates specific to Ireland: prescribing changes, new safety warnings and direct healthcare professional communications. Dynamic content requiring versioned ingestion.
HPRA Interchangeable Medicines ListStructured dataThe statutory list of generic substitution permitted products, directly relevant to dispensing error scenarios involving generic switching or brand-change queries.
HPRA Medicines Shortage NotificationsStructured dataLive list of current and recently resolved shortages, relevant to dispensing decision scenarios where the prescribed product is unavailable.
HPRA Educational Materials (EdMs)Document chunksRisk minimisation materials for high-risk medicines: isotretinoin, valproate, clozapine, thalidomide and methotrexate. Mandatory dispensing requirements and rich material for high-stakes scenario design.

Licensing note: medicines.ie is operated commercially and content licensing for AI retrieval will require a formal agreement. This is the most critical commercial relationship to establish for Irish jurisdiction accuracy.

02

Regulatory: Irish Pharmacy Legislation

Irish pharmacy law is the framework within which every scenario must operate. Most generic AI tools have no structured awareness of Irish pharmacy legislation as a coherent body of law. This pillar allows AccuRXy to generate scenarios that are not only clinically plausible but legally accurate for Ireland.

Primary legislation

SourceHeld asRole in the evidence pack
Pharmacy Act 2007 (as amended)Document chunksThe principal Act governing pharmacy regulation in Ireland: PSI functions, registration of pharmacists and pharmacies, fitness to practise. As amended by the Regulated Professions (Health and Social Care) (Amendment) Act 2020.
Misuse of Drugs Acts 1977 to 2016 (consolidated)Document chunksThe primary legislation controlling scheduled substances in Ireland. Underpins all controlled drug scenarios: opioids, benzodiazepines, stimulants and cannabis-based medicines.
Misuse of Drugs Regulations 2017, S.I. No. 173Document chunksThe operational statutory instrument for controlled drugs: schedules 2 to 5, prescribing requirements, record-keeping obligations, CD safe storage and destruction. The most clinically scenario-relevant piece of controlled drug law.
Medicinal Products (Prescription and Control of Supply) Regulations 2003, S.I. No. 540Document chunksThe primary framework for prescription-only medicines: prescription requirements, emergency supply, pharmacist and nurse prescribing. Scenarios on prescription validity or emergency supply draw directly from this.
Regulation of Retail Pharmacy Businesses Regulations 2008, S.I. No. 488Document chunksMinimum standards for pharmacy premises, record keeping and operations. Relevant to premises compliance, duty pharmacist requirements and SOP scenarios.
Medicinal Products (Control of Placing on the Market) Regulations 2007Document chunksGoverns market authorisation and supply chain integrity, relevant to falsified medicines and product recall scenarios.

PSI subordinate rules

SourceHeld asRole in the evidence pack
PSI (Retail Pharmacy Businesses) (Registration) Rules 2008Document chunksOperational rules for pharmacy registration and ongoing compliance, referenced in premises and record-keeping scenarios.
PSI (Education and Training) (Integrated Course) Rules 2014Document chunksThe legal basis for the PSI five-yearly accreditation cycle of MPharm programmes, relevant to academic governance scenarios.
03

Professional Standards: PSI Guidelines and Guidance

PSI guidelines sit between statute and professional expectation. They define what good practice looks like in specific situations and are the primary source for scenario-relevant clinical decision-making rules: what should a pharmacist do, when, and why.

Dispensing, supply and records

SourceHeld asRole in the evidence pack
PSI Guidelines on the Keeping of Records in Respect of Medicinal Products (Version 4, March 2024)Document chunksCurrent version. Daily dispensing record, controlled drug registers, Schedule 2 to 4 requirements and computerised record systems. Specific enough to generate prescription-record error scenarios at clause level.
PSI Guidelines on the Storage of Medicinal Products within a Retail Pharmacy BusinessDocument chunksCold chain, fridge management and CD safe storage, relevant to temperature-excursion and storage-error scenarios.
PSI Guidelines on the Disposal of Medicinal Products within a Retail Pharmacy BusinessDocument chunksPatient-returned medicines, waste and controlled drug destruction, relevant to safe disposal and regulatory compliance scenarios.
PSI Guidelines for Retail Pharmacy Businesses: Premises RequirementsDocument chunksConsultation room, dispensary layout and access requirements, used to contextualise community pharmacy scenarios.

Controlled drugs

SourceHeld asRole in the evidence pack
PSI/Medical Council Joint Guidance on Safe Prescribing and Dispensing of Controlled Drugs (updated 2024)Document chunksSchedules 2, 3 and 4 Part 1. Opioid prescribing patterns, methadone and buprenorphine supply, requisition formats, forgery recognition and escalation pathways. One of the most scenario-rich documents in the Irish pharmacy literature.
PSI Guidance for Pharmacists on the Safe Supply of Methadone and Buprenorphine for Opioid SubstitutionDocument chunksOperational guidance for OST dispensing: instalment dispensing, supervised consumption and emergency supply. High-stakes, high-frequency scenario territory.

Vaccination and clinical services

SourceHeld asRole in the evidence pack
PSI Guidance on the Provision of Vaccination Services by Pharmacists in Retail Pharmacy BusinessesDocument chunksThe framework for pharmacist-led vaccination: consent, anaphylaxis management and record keeping. Updated annually for each vaccination season.
PSI Guidance on the Provision of an Influenza Vaccination Service for Children Aged 6 Months and OlderDocument chunksPaediatric-specific guidance, relevant to dose selection, parental consent and age-band error scenarios.
PSI Guidance on Non-Prescription Medicinal Products Containing CodeineDocument chunksA common community pharmacy consultation scenario: appropriate supply criteria, patient questioning and misuse risk management.
PSI Guidance on the Administration of Injectable Medicines by PharmacistsDocument chunksRelevant to clinical services training scenarios involving injectable administration by pharmacists.

Patient safety and error learning

SourceHeld asRole in the evidence pack
PSI Learning from Inquiries (Professional Conduct Committee case summaries)Document chunksAnonymised real Irish disciplinary proceedings involving dispensing errors, wrong patient, wrong dose and controlled drug irregularities. Arguably the most valuable source in the entire inventory for authentic Irish medication error scenarios.
PSI Inspection Findings OverviewDocument chunksPatterns emerging from actual PSI pharmacy inspections, giving scenario contexts grounded in real compliance failures in Irish pharmacies.
PSI Practice UpdatesDocument chunksRolling publication series covering dispensing and medication errors, patient safety, recalls, falsified medicines and practice information. Requires regular re-ingestion.

Academic reviewer note: PSI Learning from Inquiries cases are publicly available, but their use in AI-generated educational scenarios may benefit from PSI engagement or ethical framing, even though all cases are anonymised.

04

Medicines Information: Clinical Reference Sources

These are the sources Irish pharmacists and pharmacy students use at the point of care. Including them ensures AccuRXy generates scenarios using the same reference frameworks that students are trained on and assessed against.

SourceHeld asRole in the evidence pack
Irish Medicines Formulary (IMF)Structured data and document chunksThe only Ireland-specific, fully independent clinical medicines reference. Covers all licensed and marketed products: indications, dosages, interactions, contraindications, side effects, HSE reimbursement status and HPRA interchangeability. Updated twice yearly. AccuRXy should pursue a licensing agreement with IMF Media.
MIMS IrelandLinked / referencedA prescribing and drug directory reference used by Irish GPs, relevant to the prescriber-side information environment that pharmacy students must navigate.
National Medicines Information Centre (NMIC) bulletinsDocument chunksBased at St James's Hospital Dublin. Evidence-based drug information bulletins and responses to medicines information queries, giving Ireland-specific context on interactions, dosing in special populations and therapeutic alternatives.
HSE Medicines Management Programme guidanceDocument chunksPreferred medicines lists, best-value biologicals and managed access protocols for high-cost medicines. Directly relevant to reimbursement and High Tech dispensing scenarios.
HSE Primary Care Reimbursement Service (PCRS) schemesStructured dataGMS, DPS, LTI, High Tech Arrangement and Hardship schemes. Essential for scenarios involving reimbursement eligibility, scheme switching or patient payment queries.
HSE National Clinical Guidelines (NCEC series)Document chunksNumbered, evidence-based, HSE-approved guidelines, providing clinical context for monitoring-heavy scenarios such as anticoagulation, diabetes and asthma.
05

Education and Competency Framework

This pillar is what makes AccuRXy academically credible to pharmacy faculty. Every scenario's learning objectives can be mapped to the national competency framework and institutional curricula. This is the answer to how the platform knows a scenario is appropriate for a Year 3 student at a given school.

SourceHeld asRole in the evidence pack
PSI Core Competency Framework for Pharmacists (Version 2, December 2022, updated September 2023)Structured dataThe definitive competency map for Irish pharmacy. Four domains, Personal, Professional, Organisation and Management Skills and Pharmacy Care, plus Public Health. Embedded throughout MPharm programmes at TCD, UCC and RCSI, and every scenario maps its objectives to specific competencies.
PSI MPharm Accreditation StandardsDocument chunksThe standards against which MPharm programmes are evaluated, defining what a PSI-accredited pharmacy graduate must demonstrate. Used to calibrate scenario difficulty and coverage.
PSI Accreditation Report: TCD (2025)Document chunksThe most recent TCD accreditation report, with observations on learning outcome alignment, OSCE design and clinical placement assessment.
PSI Accreditation Reports: UCC and RCSIDocument chunksEquivalent reports for UCC and RCSI, essential for institution-specific scenario calibration.
APPEL (Affiliation for Pharmacy Practice Experiential Learning) placement frameworksDocument chunksNational body managing practice placements for TCD, UCC and RCSI. Placement competency frameworks and assessment tools define what students must demonstrate in community and hospital settings.
MPharm national learning outcomes (PSI-defined)Structured dataThe national learning outcome framework from which scenario objectives derive, enabling year-appropriate calibration.
PSI Professional Registration Examination (PRE) blueprintDocument chunksThe assessment framework for the final-year pre-registration examination. The highest-stakes assessment context for AccuRXy scenarios.
New pharmacy programmes: ATU Sligo, SETU Waterford, University of Galway (announced 2024)Linked / referencedThree new MPharm programmes designed to double Ireland's annual pharmacy graduate output. Documentation as it becomes available.
06

Medication Safety: Error Taxonomy and Incident Learning

This is where AccuRXy's most distinctive intellectual property develops. The combination of a formal Irish-applicable error taxonomy with real Irish incident evidence produces scenarios that are both academically valid and clinically authentic. No generic AI tool can replicate this.

Irish incident evidence

SourceHeld asRole in the evidence pack
PSI Learning from Inquiries (cross-referenced from Pillar 3)Document chunksUsed here to extract error type, medicine class, dispensing stage, contributing factors and outcome, populating the AccuRXy proprietary error pattern library.
HPRA Drug Safety Newsletters (pharmacovigilance signals)Document chunksSafety signals that have reached regulatory action level in Ireland, providing evidence for plausible real-world consequences in scenarios.
State Claims Agency (SCA) Clinical Indemnity Scheme annual reportsDocument chunksPharmacy and medicines-related claim patterns for Irish healthcare settings, showing the most clinically and financially consequential error types.

Error classification and taxonomy

SourceHeld asRole in the evidence pack
WHO Medication Without Harm, Global Patient Safety ChallengeDocument chunksThe international WHO framework for medication safety, providing the overarching classification structure: patient factors, medication factors, healthcare provider factors and system factors.
ISMP error taxonomyStructured dataThe most widely used medication error classification system. Maps onto Irish practice and forms the basis for the AccuRXy proprietary error taxonomy.
NCC MERP index for categorising medication errors (categories A to I)Structured dataThe standard severity classification, from no-harm near miss through to patient death. Used in scenario metadata to tag difficulty and consequence level.
Dispensing error classificationsStructured dataAccuRXy proprietary taxonomy layer covering wrong patient, wrong drug, wrong dose, wrong strength, wrong formulation, wrong route, wrong frequency, wrong duration and wrong label. Each error type maps to scenario templates, clinical cues, common distractors, learner interventions and relevant legislation.

Drug-specific safety materials

SourceHeld asRole in the evidence pack
HPRA Educational Materials for high-risk medicinesDocument chunksCross-referenced from Pillar 1. Define the specific dispensing interventions required for high-risk medicines and underpin high-stakes scenario design.
HSE warfarin anticoagulation information packsDocument chunksHSE-produced patient information, the standard reference for anticoagulation monitoring scenarios, one of the highest-frequency dispensing error risks in Irish community pharmacy.
Joint Formulary Committee anticoagulation guidelinesDocument chunksClinical context for warfarin dose adjustment, INR interpretation and patient counselling when INR is out of range.
07

Linked and Referenced: External Sources

These sources are relevant to Irish pharmacy education but are not Irish-jurisdiction-specific, or require access or licensing arrangements AccuRXy does not yet hold. They are listed because pharmacy academics will reference them and scenario reviewers should understand where they sit in the source hierarchy.

SourceHeld asRole in the evidence pack
EMC (Electronic Medicines Compendium) via Eolas MedicalLinked / referencedUK SmPCs, live on the Eolas clinical platform. Relevant context where UK and Irish licences differ. Not the primary Irish source; medicines.ie and HPRA hold that role.
British National Formulary (BNF) and BNF for ChildrenLinked / referencedUsed contextually in Irish hospital settings even though not Ireland-specific. Acknowledged but never the primary evidence source in Irish scenarios.
EMA PRAC meeting highlightsLinked / referencedEU-level safety signals that precede HPRA regulatory action, useful for advanced scenarios involving emerging safety concerns.
Cochrane Library systematic reviewsLinked / referencedEvidence-based context behind therapeutic decisions, for scenarios that require justification of clinical choices in a specific patient profile.
Irish Pharmacy Journal (IPJ)Document chunksThe official PSI publication: practice articles, research, professional guidance and continuing education content.
ISMP Medication Safety Alerts and High-Alert Medications listLinked / referencedInsulin, anticoagulants, concentrated electrolytes, chemotherapy and opioids. Internationally recognised and used in Irish hospital pharmacy.
WHO Essential Medicines List (Irish comparative context)Linked / referencedA reference framework for which medicines represent core therapeutic competence expectations, informing which drug classes the scenario library prioritises.

08. Priority

The source hierarchy

When AccuRXy retrieves an evidence pack for an Irish scenario, it draws from the pillars in a defined priority order. Higher layers constrain what lower layers can produce.

01

Legislation

Pharmacy Act 2007, Misuse of Drugs Regulations 2017, Medicinal Products Regulations 2003

02

Regulatory authorisation

HPRA medicines database, medicines.ie SmPCs, EMA product information

03

Professional standards

PSI guidelines, PSI/Medical Council joint guidance, PSI Practice Updates

04

Competency framework

PSI Core Competency Framework, MPharm learning outcomes, PRE blueprint

05

Clinical medicines reference

Irish Medicines Formulary, NMIC bulletins, HSE MMP guidance

06

Incident learning

PSI Learning from Inquiries, HPRA Drug Safety Newsletters, SCA data

07

AccuRXy proprietary layer

Error taxonomy, scenario templates, difficulty calibration, distractor rules

When AccuRXy retrieves an evidence pack for an Irish scenario it draws from the pillars in this priority order. Higher layers constrain what lower layers can produce.

09. Review

Recommended next steps following academic review

Actions arising from review of the source inventory, spanning academic annotation, licensing, legal advice, data schema and partnerships.

Academic reviewers

Annotate gaps, corrections or additional sources expected for their institution's MPharm programme.

Commercial

Initiate the licensing conversation with IMF Media, the single most important source relationship for Irish jurisdiction accuracy.

Legal and compliance

Take advice on the use of PSI Learning from Inquiries cases in AI-generated educational scenarios, even though all are anonymised.

Development

Define the metadata schema for each source type (version, effective date, jurisdiction, document type, approval status) before ingestion begins.

Product

Map the source pillars to the Scenario Design Lab parameter set so each selection triggers the appropriate retrieval filters.

Partnerships

Explore a formal relationship with the PSI Education team, positioning AccuRXy as the national digital scenario platform ahead of the three new pharmacy school launches.

In plain language

AccuRXy uses a governed retrieval-augmented generation architecture. Approved academic, regulatory and medication-safety sources are stored in a version-controlled evidence platform and supplemented by structured medicines data and clinical validation rules. When a scenario is requested, AccuRXy retrieves a jurisdiction-appropriate evidence pack, instructs an approved language model to generate a structured scenario within that pack, validates the result, maps material claims to supporting source passages and submits the scenario for academic review.

The reference resources are not just accessible, they are central to every scenario from the first line of evidence retrieval through to the final audit record.