Why it matters
Without this system, missing plants, mislabelled accessions, and inconsistent records can arise, leading to incomplete inventories and audit difficulties during accreditation and annual reporting.
From design to safety
Design
AI designs stocktake workflow
AI models define data requirements, identify tag types, and plan data fields for plant records, reconciliation, and stock thresholds to guide the initial design.
- accession records database
- barcode tags on plant labels
- GIS-based garden map layer
Build
Install tagging and data pipelines
Data capture schemes integrate with accession records, update tag inventories, and connect handheld scanners to the GIS map for real-time walkthrough input.
- barcode tags on plant labels
- RFID tag system
- handheld barcode/RFID scanner
Operation
Run daily inventory monitoring
The system reads plant tags, cross-checks with records, flags discrepancies, and generates daily reconciliation reports for review.
- accession records database
- RFID tag system
- GIS-based garden map layer
Maintenance
Maintain data hygiene and tag integrity
Regular audits verify tag legibility, update missing plant status, and reconcile changes with the live plant records ledger.
- handheld barcode/RFID scanner
- reconciliation records ledger cabinet
- barcode tags on plant labels
Safety
Ensure traceability and accountability
All stocktake actions are logged, flagged issues routed to the curator for approval, and audit trails maintained for regulatory compliance.
- accession records database
- GIS-based garden map layer
- reconciliation records ledger cabinet
Part by part
- 1
accession records database
- Watch for
- mismatched accession numbers, duplicate records, or missing metadata during stocktake
- AI action
- read tag IDs and compare to accession IDs; trigger discrepancy alerts; log audit event with timestamp and operator ID; initiate reconciliation workflow
- 2
barcode tags on plant labels
- Watch for
- faded or detached tags, unreadable barcodes, tag drift from plant
- AI action
- scan and validate tag readability; flag unreadable or out-of-sync tags; prompt replacement order and capture replacement history
- 3
RFID tag system
- Watch for
- tag failures, unread RFID zones, inconsistent tag IDs
- AI action
- poll RFID readers; surface discrepancies to records for reconciliation; store signal strength and tag read count for trend analysis
- 4
handheld barcode/RFID scanner
- Watch for
- incomplete sweeps, missed plants, or scanner downtime
- AI action
- prompt operator to complete walk route and re-scan missing plants; append geolocation to each scan row
- 5
GIS-based garden map layer
- Watch for
- bed misalignment, plant location drift, or bed renaming
- AI action
- update map coordinates on confirmed moves and label changes; trigger notifications for location conflicts
- 6
reconciliation records ledger cabinet
- Watch for
- unsigned changes, unapproved edits, or lagging records
- AI action
- route changes to curator for sign-off and archive old entries; enforce two-person validation for transfers
Worked example
Catching a mislabel before an audit
Illustrative scenario with sample data.
A routine stocktake detects a barcode mismatch on a specimen in bed 12 and a historic accession record shows a different owner. Immediate reconciliation is required to maintain accuracy.
Signal in
acession_id: A-1023, plant_tag: B-457, bed_location: Bed 12, date: 2026-11-01, status: active, last_record_update: 2026-09-15, origin: wild-origin, provenance_type: ex situ; location_history: Bed 9→Bed 12
System out
Discrepancy found: accession A-1023 tagged as B-457 in Bed 12; record shows A-1023 for Bed 9. Flagged for curator review; reconciliation log entry created; tag re-scan scheduled; provenance updated to reflect move and new location.
Human sign-off: Head of Living Collections reviews discrepancy, approves change, and updates the accession record with Bed 12 confirmation; provenance_type remains ex situ with updated origin notes.
How to roll it out
Phase 1
Weeks 1 to 2
Baseline data integration, inventory map alignment, and tag inventory calibration to establish a trustworthy starting dataset.
Phase 2
Weeks 3 to 6
Pilot stocktake in two zones, refine reconciliation workflows and alarms, validate data flows from tags to the ledger.
Phase 3
Weeks 7 to 10
Full garden rollout, operator training, and live-dashboard deployment with audit-ready exportable reports.
Where people stay in charge
- Decisions on accession status changes, transfers, or removals must be signed by the Head of Living Collections.
- Discrepancies impacting safety or critical accessions require immediate notification to the Garden Director.
- All data changes must be logged with a timestamp, user, and rationale.
Questions owners and crews ask
How does AI help keep accession records accurate across cycles?
AI continuously reads tag data, cross-checks with the central accession database, and flags inconsistencies for curator review, ensuring records reflect live plant positions and movement history.
What happens when a tag is unreadable during a stocktake?
The system prompts a manual read, schedules a tag replacement, and logs the issue for the records team to update provenance, status, and restoration steps.
Who approves changes identified by the stocktake AI?
The head of living collections or garden director approves major changes, with the plant records officer validating data integrity and cross-checking provenance notes.
How are discrepancies reported to governance and trustees?
Automated reconciliation reports are prepared for stewardship review, including actions taken, current stock status, and upcoming verification steps suitable for board reporting.