Why it matters
Without AI, mismatches in condition, inadequate facilities, or travel risks can cause damaged objects, failed indemnity, lost loans, or delayed exhibitions, compromising accreditation and public trust.
From design to safety
Design
Define data intake and risk criteria
Specify data fields from condition reports, travel risk, and borrower facilities; set scoring thresholds, escalation rules, metadata schemas, and documentation standards for loan suitability. Include provenance fields, version control, and data quality rules; align with Spectrum procedures and RH/Lux baselines.
- loan request intake desk
- condition report binder
- travel risk assessment binder
Build
Integrate data feeds and rules engine
Ingest standardized records into a centralized AI-ready repository; configure workflow for loan committee review; auto-generate draft documents, checklists, and provenance stamps. Implement access controls, data lineage, and disaster recovery plans.
- borrower facility reports binder
- curatorial case file
- loan file cabinet
Operation
Automated screening and scoring
AI evaluates incoming requests in real time against RH/Lux guidelines, condition baselines, and facility readiness; produces action lists, mitigations, and adaptive timelines; cross-checks with indemnity eligibility and Government Indemnity constraints.
- loan file cabinet
- condition report binder
- curatorial case file
Maintenance
Data quality and calibration checks
Regularly refresh reference datasets (facility ratings, borrower reputational checks, indemnity histories) and validate model outputs with expert sign-off and versioned records. Schedule quarterly rebaselining and annual governance audits.
- travel risk assessment binder
- condition report binder
- loan file cabinet
Safety
Audit, compliance, and sign-off trails
Maintain immutable logs of AI decisions, trigger independent review by loan committee and registrar for high-risk cases, and archive all approvals with timestamps and signatures. Ensure traceability to Spectrum, RH/Lux thresholds, and indemnity documentation.
- loan committee minutes binder
- curatorial case file
- borrower facility reports binder
Part by part
- 1
loan request intake desk
- Watch for
- Mismatched borrower details or missing project dates that trigger incomplete reviews.
- AI action
- Flag incomplete requests and assign reviewer via the loan workflow system.
- 2
condition report binder
- Watch for
- Significant deterioration not reflected in preliminary notes or divergent readings between instruments.
- AI action
- Flag discrepancy, prompt re-measurement and curator review.
- 3
travel risk assessment binder
- Watch for
- Unassessed travel routes or documented access limitations that could affect transport.
- AI action
- Request updated risk profile and proposed mitigations from registrar.
- 4
borrower facility reports binder
- Watch for
- Facility deficiencies in borrower venue that violate RH/lux or humidity thresholds.
- AI action
- Pause loan approval and request remediation plan.
- 5
curatorial case file
- Watch for
- Missing curatorial justification or provenance gaps.
- AI action
- Generate curatorial appendix and trigger committee note.
- 6
loan committee minutes binder
- Watch for
- Decision bottlenecks or unclear sign-offs on high-risk loans.
- AI action
- Notify chair and surface required approvals.
Worked example
Outgoing Loan Assessment for Regional Painting Collection
Illustrative scenario with sample data.
A regional museum evaluates a painting loan to a partner gallery with reported RH control and lighting claims; the system cross-validates against Spectrum standards and indemnity constraints.
Signal in
dateReceived: 2026-10-12; objectID: 2849; condition: minor varnish lift; travelRiskScore: 72; borrowerFacility: RH 50-60%, lux 300-350; curatorialCase: provenance annotated; facilityReport: compliant; indemnityOK: true; RHLuxOK: true; version: v2.1
System out
AI-generated loan package: risk score 78, recommended conditions (RH 50-60%, lux 300-350, no direct sunlight), travel plan, and required mitigations; dashboard flags for condition check and facility readiness; committee notified; provisional approval with caveats.
Human sign-off: Registrar signs off the package after review with conservation and curatorial leads; loan file is versioned as v2.1 and scheduled for governance review in the next committee cycle.
How to roll it out
Phase 1
Weeks 1 to 4
Baseline data model established; core datasets wired and tested with sample loans, including condition report feeds and travel risk data feeds.
Phase 2
Weeks 5 to 8
Workflow templates activated; staff training and sign-off procedures validated, including compliance checks with Spectrum and indemnity rules.
Phase 3
Weeks 9 to 12
Live pilot on non-capital loan requests; governance checks completed; audit trails populated and tested for sign-off routing.
Where people stay in charge
- Decisions affecting indemnity, loan release, or high-risk loans must be approved by the loan committee and registrar.
- All AI-generated recommendations must be validated by conservation and curatorial leads before final approval.
- Sensitive data handling and access controls must be maintained by the registrar and IT security officer to protect provenance and borrower information.
Questions owners and crews ask
What types of loan requests can the system assess?
The system evaluates standard outgoing loans, including object condition, travel risk, borrower facility readiness, and curatorial justification to propose an initial decision and required mitigations.
Who signs off on AI-assisted loan decisions?
The registrar and loan committee sign off on AI-informed recommendations, with conservation and curatorial leads providing essential validation before approval.
How does the system handle condition changes before transport?
If readings indicate deterioration, the system escalates to conservation reviews, triggers revised condition reports, and can pause the loan until recalibration is complete.
What records are kept after a loan decision?
All inputs, AI assessments, approvals, and amendments are stored in the loan committee docket with provenance and sign-off timestamps.