Why it matters
Without AI triage, condition drift, missed loans, and uncertain treatment planning risk delays, misallocated resources, and disputes over archival integrity and insurance claims.
From design to safety
Design
AI-driven triage criteria defined
Define data inputs (condition flags, marginalia, humidity exposure, previous treatments) and decision rules for prioritizing items for conservation and encapsulation. Establish metrics, provenance, and versioned treatment pathways.
- condition assessment checklist
- treatment pathway schema
- triage survey app
Build
Data capture and workflow scaffolding
Implement scanners, planar flattening press, encapsulation station, and bulk storage integrations; configure metadata schemes and automated QA checks. Create secure backups and offline mirrors for disaster recovery.
- high-resolution scanner
- planar flattening press
- encapsulation station
Operation
Live triage monitoring and prioritization
AI reads condition scores, scan data, and user notes to rank items, generate treatment proposals, and produce work orders. Track item histories and ensure traceability of decisions.
- digital triage tablet
- bulk archival storage rack
- acid-free folder stock
Maintenance
Calibration and data hygiene cycles
Regularly recalibrate imaging sensors, update condition thresholds, refresh metadata templates, and audit AI decision logs for traceability. Validate data integrity after system updates.
- condition monitor
- metadata repository
- quality assurance module
Safety
Ethics, COSHH, and treatment reversibility
Enforce reversibility of interventions, document COSHH assessments for solvents, and ensure staff sign-off on potentially irreversible steps. Maintain PPE compliance and staff vaccination status where applicable.
- safety data sheet binder
- fume extraction unit
- reversibility checklist
Part by part
- 1
Planar flattening press
- Watch for
- Cord snagging, edge distortion, creasing that intensifies with humidity changes
- AI action
- Detect flattening-induced stress and flag items needing alternate flattening approaches
- 2
Encapsulation station
- Watch for
- encapsulation adhesive delamination, trapped air bubbles, yellowing
- AI action
- Recommend encapsulation method and material batch based on paper type and moisture data
- 3
High-resolution scanner
- Watch for
- Scan artifacts, transparency loss, edge skew
- AI action
- Trigger re-scans and auto-generate condition stamps for each page surface
- 4
Acid-free folder stock
- Watch for
- Degradation of storage folders, cellulose acidity
- AI action
- Suggest inventory replacement schedule and material substitutions for long-term stability
- 5
Bulk archival storage rack
- Watch for
- Shelf wear, stacking pressure hotspots, access bottlenecks
- AI action
- Optimize storage layout and generate upcoming-access plans for triage rounds
- 6
Digital triage tablet
- Watch for
- Incomplete field data, missing borrower notes, inconsistent labeling
- AI action
- Prompt field-data completion and auto-validate against standard metadata templates
Worked example
Catching a flood-damaged map before deacidification
Illustrative scenario with sample data.
A large rolled map shows blistering along the fold lines after a humidity spike; immediate triage is needed to avoid irreversible tailoring or staining. The team must stabilize, document, and plan extraction without damaging fragile ink lines.
Signal in
collection: 1920s rolled maps; location: storage vault B; data: humidity=68%, temperature=22C, foldLineScores=4-6, previousTreatment=scotch tape, conservationHistory=none, inkIntegrity=0.85, backingPaper=frayed
System out
priorityOrder: flattening and stabilization; recommended encapsulation restart; inventory updated with new roll counts; work order created with required materials and staff assignments; monitoring plan issued; risk register updated; curatorial note attached.
Human sign-off: Head of Conservation reviews the proposed triage path, signs off the work order, and approves material substitutions and the expanded monitoring plan.
How to roll it out
Phase 1
Weeks 1 to 4
Baseline setup of imaging, flattening, and encapsulation stations; initial data templates and QA workflows defined.
Phase 2
Weeks 5 to 12
Full AI-driven triage cycles begin; integration with condition reports and digital logbooks; staff training completed.
Phase 3
Weeks 13 and beyond
Routine triage operations with continuous improvement through feedback loops and audit trails.
Where people stay in charge
- Decision on irreversible treatments must be signed off by the accredited conservator.
- COSHH and reversibility checks must be completed by the safety officer before any solvent use or adhesive application.
- AI outputs must be reviewed against existing condition reports by a conservator before work orders are issued.
Questions owners and crews ask
What data does AI use to triage large paper collections?
AI uses condition flags, humidity and temperature readings, paper type, age, prior treatments, marginalia, and imaging data to assign priority and propose conservative actions.
How does AI decide when to flatten or encapsulate items?
AI weighs flattening risk versus damage from curling, moisture, and handling; it proposes flattening or encapsulation with conditions for reversibility and long-term stability.
Who signs off on AI-generated treatment proposals?
The accredited conservator reviews and signs off treatment proposals, ensuring reversibility, ethical practice, and PACR accreditation requirements.
How is data kept safe and auditable?
All condition records, scans, and work orders are stored in a secured metadata repository with versioning, access logs, and explicit sign-off trails.