Why it matters
Without AI, assemblage analysis risks inconsistent classification, missed refits, unreported typologies, and backlog in post-excavation assessment, delaying archive deposition and HER updates.
From design to safety
Design
AI-informed typology planning
AI reads typological data and refit potential from initial finds, suggesting sampling strategies, recording templates, and refurbishment of the written scheme of investigation, aligning with project briefs and regulatory expectations, with explicit data provenance and versioning.
- Flint flakes
- Cores
- Debitage
- Harris matrix draft board
Build
Automated field recording guides
AI interprets real-time input from context sheets, photogrammetry, and plan drawings to steer context definition, watch for safety, and ensure consistency with the SWI, enabling rapid field decisions and traceable edits to context records.
- Tools
- Debitage
- Refits
- Context recording binder
Operation
On-site artefact classification
AI processes colour, surface texture, and morphological measurements from imaging and microscopes to assign preliminary typology and flag ambiguous items for specialist review, generating provisional catalog numbers and linking to plan views.
- Flint flakes
- Tools
- Typology
- Digital catalog server
Maintenance
Post-excavation checks and curation
AI monitors catalogue integrity, checks refits, updates context records, and flags items for re-binning before archive deposition and post-excavation assessment, ensuring dataset hygiene for HER submission and long-term storage.
- Cores
- Refits
- Typology
- Archive storage jars
Safety
Quality gates and safety flags
AI flags unstable finds, potential hazardous contamination in artefacts, and ensures compliance with CIfA standards during storage, transfer, and welfare handling, with escalation protocols to the site safety adviser.
- Flint flakes
- Tools
- Refits
- Safety signage panel
Part by part
- 1
Flint flakes
- Watch for
- Unusual fracture patterns or cortical changes indicating knapping stages or raw material source shifts.
- AI action
- AI detects retouch and platform attributes, flags potential new typologies for curator review.
- 2
Cores
- Watch for
- Multiple striking platforms or thinning flakes indicating core reduction strategy.
- AI action
- AI estimates reduction sequence and suggests sample depths for refit analysis.
- 3
Tools
- Watch for
- Tool morphology changes suggesting utilitarian groups or functional classes.
- AI action
- AI assigns preliminary tool types and links to known typologies.
- 4
Debitage
- Watch for
- Concentration patterns by size class and platform types.
- AI action
- AI maps assemblage reduction stage and prioritizes samples for radiocarbon or use-wear studies.
- 5
Refits
- Watch for
- Fragment connectivity suggesting multiple episodes of reuse or site occupation.
- AI action
- AI proposes probable refit chains and visualizes Harris matrix connections.
- 6
Typology
- Watch for
- Ambiguities between standard classifications and regional variants.
- AI action
- AI proposes diagnostic traits and provisional typology placements for expert review.
Worked example
Settlement-scale lithic assemblage processing on a 2.3 ha site, Mesolithic to Early Neolithic horizons
Illustrative scenario with sample data.
An excavation-led trench system yields a lithic assemblage spanning Mesolithic to Early Neolithic, with dense debitage and multiple refits.
Signal in
siteAreaHa: 2.3; periodsFound: Mesolithic, Early Neolithic; contextSheets: 120; photogrammetryModels: 14; lithicItemsCatalogued: 460; preliminaryTypology: 28 items flagged as diagnostic.
System out
AI-generated report: 12 refits identified; tentative typology alignments to Mesolithic bladelets and Early Neolithic cores; dataset prepared for post-excavation assessment and archive deposition.
Human sign-off: Finds officer approves the AI-generated typology list and authorizes queue for curator review and HER transfer.
How to roll it out
Phase 1
Weeks 1 to 4
Baseline data ingestion, SWI alignment, and pilot AI-assisted context recording for lithics.
Phase 2
Weeks 5 to 12
Full integration with photogrammetry pipeline and refit-network visualization; automated curation checks begin.
Phase 3
Weeks 13 onward
Operational AI governance, post-excavation planning, and archive deposition readiness.
Where people stay in charge
- Decisions affecting classification and typology must be confirmed by the finds specialist or curator.
- All safety and welfare decisions must be signed off by the site safety adviser and project manager before endorsement.
- The archive deposition and HER transfer steps require written sign-off by the curator and heritage consultant prior to final submission.
Questions owners and crews ask
What data does the lithic AI read during field recording?
The AI reads context sheets, plan views, photogrammetry models, micromorphology scans, and preliminary measurements to seed typology and flag unusual items for specialist input.
Who approves AI-generated typology assignments?
The finds officer or curator reviews AI-suggested typologies and refit chains before they are recorded in the context database and HER.
How does the AI help with refit analysis?
AI links fragment edges and platform indicators to propose probable refit chains, generating potential Harris matrix connections for expert review.
What safety decisions remain with qualified personnel?
Anything affecting trench stability, buried services, contamination, or welfare planning must be signed off by the designated safety officer and project manager.