Why it matters
Without AI, misselection of samples and inconsistent modelling can create delays, unreported age estimates, and backlogs in post-excavation assessment and archive deposition.
From design to safety
Design
AI informs sampling strategy
AI reviews draft sampling plans, existing contexts, and written scheme of investigation to select radiocarbon samples that maximize dating precision, minimize cost, and align with project sign-off workflows across HER. It integrates site-wide context maps, stratigraphic relationships, and lab capacity calendars to ensure compatible submission windows, chain-of-custody tracking, and contingency options.
- radiocarbon samples
- short-lived material
- calibration
- dating reports
Build
Assessing samples in field
AI supports on-site assignment of context sheets and sample labels, ensuring chain-of-custody, volume checks, container IDs, and immediate suitability flags attached to the context record. It prompts barcode scans, photo-verification, and dual-tag reconciliation before shipment to the AMS facility.
- radiocarbon samples
- short-lived material
- calibration
- dating reports
Operation
Modelling and calibration in real time
AI links observed dates with calibrated curves, runs Bayesian updates, cross-validates with existing context networks, and flags anomalies for curator review before final interpretation. It maintains a live audit trail, stores posterior results, and previews alternative priors for sensitivity analysis.
- Bayesian modelling
- OxCal
- calibration
- dating reports
Maintenance
Recalibration and reprocessing
AI manages re-reads, updated calibration curves, versioned dating reports, archive deposition drafts, and differential provenance notes to support post-excavation assessment. It schedules monthly revalidation, logs curve versioning, and integrates with HER back-office systems for deposition-ready outputs.
- Bayesian modelling
- OxCal
- calibration
- dating reports
Safety
Documentation integrity and compliance
AI ensures traceable dating records, sign-off by qualified persons, alignment with HER and CIfA standards, and retention of auditable logs for planning authorities. It generates compliance dashboards and flags missing sign-offs before submission deadlines.
- calibration
- dating reports
- Bayesian modelling
- OxCal
Part by part
- 1
radiocarbon samples
- Watch for
- Mislabelled samples or degraded material causing age drift or contamination risk.
- AI action
- AI flags mislabelled contexts and triggers re-sampling workflow.
- 2
short-lived material
- Watch for
- Inclusion of long-lived material that biases age estimates.
- AI action
- AI filters material types and re-prioritizes timezone-suitable short-lived samples.
- 3
Bayesian modelling
- Watch for
- Non-convergent models or poor prior choices compromising date ranges.
- AI action
- AI proposes priors, tests convergence, and suggests model adjustments.
- 4
OxCal
- Watch for
- Outdated code or calibration curve mismatches impacting results.
- AI action
- AI applies current OxCal routines and flags deprecated steps.
- 5
calibration
- Watch for
- Calibration curve gaps or region-specific offsets affecting ages.
- AI action
- AI interpolates with latest curves and documents uncertainty.
- 6
dating reports
- Watch for
- Incomplete provenance or missing sign-offs delaying deposition.
- AI action
- AI auto-generates structured dating reports with version history.
Worked example
Refining a late Iron Age chronology on a 8.2 ha site
Illustrative scenario with sample data.
On a development-led site, new curve revisions demand rapid re-calibration of a subset of radiocarbon dates from short-lived cereal remains.
Signal in
siteCode: CR-88, contextId: C-312, sampleId: S-522, material: charred barley, dateAge: 2450±30 BP, curve: IntCal20, OxCalVersion: 4.4, lab: AMS-01
System out
alert: calibration curve updated; action: re-run with IntCal20. Updated report P-522-CR-88-2026.pdf; signOff: HER Archaeologist.
Human sign-off: Project manager authorizes re-run; planning archaeologist confirms alignment with Written Scheme of Investigation and HER consent.
How to roll it out
Phase 1
Weeks 1 to 4: Baseline
Ingest baseline sample metadata, QA checks, and initial selection rules for radiocarbon materials across contexts.
Phase 2
Weeks 5 to 12: Modelling
Construct Bayesian models; configure OxCal; update calibration curves; run convergence checks and document priors.
Phase 3
Weeks 13 to 20: Reporting
Draft dating reports with provenance chains, curate sign-offs, prepare archive deposition files and distribute to stakeholders.
Where people stay in charge
- Decisions on sample selection and modelling must be signed off by a qualified archaeologist or project supervisor due to interpretive risk.
- All calibration decisions must be reviewed against the current HER guidance and CIfA standards before reporting.
- Any anomaly flagged by AI must be reviewed by a curator or senior finds specialist to prevent misinterpretation.
Questions owners and crews ask
What data does the AI read to select radiocarbon samples?
The AI reads context sheets, stratigraphic relationships, material type, context age estimates, sample provenance, and contextual associations to rank suitability for radiocarbon dating.
How does AI handle calibration updates?
AI applies the latest calibration curves, tracks version history, validates model convergence, and flags changes requiring model re-runs or updated reports for sign-off.
Who validates the final dating report?
A senior finds specialist or curator signs off, with the project manager ensuring client and planning authorities are informed and archive deposition ready.
What if AI flags a controversial date?
The flagged result triggers a review by the planning archaeologist or HER officer, with additional samples or alternative modelling considered before publication.