Module 10: Geotechnical Engineering Use Cases
Learning Objectives
- Use AI to organize and review borehole logs, laboratory results, field tests, groundwater observations, and geotechnical reports.
- Prompt for traceable soil-parameter tables and correlation applicability checks.
- Distinguish measured properties, derived values, correlations, assumptions, recommendations, and final design selections.
- Apply AI to foundation and slope workflow review without replacing site-specific engineering judgment.
- Build uncertainty registers that expose investigation limits and parameter sensitivity.
Geotechnical Parameter
A geotechnical parameter is a measured, derived, correlated, or selected property used to represent soil, rock, groundwater, or interface behavior in analysis or design. Its reliability depends on sampling, testing, spatial variability, stress state, drainage condition, and interpretation.
Parameter Provenance
Parameter provenance is the traceable record showing how a geotechnical value was obtained, including borehole or sample identifier, test method, depth, correlation or interpretation method, assumptions, and the engineer's selection process.
High-Value Geotechnical AI Tasks
AI can assist with:
- extracting strata, depths, test results, and groundwater observations from logs and reports;
- standardizing units and terminology while preserving source IDs;
- comparing field and laboratory classifications;
- building parameter tables with provenance labels;
- checking whether correlations are being applied outside stated conditions;
- mapping report recommendations back to investigation evidence;
- reviewing foundation or slope analysis inputs and assumptions;
- identifying missing investigation data and spatial coverage gaps;
- organizing settlement, movement, or groundwater monitoring data.
Geotechnical Evidence-to-Decision Workflow
A controlled path from investigation data to reviewed geotechnical interpretation.
Geotechnical conclusions should move from traceable site evidence to parameter selection, analysis, and qualified review; missing evidence should stop the workflow.
Define site question and consequence → Collect borehole, lab, groundwater, and site evidence; Collect borehole, lab, groundwater, and site evidence → Select parameters with provenance and uncertainty; Select parameters with provenance and uncertainty → Run bounded analysis or comparison; Run bounded analysis or comparison → Evidence and sensitivity support conclusion?; Evidence and sensitivity support conclusion? — Yes → Geotechnical professional review; Evidence and sensitivity support conclusion? — No → Record data gap or request investigation; Record data gap or request investigation → Collect borehole, lab, groundwater, and site evidence
- Define site question and consequence: terminator
- Collect borehole, lab, groundwater, and site evidence: data
- Select parameters with provenance and uncertainty: process
- Run bounded analysis or comparison: process
- Evidence and sensitivity support conclusion?: decision
- Record data gap or request investigation: document
- Geotechnical professional review: terminator
Geotechnical Evidence Prompt Pattern
Use only the supplied investigation data and report.
Create a table with borehole/sample ID, depth/elevation, material description, test data,
groundwater observations, interpreted parameters, and source location.
Label each parameter as measured, derived, correlated, assumed, recommended, or selected.
Do not fill gaps silently.
Flag spatial variability, conflicting classifications, missing provenance, and parameters whose applicability requires engineering review.
Geotechnical Parameter Provenance Stack
The evidence layers behind a defensible design parameter.
Parameter values are meaningful only when the source, test condition, location, and interpretation method remain traceable.
Site Variability and Temporal Conditions Are Central
AI should help expose rather than smooth over:
- borehole spacing and coverage limitations;
- abrupt changes in strata;
- fill or disturbed-ground uncertainty;
- groundwater variation with time and season;
- disturbed versus undisturbed sampling limitations;
- differences between field and laboratory behavior;
- sensitivity of conclusions to parameter selection;
- construction-stage conditions not represented by the investigation.
High-Value Geotechnical AI Tasks
Reviewable uses of AI in geotechnical workflows.
These AI tasks support organization and checking, but they do not remove the need for site-specific geotechnical judgment.
Distinguish Data from Interpretation and Design Selection
A soil description, laboratory result, empirical correlation, interpreted parameter, recommended value, and final design parameter are not equivalent. AI-generated summaries should preserve this chain so the reviewer can trace how each value entered the analysis.
Do Not Generalize Beyond the Investigation
AI should not extrapolate a borehole condition across an entire site without engineering justification. Subsurface conditions can change substantially over short distances, especially in fills, residual soils, alluvium, slopes, reclaimed areas, and previously developed sites.
Foundation and Slope Review Controls
When AI supports a foundation or slope workflow, require explicit review of:
- geometry and stratigraphy;
- groundwater condition and observation timing;
- loading and construction sequence;
- drained versus undrained behavior;
- parameter source, variability, and sensitivity;
- failure mechanism or limit state being analyzed;
- model geometry and boundary assumptions;
- observational or monitoring controls where applicable.
The AI should organize the review and identify evidence gaps, not silently select a design parameter or certify stability.
Evidence Versus Extrapolation
Separate supported site interpretation from claims that need additional investigation.
Spatial variability and changing groundwater conditions make unsupported extrapolation especially risky in geotechnical work.
- Tested sample result
- Observed groundwater level
- Mapped stratum at investigated point
- Uninvestigated zone assumption
- Permanent groundwater from one reading
- Site-wide parameter from isolated test
Geotechnical Review Loop
Site interpretation improves through evidence, sensitivity, and professional review.
When sensitivity or field evidence conflicts with the initial interpretation, return to parameters and evidence before finalizing.
Worked Use Cases and Reusable Prompts
The companion Worked Examples contains prompts for borehole-log extraction, soil-classification consistency, groundwater summaries, parameter provenance, correlation comparison, recommendation evidence maps, slope-input audits, and geotechnical uncertainty registers.
Geotechnical Agents Must Preserve Spatial and Investigative Context
A soil value without borehole, depth interval, test method, specimen condition, coordinate, groundwater context, and source revision can be dangerously misleading. Require the agent to carry these identifiers through extraction, correlation, parameter selection, and reporting.
Represent Parameter Uncertainty Explicitly
Do not force one deterministic soil parameter when the evidence supports a range or competing interpretations. Record source values, exclusions, transformation method, selected characteristic or design value, sensitivity cases, and the engineer who accepts the parameter basis. Agentic workflows should make uncertainty visible rather than compress it away.
- AI is useful for organizing geotechnical evidence and exposing inconsistencies, provenance gaps, or missing data.
- Parameter provenance must remain visible from test result to design selection.
- Spatial variability and groundwater uncertainty are fundamental, not secondary details.
- Correlations require applicability checks and should not be treated as direct measurements.
- Final geotechnical recommendations remain site-specific professional judgments.
References
- AI and Civil Engineering — ASCE. ASCE resources on AI applications and professional considerations across civil engineering practice.
- Artificial Intelligence and engineering responsibility — ASCE. ASCE policy affirming that AI cannot replace the professional judgment or responsibility of the civil engineer.
- Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile — NIST. Risk-management guidance supporting uncertainty disclosure, validation, and human oversight.