farm resilience / Farm owners, managers, operators, agronomists, livestock professionals, equipment and facility teams, advisers, AI and data teams, safety professionals, vendors, trainers, and affected workflow participants
Agricultural human–AI handoff tabletop
Run a discussion-based exercise for one farm AI handoff, testing evidence display, uncertainty, workload, competence, decision rights, override, fallback, escalation, action confirmation and feedback without touching production systems.
See the whole mission
Orient before entering the field.
Connect the intended outcomes, operating stages, stop conditions, and supporting technology concepts before opening the detailed action sequence.
- 01Test whether reviewers receive usable evidence and enough time
- 02Expose authority, competence, workload and automation-bias gaps
- 03Exercise abstention, conflicting evidence, override, fallback and escalation
- 04Connect decisions and executed actions to outcome and governance feedback
- 01Design a safe bounded scenario
A tabletop should expose decision gaps without manipulating live accounts, models, machinery or farm operations.
3 field actions ↓ - 02Exercise evidence presentation
A model output without freshness, applicability and uncertainty context encourages false confidence.
3 field actions ↓ - 03Exercise decision and control
Human presence is not meaningful if the reviewer cannot reject, pause or recover.
3 field actions ↓ - 04Close the learning loop
A handoff is incomplete if disagreement, near miss and outcome evidence never reach system governance.
3 field actions ↓
- G01
This is a discussion exercise, not a live model, automation, machinery or emergency-control test.
- G02
It does not design a safety function, set staffing, assign legal authority or validate AI performance.
- G03
Override and participant records may be sensitive; use a no-fault improvement boundary and appropriate access.
Design a safe bounded scenario
A tabletop should expose decision gaps without manipulating live accounts, models, machinery or farm operations.
- 01Choose one fictional AI-assisted decision, exact workflow, system version, users, consequence and no-live-action boundary
- 02Name facilitator, participants, objectives, assumptions, decision owner, domain expert, technical contact and observers
- 03Prepare injects for missing data, low confidence, conflicting field evidence, workload, unavailable expert, connectivity loss and delayed outcome
Exercise evidence presentation
A model output without freshness, applicability and uncertainty context encourages false confidence.
- 01Ask what system status, model version, operating domain, input source and quality, output units, uncertainty and limitations are visible
- 02Test whether users can distinguish observation, model inference, recommendation, approved instruction and executed action
- 03Check accessibility, language, timing, alarm volume, competing work and whether abstention is recognizable
Exercise decision and control
Human presence is not meaningful if the reviewer cannot reject, pause or recover.
- 01Have participants accept, modify, reject, defer or escalate with reason, evidence and authority
- 02Test override effect, safe alternative workflow, action confirmation, downstream automation boundary and communication
- 03Introduce automation bias, out-of-loop state, authority conflict and failed fallback without blaming individuals
Close the learning loop
A handoff is incomplete if disagreement, near miss and outcome evidence never reach system governance.
- 01Trace decision, instruction, executed action, observation, delayed effects, correction, complaint, incident and appeal records
- 02Identify monitoring, model, interface, training, staffing, policy and vendor improvements with owners and dates
- 03Record no-go conditions, time-bound exceptions, next exercise and accountable acceptance without claiming safety certification
Continue through the operation
See where this field guide fits.
Move beyond one task into the complete evidence, technology, operating, and review sequence around it.
Govern agricultural AI and human oversight
Move from a bounded farm problem through performance evidence, representative use, monitoring and meaningful human control without giving a model agricultural decision authority.
- 01 / GOVERNUnderstand agricultural AI use-case governanceTechnology→
- 02 / REVIEW USEReview one proposed AI use caseField guide→
- 03 / EVALUATEUnderstand AI performance evidenceTechnology→
- 04 / AUDIT CLAIMAudit one performance claimField guide→
- 05 / APPLYPlace the model inside decision supportTechnology→
- 06 / MONITORUnderstand model drift assuranceTechnology→
- 07 / REVIEW DRIFTReview one deployed modelField guide→
- 08 / HAND OFFUnderstand meaningful human controlTechnology→
- 09 / EXERCISERun the human–AI handoff tabletopField guide→
- 10 / PROTECTReconnect to automation safetyTechnology
Run the human–AI handoff tabletop
Test evidence, workload, competence, authority, abstention, fallback and feedback without touching production systems.