CONTEXT(CLONE)
Sign in

Watch repeat site surveys and recheck only what changed

For: Facilities or infrastructure asset manager with sites surveyed on a schedule

Pattern: WatchtowerNeeds physical AI: plannedDesigned for 30 to 2,000 agents
Get early accessSign in to use

The pain today

Roofs, substations and embankments are surveyed by drone or fixed camera every period. Each survey is reviewed from scratch or not at all, so slow change such as ponding, vegetation or cracking is missed.

The ask

“Watch the scheduled survey imagery for our sites. When an asset looks different from the last survey, show me both images, describe the change, and recheck only the maintenance plans and risk entries that depend on that asset.”

Plain words, as you would say it to a colleague. Edit it to fit your case before you send it.

What you attach or connect

  • Scheduled survey imagery by asset
  • Previous survey findings
  • Maintenance plans and risk register entries
  • Asset register

The unit of work

One worker task per one asset's image pair across two surveys.

Why a swarm fits

Most assets look the same from one survey to the next. Each comparison needs two images of one asset, and only a detected change triggers work on the plans that depend on it.

Not for

Fast-moving hazards. A scheduled survey cannot replace sensors or alarms.

The decision tree

6 typed decisions, each with an action for every answer

At fixed moments in a run, the engine puts one narrow question to a decision model. The decision model never writes text: it answers yes or no with a probability, picks from listed options, or gives a score, about a small slice of the material. The engine then does exactly what this tree says, which is what makes the run auditable. The thresholds are the template's design values, not measured results.

  1. Planner, while planning

    Scope checkA choice among options

    Before work starts on a unit

    Do the survey records show this asset was imaged in both the previous and the latest survey?

    Sees only: One asset's entries in the two survey indexes

    Why: Planned: an asset missing from a survey is reported as not surveyed, not as unchanged.

    • In both surveysthenAccept
    • Missing from latest surveythenMark unresolved
    • No earlier image to comparethenSkip this unit

  2. After workers, the judge checks

    Evidence checkA choice among options

    After a worker answers

    Does the written comparison describe a change to the asset itself, rather than a difference in light, season, weather or camera angle?

    Sees only: Planned design: the written comparison of one asset's image pair, with both capture conditions

    Why: Planned: most apparent change is lighting or season; this keeps the change notes worth reading.

    • Change to the assetthenAccept
    • Imaging difference onlythenReject and retry
    • Cannot separate the twothenMark unresolved

  3. Evidence checkYes or no, with a probability

    After a worker answers

    Does the quoted maintenance plan or risk entry rest on the condition of this asset that has changed?

    Sees only: The accepted change description and one quoted plan or risk entry

    Why: Planned: only plans that depend on the changed condition are reopened.

    • Yes: 0.85 or higherthenAccept
    • Unsure: 0.50 up to 0.85thenEscalate to a strong model
    • No: below 0.50thenReject and retry

  4. Reconciler, while merging

    Conflict checkYes or no, with a probability

    While reconciling

    Does the described change contradict a condition the maintenance plan or risk entry states as its assumption?

    Sees only: The change description and the plan's stated assumption about the asset

    Why: Planned: shows where a plan's basis no longer holds.

    • Yes: 0.70 or higherthenMark unresolved
    • Unsure: 0.30 up to 0.70thenMark unresolved
    • No: below 0.30thenAccept

  5. Run control, between rounds

    Another round?Yes or no, with a probability

    Between rounds

    Have the last several surveys each described further change of the same kind to this asset?

    Sees only: One asset's accepted change descriptions across recent surveys

    Why: Planned: progressive change keeps an asset on the recheck list; stable assets drop off.

    • Yes: 0.60 or higherthenContinue
    • Unsure: 0.30 up to 0.60thenContinue
    • No: below 0.30thenStop

  6. Accountable person, before anything is settled

    Person decidesYes or no, with a probability

    Before anything is reported as settled

    Does this change note describe a change to an asset, or a broken assumption in a maintenance plan or risk entry?

    Sees only: One site's change note with image pair references

    Why: Planned: the asset manager confirms each change and decides inspection, repair and risk updates.

    Accountable: The asset manager confirms each change and decides inspections, repairs and risk register updates.

    • Yes: 0.20 or higherthenAsk a person
    • Unsure: 0.05 up to 0.20thenAsk a person
    • No: below 0.05thenAccept

The fleet: who does what

Model tiers by role, not brands: you choose the models. Strong reasoning models plan and reconcile, small fast models do the wide work, and the judge is a decision model from a different family, so it does not share the workers' blind spots.

  1. Planner

    A strong reasoning model keeps the map from assets to the maintenance plans and risk entries that depend on them.

    Decisions here:1. Scope check

  2. Workers

    A perception model for video and images, planned, one task per asset image pair, describing visible change.

    Designed for 30 to 2,000 agents, one worker task per one asset's image pair across two surveys. Each worker receives only its own unit.

  3. Judge, from a different model family

    A decision model from a different family checks that the described change is more than light, season or camera angle.

    Decisions here:2. Evidence check3. Evidence check

  4. Reconciler

    A strong reasoning model issues one change note per site and lists the dependent plans that were rechecked.

    Decisions here:4. Conflict check5. Another round?

  5. Accountable person

    The asset manager confirms each change and decides inspections, repairs and risk register updates.

    Decisions here:6. Person decides

Checked before anything is accepted

  • Every change note shows both images with survey dates
  • Differences in lighting, season and camera angle are ruled out before a change is reported
  • Assets missing from a survey are reported as not surveyed
  • Dependent plans judged unaffected are listed with the reason

What comes back

  • Change note per site with image pairs
  • Maintenance plans and risk entries that depend on changed assets
  • Assets not covered by the latest survey
  • Changes the judge could not separate from imaging differences

What to measure

  • Share of reported changes confirmed on a site visit
  • Deterioration found on site that earlier surveys already showed
  • Share of survey imagery that gets compared
  • Cost per asset per survey

Names of measures only. No result is claimed for this template.

Templates open in the workspace chat with the ask filled in. Nothing runs until you send it.

Get early accessSign in to use

More in Physical AI and field operations

Check site photos against the construction programme

For: Construction project manager or owner's representative tracking progress across sites

Progress is reported as a percentage in a spreadsheet.

Pattern: Map, verify, reduceNeeds physical AI: planned5 decisionsDesigned for 30 to 800 agents

Roll up drone inspection footage across an asset network

For: Asset integrity manager for overhead lines, pipelines or telecom towers

Drone flights return footage for every span and tower.

Pattern: Hierarchical decompositionNeeds physical AI: planned6 decisionsDesigned for 50 to 3,000 agents

Review plant walkdown video from four specialist angles

For: Site HSE manager or operations manager at a process or manufacturing plant

Walkdowns are filmed but reviewed by one person with one checklist.

Pattern: Specialist panelNeeds physical AI: planned5 decisionsDesigned for 8 to 300 agents