Skip to content
HoldField

Physical AI · Automation ladder

Evidence, from capture to go/no-go

Five stages, one controlled chain. Each stage captures or proves something, hands a human the evidence, and keeps the PLC and safety circuit authoritative. AI Sense observes across all of it — and commands none of it.

Stage 01

Edge AI inspection

Capture, detect, measure, and prove defects at the station.

Inputs

  • Station camera frames
  • Part under inspection
  • Inspection recipe

Proof generated

  • Frame reference
  • Overlay reference
  • Defect measurement
  • Inspection receipt
  • Evidence bundle

Where it appears in the app

  • Inspect
  • Continuous
  • Review
  • Evidence

AI Sense support

  • Finds missing evidence
  • Explains low confidence
  • Detects repeated defect patterns
  • Flags review bottlenecks

Safety boundary

  • Cannot force PASS, clear recovery, or bypass station gates.
Explore Edge AI inspection →

Stage 02

PLC-connected routing

Send governed quality results to the line without bypassing the PLC or safety authority.

Inputs

  • Governed inspection verdict
  • PLC route policy
  • Safety-chain state

Proof generated

  • PLC signal evidence
  • Route agreement receipt
  • Governance decision
  • Commissioning blocker state

Where it appears in the app

  • Governance
  • Commissioning
  • Stations
  • Ops metrics

AI Sense support

  • Finds missing PLC evidence
  • Explains routing risk
  • Flags governance gaps
  • Detects authority rejection spikes

Safety boundary

  • Cannot write PLC outputs or override safety authority.
Explore PLC-connected routing →

Stage 03

Motion-assisted coverage

Use indexed presentation to prove the required views. No coverage, no PASS.

Inputs

  • Rotary or indexed fixture position
  • Required-view checklist
  • Captured views

Proof generated

  • Coverage proof
  • View completeness
  • Evidence bundle
  • Vision Twin comparison

Where it appears in the app

  • Vision Twin
  • Review
  • Work packages

AI Sense support

  • Detects repeated incomplete coverage
  • Ranks station coverage risk
  • Generates human review checks

Safety boundary

  • Cannot mark a station safe or approve a PASS without evidence.
Explore Motion-assisted coverage →

Stage 04

Robot-assisted handling

Use commissioned skills for bounded handling. No arbitrary robot motion.

Inputs

  • Commissioned robot skill
  • Part presentation request
  • Bounded motion envelope

Proof generated

  • Commissioned skill reference
  • Robot boundary evidence
  • Station capability status
  • Safety receipt

Where it appears in the app

  • Stations
  • Commissioning
  • Vision Twin
  • Governance

AI Sense support

  • Explains readiness blockers
  • Flags missing commissioning evidence
  • Prepares engineer work packages

Safety boundary

  • Cannot issue natural-language robot commands or arbitrary motion.
Explore Robot-assisted handling →

Stage 05

Controlled deployment

Run supervised deployments with evidence-backed go/no-go readiness before any station is promoted.

Inputs

  • Shadow-mode inspection stream
  • Uncertainty queue
  • Review decisions
  • Drift signals

Proof generated

  • Deployment metrics
  • Review analytics
  • False-reject candidates
  • Possible escape candidates
  • Drift summary
  • Go / no-go packet

Where it appears in the app

  • Review analytics
  • Work packages
  • Commissioning
  • Ops metrics
  • SignalOps

AI Sense support

  • Ranks deployment blockers
  • Finds missing evidence
  • Explains drift risk
  • Prepares support and QA handoffs

Safety boundary

  • Cannot approve production or hide escape candidates.
Explore Controlled deployment →

AI Sense

One reading layer across every stage

Observes evidence, finds missing proof, explains uncertainty, ranks human checks, and prepares handoffs — it never commands hardware.

Reads

  • Evidence bundles
  • Review events
  • QA decisions
  • Vision Twin drift
  • Commissioning blockers
  • Governance decisions
  • Station registry
  • Ops metrics

Produces

  • Findings
  • Evidence-gap warnings
  • Work-package hints
  • Commissioning questions
  • Support summaries

Never

  • No PLC writes
  • No force PASS
  • No recovery clear
  • No robot commands
  • No camera/light commands
  • No production approval
  • No evidence mutation
  • No QA decision mutation

AI Sense observes evidence and guides humans — it records nothing and changes nothing. It does not command a station, write a PLC, clear recovery, reset safety, force a pass, approve production, sign off, or mutate any review, QA decision, commissioning, governance, evidence, or runtime state. Every recommendation is a suggestion for a human to carry out; the PLC and safety circuit remain authoritative.

Signed-in teams see these capabilities operationalized in the HoldField app, under Capabilities. Open the workspace →