The operational shift is capture-plus-classification automation feeding, not replacing, engineering judgment.
AI value · Expected annual consumer savings
Company-reported
National Grid anticipates about £630,000 in annual savings.
Forecast rather than audited realized savings.
Before
Captures close-up pylon images. → Manually grades corrosion from one to six.
After
Captures high-definition images near live pylons under central control. → Automatically grades corrosion; engineers use results for maintenance and investment decisions.
Human boundary
Engineers decide asset condition, work priority, and intervention; aviation operators retain flight authority.
Why it matters
Central BVLOS drones/AI automate routine pylon imaging and corrosion grading; engineers decide maintenance.
How the work changed
Before
How the work ran before the change.
Step 1 of 2
Helicopter or drone crew
Captures close-up pylon images.
ControlAviation and live-infrastructure safety.
Step 2 of 2
Asset analyst
Manually grades corrosion from one to six.
ControlEngineering judgment.
What changed
Central BVLOS drones/AI automate routine pylon imaging and corrosion grading; engineers decide maintenance.
Decision rightAI acts within a human backstop
After
How the same work runs now.
Step 1 of 2
BVLOS drone system
Captures high-definition images near live pylons under central control.
ControlCAA permission, flight controls, and infrastructure standoff rules.
Step 2 of 2
AI model and engineer
Automatically grades corrosion; engineers use results for maintenance and investment decisions.
ControlHuman engineering authority.
Process model built from the published workflow evidence for National Grid Electricity Transmission. Every step, actor, and control appears in full below.Every step, actor, and control
Exception path
Poor imagery, uncertain grades, flight constraints, or critical findings trigger repeat capture or human inspection.
Decision authority
Engineers decide asset condition, work priority, and intervention; aviation operators retain flight authority.
Before
#
Actor
Action
Control
01
Helicopter or drone crew
Captures close-up pylon images.
Aviation and live-infrastructure safety.
02
Asset analyst
Manually grades corrosion from one to six.
Engineering judgment.
After
#
Actor
Action
Control
01
BVLOS drone system
Captures high-definition images near live pylons under central control.
CAA permission, flight controls, and infrastructure standoff rules.
02
AI model and engineer
Automatically grades corrosion; engineers use results for maintenance and investment decisions.
Human engineering authority.
Work that left the path
Much routine helicopter capture
Manual first-pass corrosion grading
Human role before
Crews captured images and analysts manually graded each asset.
Human role after
Remote pilots supervise capture; engineers validate asset health and decide interventions.
AI role
Decision mode: inspection classification. Grades corrosion from drone imagery; it does not authorize maintenance or outages.
Outcomes
Expected annual consumer savings
Company-reported
Helicopter capture and manual image assessment.→National Grid anticipates about £630,000 in annual savings.
Business adoption announced in 2025. · 21,900 steel lattice towers requiring regular inspection.
Forecast rather than audited realized savings.
Workflow maturity
Verified
Four-year innovation and trial program.→Transitioned to business-as-usual national rollout.
September 2025. · England and Wales transmission network.
Initial deployment scale and independent accuracy metrics were not disclosed.
What leaders can reuse
Anti-pattern
Publishing forecast savings as realized or implying AI authorizes maintenance.
Questions
01How is grade accuracy audited?
02What assets still require physical inspection?
03When will savings be verified?
Portability conditions
Aviation permission
High-quality imagery
Validated grading model
Engineering review
Reputation risk
medium
Evidence and authority
What the public record supports.
Current · updated
2 primary; publication outcomes are verified and reported.
Bundle 1.0.0 · reviewed 2026-09-06 · stable ID 427201159669c175