One week with a utility drone program that earns its keep every single day, not just on storm day. Dock-based drones flying daily substation inspections with reports that write themselves, distribution and transmission patrols where vision AI reads every insulator, a storm-response swarm that hands the OMS a damage map in 3 hours instead of 2 days, and the two missions almost nobody has thought of: knocking ice off transmission conductors before they gallop, and installing bird deterrents on energized towers without an outage, a climber, or a helicopter. Every flight autonomous; every finding a work order; every report filed before the pilot's coffee is cold.
| Without | With GridCORTEX | Δ |
|---|
| Without | With GridCORTEX | Δ |
|---|
The scenario is one full week, Monday through Sunday, at a fictional utility that owns good drone hardware and gets almost nothing from it: 6 docks, meaning weatherproof garages that drones launch from and recharge in without a person present, 14 aircraft, 2 licensed pilots, and a track record of 4 flights a month, each followed by a week of a human reviewing images and writing the report. Four decisions turn that hardware into a program, and a human mission commander approves every flight plan before anything takes off.
Monday morning, the first decision launches the daily routine: 6 dock-based drones inspect every substation each morning with heat-sensing and visual cameras, patrol neighborhood power lines in order of asset risk, and sweep the big transmission lines twice a week, with vision software reading every image as it arrives and reports filing themselves. The first morning covers all 8 substations before 10 AM and finds three problems: an electrical connection at the RIVERSIDE substation running 41 degrees Celsius hotter than the surrounding air, an early sign of failure; an oil seep on a NORTHGATE transformer; and the start of a bird nest at LAKELINE. Each becomes a work order automatically. Tuesday's patrols cover 31 miles of line and 6,100 images with 9 findings, including a cracked insulator that has grown 2 millimeters since March, flagged even though it still technically passes inspection standards. Wednesday, the second decision flies a mission almost no utility flies: installing bird deterrents on 12 energized high-voltage towers, where osprey nest starts would soon gain federal protection and force outages all season. Drones place 41 deterrents in 4.2 hours with no outage, no climber, and no helicopter.
Thursday evening a storm crosses the territory with 61-mile-per-hour gusts. The third decision stages a storm swarm to launch the moment winds drop: 8 aircraft sweeping assigned sectors in the dark, 140 miles of line with heat-sensing cameras and spotlights. The software classifies 2 broken poles, 11 downed wire spans, 23 trees on lines, and 1 failed transformer, and feeds each into the outage management system, the software that tracks outages and dispatches crews, as a confirmed, located job with a crew-hour estimate. The full damage map is done in 3.1 hours, before midnight, and crews roll at dawn to known work instead of rumors. Saturday, an arctic front follows. Laser measurement finds 0.4 inch of ice coating 4 spans of a 138,000-volt transmission line, with evening crosswinds likely to set the wires galloping, bouncing in growing arcs until they touch and short out. De-icing passes, using rotor wash and a mechanical knocker, clear all four spans before sunset, verified span by span. The Sunday tally: 91 flights, 62 line-miles plus 8 substations inspected daily for 7 days, 27 findings turned into work orders, a storm mapped in 3.1 hours, 12 towers bird-proofed, 4 spans de-iced, zero human climbs, zero helicopter hours, every report filed.
The docks stay closed, and the week goes the way most utility drone programs actually go. Substation checks stay monthly and manual, so the overheating connection at RIVERSIDE keeps cooking toward a summer failure that would take the whole transformer bank down. The bird work is quoted the old way, a helicopter plus a planned outage with a 6-week lead time; by then the nests are finished, federally protected, and untouchable, and the towers face a season of bird-caused outages. The storm is assessed truck by truck: at dawn the outage system still shows 60 percent of events as unassessed, crews are dispatched on guesses, and the full damage picture takes about 2 days. The iced line gets watched and hoped over, until the 22:10 crosswind sends two spans galloping into a short circuit that cuts power to 11,000 customers. The week ends at 4 flights and a 3-week report backlog. The core failure: the bottleneck was never the flying. It is the image review and the paperwork.
The software does the flying, the reading, and the filing. Mission planning is a routing optimization across 6 docks, 14 aircraft, camera payloads, battery cycles, wind, and airspace rules (use cases 16.1 and 16.2). Vision software reads every insulator, splice, and gauge as images arrive and flags only the findings that matter. The exotic missions, ice removal and installing bird deterrents on energized towers, are computed and rehearsed in physics simulation before any aircraft goes near a live wire. The storm swarm hands the outage system a classified damage map in 3.1 hours (use case 16.6, feeding use case 3.5). A human mission commander approves every flight plan and all four recommendations, and safety-critical calls stay human. The winning numbers: 91 flights, 27 auto-filed work orders, zero climbs, zero helicopter hours, and the line that failed in the last ice storm stays up.
| Measure | Without GridCORTEX | With GridCORTEX | Delta |
|---|---|---|---|
| Flights this weekautonomous missions actually flown in the seven days | 4 | 91 | a program, not a hobby |
| Substation inspectionshow often each substation gets a full camera and heat-scan check | monthly, manual | 56 (daily × 8 sites) | overheating joint caught Monday 8 AM |
| Storm damage assessmenthow long it takes to know what the storm actually broke; OMS is the outage-tracking system crews are dispatched from | ~2 days, truck by truck | 3.1 hrs, geo-fed to OMS | crews roll to KNOWN work |
| Bird deterrents (12 towers)hardware installed on live towers to keep protected birds from nesting where they cause outages | heli + outage, 6-wk lead | one afternoon, energized | the mission nobody flies |
| Ice on the 138 kV corridorice buildup on a 138,000-volt line; wind can set iced wires bouncing in growing arcs until they touch and fail | watch and hope, then gallop | 4 spans de-iced by sunset | 11,000 customers kept |
| Findings → work orderscamera findings that became actionable repair orders | image backlog, 3 weeks | 27, auto-filed with photos | the looking was the bottleneck |
| Helicopter hourshours of hired helicopter time the old methods required | patrol + bird quote | 0 | about a tenth of the cost |
| Human climbs / confined worktimes a person had to climb a tower or work in a dangerous space | routine | 0 | the safety story |
| Report backloginspection reports waiting to be written | 3 weeks | 0: filed on landing | done when the drone lands |
| Transmission corridor recordhow detailed and current the record of the big lines is; change detection compares each flight against the last | annual snapshot | tower-by-tower, change-detected | 2 mm crack caught |
| Program audit trailthe evidence trail for aviation regulators and spending reviews; Relay is the system's decision log | flight logs in a drawer | every flight Relay-traced | FAA + rate case ready |
| Where this enters the accountwhich business conversations this program belongs in | standalone robotics pitch | through Asset, Resilience, Ops | the Trojan airframe |
The two highest consequence exposures in line inspection are low altitude manned flight and climbing energized structures. Both are replaced for routine condition assessment, and the climber only goes up once the defect is already confirmed and the repair is scoped.
Counted in units you already track:
Patrol hours and desk triage hours come back to the line department, and the inspection engineer stops reading every image and starts reading only the flagged ones.
The numbers we need from you to run that formula:
| Cost driver | How it is calculated, from a rate you supply |
|---|---|
| Patrol labor | patrol hours avoided x your loaded patroller rate |
| Aviation | manned inspection hours avoided x your all-in helicopter hourly cost, whether owned or contracted |
| Engineering triage | triage hours avoided x your loaded engineer rate |
| Avoided failure | your own cost per unplanned transmission outage x the share of failures you believe earlier detection would have caught, which you set, not us |
| Vehicle and travel | road miles avoided x your fleet cost per mile, plus per diem on remote patrols |
You pay for the flight operations, whether you own the fleet or contract it, for the GridCORTEX intelligence layer that turns imagery into ranked findings, for the integration into your asset and work systems, and for your own staff time to validate detections during the first inspection season. The intelligence layer is the smaller line item. The flight operations dominate.
Payback is usually driven by aviation and patrol labor, not by avoided failures, because avoided failure is the number you will trust least. Build the case on the two you can audit and treat avoided failure as upside.
The exposure removed is a person inside an energized yard for the purpose of looking. Routine rounds put an inspector between live buses and bushings at minimum approach distance, in summer heat or on ice, several times a year per station. Qualified crews still enter for testing, sampling, and repair, under your normal switching and clearance controls, and the robot never touches equipment.
Counted in units you already track:
Inspection and drive hours come back to the substation department, and the after hours callout to look at a yard mostly stops.
The numbers we need from you to run that formula:
| Cost driver | How it is calculated, from a rate you supply |
|---|---|
| Inspection labor | yard and drive hours avoided x your loaded inspector rate |
| Callout and overtime | after hours trips avoided x your callout minimum hours x your premium rate |
| Vehicle and travel | route miles avoided x your fleet cost per mile, plus per diem on distant stations |
| Avoided equipment failure | your own cost of an unplanned station outage plus the replacement or repair cost of the unit x the share you judge earlier thermal or acoustic detection would have caught, which you set from your own failure history |
| Outage windows not taken | inspection clearances avoided x your switching labor per clearance plus your internal cost of the outage window |
You pay for the robot and its charging dock per station, for yard preparation such as gravel, ramps, and navigation markers, for enough site network to move thermal and acoustic data off the robot, for weather hardening, and for someone to recover the robot when it gets stuck. Be clear eyed about scope: the robot reads, listens, and looks. It does not open a cabinet, pull a sample for dissolved gas analysis, or operate anything, so a person still visits for those.
Payback is driven by avoided drive time and after hours callouts, which means the remote and unmanned stations pay back first and the staffed downtown station may never pay back on labor alone. Rank your stations by round trip drive time before you rank them by asset value.
The most dangerous hours in a storm are the first ones, when assessors drive roads with downed conductor, standing water, unstable trees, and no traffic control, frequently into the night to keep the count moving. Airborne assessment in post storm conditions carries its own risk. Both are the parts of the job with the least to show for the exposure, because assessors are gathering information rather than restoring anything.
Counted in units you already track:
Assessment hours come back to the damage assessors, and, far more valuably, staged crew hours stop being idle hours because the first assignments arrive sooner.
The numbers we need from you to run that formula:
| Cost driver | How it is calculated, from a rate you supply |
|---|---|
| Staged crew idle time | idle hours removed x crews staged x your storm crew day rate, including mutual assistance and contractor rates, which are the most expensive labor hours you buy all year |
| Damage assessment labor | assessment hours avoided x your loaded assessor rate, plus per diem and lodging for assessors brought in from outside the affected area |
| Aviation | manned post storm survey hours avoided x your all in aircraft hourly cost |
| Restoration duration | customer minutes shortened x your own cost per customer minute, plus any performance based rate mechanism or penalty your commission applies |
| Materials and logistics | the cost of the second material order and expedited freight that follows a bad first count, which your storeroom can price from the last two events |
You pay for the fleet and the launch teams, for pre storm staging and the readiness you carry all year and use a handful of days, for communications in an area where the cellular network is part of what the storm took down, for the GridCORTEX layer that classifies damage and builds the dispatch sequence, and for integration into your OMS crew assignment queue. On the regulatory side, budget for Part 107 operations, for beyond visual line of sight approvals if you want real area coverage, and for the coordination work with temporary flight restrictions and emergency management that follows a declared disaster. That coordination is a relationship you build before the storm, not during it.
Payback is dominated by staged crew idle hours at mutual assistance and contractor rates, because those are the most expensive hours on your books. Model it against your worst three storm years rather than an average year, because storm frequency, not efficiency, drives whether this pays.
Damage assessment is the most dangerous driving a utility does, because it happens first, on roads nobody has cleared, with conductor on the ground and energization unknown. Aerial imagery does the finding, so assessors travel to confirmed locations instead of searching for them, and they go once the map exists rather than in the dark.
Counted in units you already track:
Assessment hours come back to the damage assessors and the storm room support staff, and the damage assessment lead stops building the picture and starts checking it.
The numbers we need from you to run that formula:
| Cost driver | How it is calculated, from a rate you supply |
|---|---|
| Assessment labor | assessment hours avoided x your loaded assessor rate x events per year |
| Storm room data entry | keying hours avoided x your loaded rate for storm room support staff |
| Crew productive time | crew hours currently spent traveling to or waiting at a location that turned out to be wrong x crew size x your loaded crew rate. You set that number from your own event debriefs |
| Restoration duration | hours cut off the event x your fully loaded restoration cost per hour, including contractor and mutual assistance crews already on the clock |
| Aerial operations | a cost added rather than avoided: flight hours per event x your all in aerial cost per hour x events per year |
You pay for the aerial imagery collection, which you may already be buying, for the vision service that turns imagery into ranked damage points, for the outage management system and GIS integration, and for your own assessors' time validating detections through the first two or three events. Imagery collection is usually the largest line item and the integration is the one that takes longest.
Payback is driven by assessment labor and by hours cut off total restoration duration, because those two are visible in your event cost accounting. Build the case on those and treat recovered crew time as upside, since misdirection is the hardest number to audit.
What is this, exactly? It is AI software: intelligent agents and models built and delivered by SoftServe, running on NVIDIA accelerated computing. It is not a hardware appliance and it does not replace the systems you run today. It deploys in your own cloud or on your premises, connects read-only to your existing systems, and recommends; your people approve every action, starting in shadow mode until it earns trust.
An autonomous inspection service for transmission asset managers: drones fly the corridor on a schedule and deliver a prioritized, georeferenced defect list with imagery evidence as a repair queue for maintenance planners. The demo above uses synthetic data; everything below describes what the real deployment needs from your organization.
| Your system | Typical products | How we connect |
|---|---|---|
| Geographic Information System (GIS) | Esri ArcGIS Utility Network, GE Smallworld | read-only API |
| Asset / work management (EAM/CMMS) | IBM Maximo, SAP PM, Oracle WAM | read-only API; recommendations are written back only after a person approves, via your existing system's own interface |
| Document and knowledge stores | helicopter patrol reports, inspection imagery archives | document upload |
| Weather and environment | National Weather Service feeds, satellite and LiDAR imagery | read-only API |
| SCADA historian | AVEVA PI System, GE Proficy | historian mirror (one-way feed) |
| Outage Management System (OMS) | GE PowerOn, Oracle NMS, ADMS outage module | read-only API |
| Field and crew systems | crew scheduling (ARCOS), vehicle location (AVL) | read-only API |
AI runs on the drones at the edge; imagery is processed on GPU instances in your own cloud account, with read-only connections through your existing data zone and no connection to control systems. Flights operate under your aviation program with FAA approvals.
The Approve button you just clicked in the demo above is the real workflow. This is what it looks like on the screen of a transmission maintenance planner in the GridCORTEX console:
Approve creates draft work orders in your EAM through its API in its native pending status, with GIS locations and imagery attached; your maintenance planner prioritizes and releases them under normal scheduling rules. GridCORTEX never dispatches crews.
Defects flow automatically from the drone sweeps; planners can reject or reclassify a defect with one click.
Scores come from the latest sweep and each defect card shows its flight date; weather feeds update every 15 minutes for flight planning.
Lives as a defect queue in the EAM with imagery in the GridCORTEX console; critical-scored defects push to email and mobile. The console runs in a browser beside your existing screens on day one; embedding into your own systems is a roadmap step once the read-only phase has earned trust. Approve, Modify, and Decline are all captured in an audit trail your compliance team can pull, and GridCORTEX never blocks or overrides anything in the systems you run today.
The fair question from any T&D ops or safety leader: "We already own drones and a couple of licensed pilots, what's new here?" Here's the honest answer, and it's why most utility drone programs plateau at four flights a month.
When someone asks "what did it actually calculate?", this is the list. In the simulation these factors drive the storyline; in a pilot they are computed from your asset registry, imagery history, and airspace.
Presenter's one-liner: "Ninety-one flights in a week: every substation daily, sixty line-miles read frame by frame, a storm mapped in three hours, ice knocked off four spans before they galloped, and bird deterrents on twelve energized towers without a single outage or climber. The drones were the easy part; the AI that plans, reads, and files is the program."