Dana runs quality at a mid-sized maintenance shop outside Phoenix. She’s made up, but her worst week was real enough for a lot of people in this industry. In the summer of 2023, her inbox filled up with alerts about engine parts sold by a London broker called AOG Technics. Over four years, the company sold tens of thousands of CFM56 engine items with falsified paperwork. Those engines power most 737s and A320s. Regulators in the US, UK, and EU put out warnings, and airlines from American to Ryanair grounded aircraft for checks.
Her shop had bought nothing from AOG Technics. It still took her team the best part of a week to prove that, digging through boxes of release forms and spreadsheets two different people had kept in two different formats.
The director, Jose Alejandro Zamora Yrala, was jailed for four years and eight months in February 2026. By then Dana had rebuilt her traceability program from the ground up, and a fair bit of AI had crept in along the way. Here’s what she’d tell anyone starting the same job.
1. Fix the records before you buy the software
Every part needs a history covering where it was made, who released it, and every repair since. That usually means an FAA Form 8130-3 or EASA Form 1, plus inspection reports and repair paperwork. Dana’s shop had most of this, just scattered across filing cabinets and three versions of the same spreadsheet.
Document-reading AI earned its keep here. It pulled part numbers, serials and dates off thousands of scanned forms in a weekend, work that would have taken her clerks months. The catch showed up on the first batch. Where a scan was smudged, the model happily filled in a plausible-looking serial number.
So she changed the rule. The software extracts only what it can read, marks anything uncertain, and a person checks every flagged field. It’s slower, but nobody ends up installing a part on the strength of a number a machine made up.
2. Check the paperwork like it might be fake
That week made Dana a lot more suspicious. Before AOG Technics, a certificate that looked right usually counted as right. Now her team checks that each part is eligible for the exact aircraft and configuration, and that it came through an approved supplier or authorized channel.
They also have an anomaly-detection tool. It compares each new certificate against hundreds of earlier ones from the same repair station. A different font, a signature that doesn’t match or a serial number already logged on another aircraft all get flagged.
Most flags turn out to be nothing, often just a station that changed its form template. Dana’s inspectors still decide every one and write down why. “The tool asks the question,” she says. “We’re the ones who sign.”
3. Follow the part for its whole life
Buying a part is only the start. Her system logs installation dates, hours, cycles, inspections, repairs, mods and removals, right through to retirement, which lets planners see how a part has actually been used and swap it out before it hits a limit.
The predictive maintenance model sits on top of that history. It reads trends in wear and sensor data and suggests when an inspection is due. On parts with years of clean records it’s genuinely useful. On the odd part with a gap in its history, its suggestions get vague, which is roughly what you’d expect.
Don’t forget the jacks
Dana’s auditor caught something she’d missed. The shop tracked every rotable part on the aircraft, but the ground equipment lifting those aircraft had almost no history at all.
Take a hydraulic tripod jack like the Tronair 02-0511C0132. It’s a 5-ton multi-stage unit rated at 10,000 lb, used at forward fuselage and main wing jacking points. The reseller lists a Certificate of Conformance and an FAA 8130-3 as available with it, and that’s paperwork worth keeping.
Her team now gives every jack, cart, and stand the same treatment as an aircraft part. Each gets a serial number in the asset system, its conformity documents stored, and every pressure test and inspection logged. If something ever goes wrong under a wing, that’s the first history anyone will ask for.
4. Know your suppliers’ suppliers
AOG Technics sold through brokers to other brokers, and plenty of buyers had no idea where their parts had started out. That’s the gap Dana worries about most.
She now treats supplier vetting as proper third-party risk management. New vendors get asked how they verify what they buy, and anyone who gets cagey about it drops off the approved list.
The shop also runs incoming parts against recall notices, service bulletins and known suspect batches automatically. When the next alert lands, she reckons it’ll take an afternoon to find any affected stock, rather than a week of boxes.
5. Be careful with the chatbot
The newest feature in her maintenance platform lets technicians type questions in plain English, like “every repair on this actuator since 2019.” It’s brilliant when it works. In testing, though, it once summarized a repair that had never happened, stitched together from two similar part numbers.
Now every AI answer has to link to the source document it came from, or it doesn’t count. The team spent an afternoon learning how to check AI-generated content, which mostly comes down to never trusting a summary you can’t trace back.
Barcodes and RFID tags have caused far less drama. They just cut down typing errors, and that’s all anyone wanted from them.
6. Make audit prep boring
Dana used to dread audits. These days, a background check runs every night, looking for missing signatures, expired certificates, and serial numbers that don’t match. Problems get fixed while the technician who did the job still remembers it.
Auditors have started asking a new question too: which AI tools touched these records, and where do humans sign off? She keeps a one-page answer ready.
7. Write it down and train everyone
Her shop now has one written procedure for receiving, inspecting, documenting, storing, tracking, and retiring parts, and buyers, store staff, technicians, and inspectors all work from the same document.
Training covers the AI tools as well, including what they look for and when to overrule them. The hardest lesson was about false alarms. Early on, the anomaly tool flagged so much harmless noise that people started clicking past it. Dana had it returned and sat the team down to explain why ignoring a flag is the one thing that could really hurt them.
She still keeps one AOG Technics alert pinned above her desk. The forged forms got past plenty of experienced buyers, and the people who finally caught them weren’t using any clever software. They were just inspectors who looked at the paperwork and asked who had actually signed it.
Related: Apple’s Free On-Device AI Has a Catch: Which iPhones Are Left Out?
