From Mill Cert to Tail Number: Automating Supplier COAs and SAP QM Results Recording for AS9100 Manufacturers

A titanium forging shows up at the dock with a mill certificate, a NADCAP heat treat certification from a sub-tier processor, and a supplier certificate of conformance clipped on top. Three documents, three formats, three different ways of describing the same chemical composition and mechanical property values. The receiving inspector reads all three, checks the numbers against the engineering drawing and the governing AMS specification, then keys the results into SAP QM before the part can move to the machine shop. If one characteristic falls outside tolerance, or if the NADCAP accreditation on that heat treat certificate quietly expired last month, the part should never make it to the floor. Whether it actually does often comes down to how carefully one inspector read one PDF on a Tuesday afternoon between two other jobs.

This is a different problem than the one Artificio addressed in the piece on pharma GxP document automation. Batch records and deviation reports in pharmaceutical manufacturing revolve around data integrity under 21 CFR Part 11, tight release timelines, and getting a batch out before a shelf-life clock runs out. Aerospace quality data is shaped differently. A single machined part can carry paperwork from four or five supplier tiers, each responsible for one segment of a genealogy chain that has to stay intact from a raw material heat lot all the way to a serialized part installed on a specific aircraft tail number. The pressure here is not a submission deadline. It is AS9100 clause 8.4, control of externally provided processes, products, and services, combined with a standing obligation under AS5553 to keep counterfeit and suspect parts out of the supply chain before they ever reach an assembly line.

Why the Aerospace Paper Trail Works Differently

Ask a quality engineer at an aerospace machine shop what lands on their desk in a given week and the list looks nothing like a batch record folder. A certificate of conformance from the raw material supplier states that the material meets the purchase order requirements and the applicable specification, but it rarely includes the actual test data behind that claim. A material test report, sometimes called a mill certificate, carries the real numbers: chemical composition percentages, tensile strength, yield strength, elongation, hardness, each tied to a specific heat or lot. A special process certificate from a NADCAP-accredited heat treater, plater, or NDT provider confirms that a specific operation was performed within an accredited scope and according to a specific process specification. None of these three documents look alike, none of them come from the same supplier, and all three have to reconcile against the same part before an inspection lot in SAP QM can move toward a usage decision.

Add supplier tiers and the genealogy problem compounds. A forging house buys billet from a mill, which cites a heat lot number tied to the mill's own test data. The forging house performs its own operation and issues a certificate that references the mill's heat lot. A NADCAP-accredited heat treater processes the forging and issues a certificate referencing the forging traveler. The machine shop receives the forging, machines it, and has to produce a certificate of conformance that traces all the way back to that original heat lot, because a customer audit or a fleet-wide corrective action five years later might need to identify every part made from that exact heat of material. Pharma traceability generally stops at a batch number and a manufacturing date. Aerospace traceability has to survive four or five ownership changes and still point back to a single furnace charge. 

None of this shows up as clean, structured data. Every supplier formats certificates differently. Some still fax paper documents that get scanned into PDFs with skewed pages and handwritten values in the margins. Units vary. Some report tensile strength in ksi, others in MPa. Specification callouts get abbreviated inconsistently, AMS 4928 might appear as "AMS4928," "AMS 4,928 Rev K," or just "4928." A human reviewer has to normalize all of that before a single value can be compared against an engineering tolerance, and normalizing it by hand at volume is exactly where errors creep into a system with almost no tolerance for them.

Where Document Intelligence Actually Fits

Artificio's document intelligence layer reads incoming certificates the way an experienced quality engineer would, except it does not get tired by the fourth document of the morning. The system ingests certificates of conformance, material test reports, and special process certifications regardless of layout, extracts the chemical and mechanical property values along with the specification references, heat or lot numbers, and supplier identifiers, then normalizes units and specification formatting before anything touches SAP.

That normalized data does not go straight into a results recording transaction. It passes through a validation layer first, built specifically around what AS9100 environments actually require. The layer checks the extracted values against the master inspection characteristics tied to the part's inspection plan in SAP QM, confirms the values fall within the tolerance defined by the governing engineering drawing or AMS specification, verifies that any cited special process certification comes from a supplier currently listed on an approved supplier list with an active NADCAP accreditation scope covering the specific process performed, and traces the heat or lot number backward through whatever genealogy records already exist in the system. Only after all four checks pass does the data post into the inspection lot as a results recording entry, ready for a usage decision.

The connection to the quality info record matters here too. In SAP QM, a quality info record links a specific vendor and material combination and determines whether goods receipt inspection triggers at all, along with which certificate types that vendor is expected to submit. When a certificate arrives, the validation layer checks it against the active quality info record for that vendor-material pair, confirming the right document type showed up and that the vendor still holds the release status the info record expects. A vendor whose quality rating has slipped, or one operating under a skip-lot arrangement that just lapsed, gets caught here rather than three steps later.

Results Recording at the Characteristic Level, Not the Document Level

Most manual COA processing treats a certificate as a single object. Someone reads it, someone decides it looks fine, someone files it. SAP QM was never built to work that way. An inspection lot in SAP QM breaks a part down into individual master inspection characteristics, each tied to a specific dimension, material property, or process requirement pulled from the inspection plan. A titanium forging might carry a dozen or more separate characteristics: tensile strength, yield strength, percent elongation, grain size, individual alloy element percentages, each with its own upper and lower specification limit drawn from the engineering drawing and the referenced AMS specification.

Automating results recording means matching extracted certificate values to the correct characteristic, not just to the correct part. Artificio's validation layer maps each value pulled from a material test report to the specific master inspection characteristic it corresponds to, checks it against the tolerance band, and records the result through the same OData services and BAPI calls that a manual results recording transaction would use. Characteristics that pass populate cleanly. Characteristics that fail, or values the system cannot confidently extract, get flagged and routed to a quality engineer for review rather than posted blind. Nothing skips the tolerance check just because the extraction happened faster than a human could read the page.

The downstream effect matters more than the speed gain. A usage decision in SAP QM depends entirely on whether the recorded results support acceptance. When every characteristic posts accurately and immediately after the certificate arrives, the usage decision can generate the same day the material hits the dock rather than sitting in a receiving queue while an inspector works through a backlog. Parts move to the machine shop faster, and the audit trail behind that movement holds up to scrutiny because every characteristic traces back to the specific line on the specific certificate that supported it.

Special Process Verification Cannot Be a One-Time Check

NADCAP accreditation is not permanent, and it does not cover every process a supplier performs. A heat treater might hold accreditation for vacuum heat treating but not for a specific alloy family, or their accreditation for a particular process might have lapsed during a renewal cycle without anyone downstream noticing. Manual verification usually happens once, when a supplier gets added to an approved supplier list, and rarely gets rechecked against every certificate that arrives afterward. That gap is exactly where an aerospace quality system becomes vulnerable, because a lapsed accreditation on a special process certificate means the process was performed outside a controlled and monitored environment, whatever the paperwork claims.

Document intelligence closes that gap by checking accreditation status against the current scope every time a certificate references a special process, not just once at supplier onboarding. When a NADCAP certificate cites a specific process specification, heat treating per AMS 2750, chemical processing per a specific AMS or military specification, non-destructive testing per a customer-specific procedure, the validation layer confirms that the issuing supplier's accreditation currently covers that exact process and that the certification date falls within the accredited window. A mismatch does not get filed away and forgotten. It gets flagged before the associated results ever post to the inspection lot, giving a quality engineer the chance to hold the part and investigate before it moves any further down the value stream.

This same verification logic extends naturally into sub-tier oversight, which AS9100 clause 8.4.3 explicitly expects an organization to maintain even when it did not select the sub-tier supplier directly. A forging house choosing its own NDT provider still leaves the prime contractor accountable for verifying that provider's accreditation and process control. Automated verification at the certificate level gives an organization visibility into sub-tier compliance that a purely manual review process, stretched across dozens of suppliers and hundreds of certificates a month, realistically cannot sustain.

Source inspection adds a government or customer representative to the same certificate review, and that representative usually shows up with far less patience for an incomplete paper trail than an internal reviewer might have. A DCMA in-plant quality assurance representative, or a customer source inspector on a commercial program, expects to walk into a receiving area and see a complete, reconciled certificate package the moment they ask for one, not a promise that someone will pull it together by the end of the week. When results recording, characteristic verification, and NADCAP scope checks already happened automatically at the moment the certificate arrived, a source inspection sign-off turns into a review of data that already exists rather than a fire drill assembled for the visit. Export control markings on the certificate itself, ITAR statements in particular, get carried through the same extraction pass so a document missing a required marking gets caught before it becomes a compliance question raised by someone outside the building.

Counterfeit Parts and the Genealogy That Has to Survive Every Handoff

AS5553 exists because counterfeit and suspect parts have entered aerospace supply chains through exactly the kind of documentation gaps this workflow is built to close. A certificate that cannot trace back to a verifiable heat or lot number, a supplier that cannot produce a chain of custody connecting its material to an original mill test report, a specification reference that does not match anything in an approved parts list, each one is a signal that should stop a part rather than let it proceed on the strength of a document that merely looks official.

DFARS 252.225-7009 adds another layer specific to specialty metals, requiring that certain materials in defense articles trace to melt and pour within a qualifying country. Verifying that manually means a quality engineer cross-referencing country of origin statements across a chain of certificates each written by a different supplier with a different level of diligence about that requirement. Automating the check means the validation layer looks for melt and pour statements, cross-references country of origin claims against the DFARS requirement tied to the specific contract, and flags any certificate where that information is missing or inconsistent with material further up the genealogy chain.

Genealogy tracking connects directly into SAP PP as well, not just SAP QM. Batch and serial number genealogy functionality in PP lets an organization link a finished, serialized part back through every production order, every sub-assembly, and every raw material lot consumed along the way. When document intelligence extracts a heat lot number from an incoming mill certificate and posts it against the correct material and batch in SAP, that number stays linked through every subsequent production step, right up to the serialized end item. A fleet-wide inquiry five years later, asking whether part serial number X traces back to heat lot Y, becomes a query against existing SAP genealogy data instead of an archive search through boxes of paper certificates.

First Article Inspection Runs on a Different Clock

First article inspection under AS9102 is a related workflow but a distinct one, and it deserves its own mention because it gets confused with routine results recording more often than it should. A first article inspection report validates that a supplier's production process can consistently produce a part conforming to every requirement on the engineering drawing, not just the requirements captured as inspection characteristics in a routine QM inspection plan. The report itself runs across three forms: part number accountability, product accountability covering raw material and special process certifications and functional testing, and characteristic accountability tied to a ballooned drawing where every dimension gets a number matched to a corresponding measured value.

That third form alone can carry a hundred or more individual characteristics on a complex machined part, each one requiring a cross-reference between a balloon number on a drawing and a measured value on a data sheet, often submitted as a scanned, ballooned drawing alongside a separate spreadsheet of measurements. Document intelligence applied to first article inspection means matching balloon numbers to the correct rows in a measurement table, checking each measured value against the drawing tolerance, and flagging any balloon number that appears on the drawing but has no corresponding measurement, or any measurement with no matching balloon. It is a heavier lift than routine results recording because the document volume per part runs so much higher, but the underlying validation logic, extract, match to a defined requirement, check tolerance, flag exceptions, follows the same architecture used for routine supplier certificates.

What Changes on the Floor and in the Audit Room

The most immediate change shows up at receiving inspection, where parts stop waiting on a document review queue and start moving as soon as automated validation clears them. The effect that matters more over time shows up during an AS9100 surveillance audit or a customer source inspection. An auditor pulling a random serialized part and asking for full genealogy back to raw material used to trigger a scramble through supplier files and inspection lot history. When every certificate has already been validated, matched to a specific characteristic, and linked through SAP genealogy at the moment it arrived, that same request becomes a straightforward query rather than a research project.

Nonconformance handling improves for a related reason. When a characteristic fails during automated results recording, the exception routes immediately to a quality engineer with the specific out-of-tolerance value, the specification limit it violated, and the source certificate already attached, instead of surfacing days later once someone finally works through a backlog of unread documents. Material review board dispositions happen faster because the investigation starts with complete information rather than a request to go find the original paperwork first. A disposition of use-as-is, rework, or return-to-supplier gets decided on facts already sitting in the system rather than facts someone has to go hunting for.

Integration with production planning closes the loop further. SAP PP order-specific inspection points, first-off, in-process, last-off, can trigger the same validation logic against in-house inspection data that gets applied to supplier certificates, so a shop's own measurement data gets checked against tolerance with the same consistency as material coming through the dock. Vendor evaluation scores in SAP improve in accuracy too, since they now reflect certificate quality and characteristic pass rates captured automatically rather than a quarterly manual review that mostly relies on memory. The genealogy chain that starts with a mill certificate and a heat lot number carries all the way through heat treat, machining, assembly, and final inspection without a single manual handoff where a number might get mistyped or a certificate might get filed in the wrong folder.

AS9100 itself is not standing still. The International Aerospace Quality Group is working toward a revised standard, expected to be rebranded IA9100 and aligned with the next ISO 9001 update, with counterfeit parts prevention and supplier control likely to get even more explicit attention than they already receive under the current Rev D. Organizations that already run automated, characteristic-level verification across their supplier document workflow will not be scrambling to build that capability when the revision lands. They will already be running it, and the transition will read more like a documentation update than an operational overhaul.

The forging that showed up at the dock this morning, the one with three certificates in three formats and a dozen characteristics that all have to check out before it reaches a machinist's hands, does not need a faster inspector. It needs a system that reads every one of those documents the same careful way every single time, catches the mismatch a tired reviewer might miss on the fourth certificate of the afternoon, and keeps the genealogy chain intact from a furnace in a mill somewhere to a specific serial number on a specific aircraft. That is a different problem than pharma batch release, and closing it takes a validation layer built around the specific things AS9100 actually asks an organization to prove, not a generic document automation tool pointed at a new industry.