Artificial Intelligence is all the rage. Now it’s transforming quality document management for aviation.
Say you buy a used 2015 Toyota Corolla for $12,500. After you complete the transaction, how important is the paperwork that comes along with that sale? Oil changes, repair shop work, body work—do you need to keep all that paperwork in case you re-sell that Corolla again in the future? Probably not. What about if you’re selling your home and put it on the market? Do you have a shoe box stuffed with paperwork that tracks household repairs or renovations? Again, not likely. In fact, paperwork in most all industries—including automotive and real estate—is not essential to maintain the value of the product. Put more simply: I can still sell a car or a home without ALL of the paperwork that gives the past history of the product.
Enter the aviation industry, where paperwork and trace documentation for “used” parts is literally king. I’ve seen an entire aircraft transfer from one party to another—and with that ownership transfer came two gigantic pallets of paperwork. Like, literally, 20 trees worth of paper. Obviously, selling a multi-million dollar aircraft is going to be complex but let’s take it down to the simple used component scenario. Even if you sell something as boring as a used tube assembly, it’s going to come with various paperwork such as an 8130 tag (which confirms it’s safe for flight) or a shop visit report. What if you don’t have the paperwork that goes along with it? Just like the scene from Frosty the Snowman where they were denied a train seat to the North Pole and were told, “No money, no ticket!” it’s the same outcome here. No paperwork, no money! If the owner of the part doesn’t have the paperwork that goes with the part, the value of that part is basically zero, nada, or zilch.
So, we’ve established that paperwork es muy importante but what does Artificial Intelligence have to do with transforming the way we manage aviation quality images and files? Remember above where I mentioned that a single aircraft can be tied to pallets of paperwork? Imagine the thousands of aircraft flying in this world and just envision the paperwork tied to these aircraft and its components. Again, that’s a lot of trees. But also imagine the technology and organizational labor involved to keep those documents organized. Documents are scanned and saved but like many manual processes, errors are prone to happen. Are documents really being saved with the correct part number and serial number information? Is that 8130 tag really an 8130 tag or should it have been classified as an EASA certificate? We believe that Artificial Intelligence is the key to detect errors and transform the way aviation document management is done.
GoDirect Trade’s conundrum
On GoDirect Trade, we have 100+ sellers and over two billion dollars in quality inventory listed. For every part listed, we have multiple product images and up to 13 types of aviation related documents loaded. In essence, we are constantly managing and uploading thousands of quality documents and product images. When we first went live, we constantly received feedback from customers related to listing quality errors. “FYI, that is NOT the 8130 tag. You’re showing a CAAC certificate for a completely different part number.” Or, my favorite: “Hey, that picture of the valve you’re selling is actually a black box recorder. Are you blind?” Suffice to say, we knew that we had a big problem. We had a gigantic pool of digital files and we had to depend on the sellers to accurately manage these files against their listings. We couldn’t go back and tell sellers to go through each document and image and check for errors—that would have required far too much friction in the user experience. So, what was the best way to solve this problem?
Enter Artificial Intelligence
Rather than place the burden of document accuracy on our seller community, our team invented Alfred, an AI-assistant on GoDirect Trade who identifies potential file inaccuracies in real-time. The technology is simple: Sellers upload images and quality documents and Alfred provides feedback such as, “Hey, your product image is named with serial number AA2 but it should be AAZ,” or, “The 8130 tag you uploaded is actually an EASA certificate with part number 3879800-2, not 3879800-3.” So, instead of making our sellers review their documents with a fine tooth comb, our marketplace does it instead, but with a high powered pressure washer. Another cool thing about Alfred is that he uses a single model methodology when he’s analyzing files. In the beginning, we tried to create separate models for each file type Alfred was trying to review (ex. one for 8130's, one for data plates, one for EASA's, etc). As you can imagine, this just screamed “too complex” so we pivoted to the single model. This allows Alfred to review just about any file type and then provide feedback based on the algorithms he continuously trains with. The image below gives a good visual on Alfred’s Big Brain solution.
Is this really a transformative technology?
There are many companies today that use Optical Character Recognition (OCR) for various things. In the aviation space, it’s primarily used for categorization and organization. At GoDirect Trade, we’re building on that foundation and using OCR and AI to drive decisions and actions. And in an industry where paperwork is king and accuracy is of utmost importance, we think this technology is just scratching the surface on the future potential of document management.
Let’s Talk Trade,