Final Thoughts on Safety and Procedures

On the front lines, you work around the clock to deliver safer and more reliable service, all at a lower cost. Since twenty twenty four, PG and E's residential electric rates have gone down five times, and for their most vulnerable customers, they've cut rates by twenty-three percent and they're still hard at work. That's PGE on the front lines for us all. Residential bundled electric rates have decreased five times since January 2024, reducing rates 13% since then for customers who get both energy supply and delivery from PGE. For PGE's most vulnerable customers, residential bundled electric rates are 23% lower than in January 2024. This message was paid for by shareholders. Hey mom, now that I have an after-school job, I think I want to buy a drum kit. Okay, Rockstar. Let's look at your account. When kids start making money, Chase has easy-to-use tools and expert bankers to help them learn how to manage it. And next month, I'll buy the bass drum and assembles. Oh, and maybe some noise canceling headphones for you. Aw, you've thought of everything. Help kids save, budget, and build financial independence all with Chase. That's good for k ids and good for parents. Visit your local Chase branch and get started. Accounts subject to approval. JP Morgan Chase Bank and a member FDIC. When I found out I was gonna be a parent, I immediately felt a lot of anxiety and worry. So I went on to BetterHelp to try to look for a therapist to help me with that. My relationship with my family and with my boyfriend and with myself were suffering. I really needed help. I was ruminating a lot, really getting those thoughts out to a therapist and getting feedback was just life-changing. Discover what BetterHelp Online Therapy can do for you. Visit betterhelp.com today . Avoiding your unfinished home projects because you're not sure where to start, ThumbTech knows homes, so you don't have to. Don't know the difference between matte paint finish and satin or what that clunking sound from your dryer is. With ThumbTech, you don't have to be a home pro. You just have to hire one. You can hire top-rated pros, see price estimates, and read reviews all on the app. Download today. Parents, have you heard your kids say, I'm not a math kid? Well with mathasium, every kid can be a math kid. They customize their math instructions so kids who are struggling are able to catch up and get ahead. And advanced kids are challenged to reach higher. Mathnasium makes math fun, so kids learn to love it. Parents say that Mathnesium has not only improved their kids' grades, it's given them a new level of confidence. In math and in school overall, visit Mathnasium.com to find a location near you . At the five five seven contact tower one two eight point one five. Caution caution. Manual fuel. Manual fuel . I'm John Golia . I'm Greg Fife. And I'm Todd Cruz . And we are the flight safety detectives. Between us, we have over a century of aviation accident investigation and safety experience to draw on as we discuss issues that affect all of us. So we are qualified to share our perspectives on accidents and incidents and what can be learned from them for the future. We're proud to say that we have two sponsors that really relate to the topic of aviation safety, the Professional Aviation Maintenance Association or PAMA, and a Vemco Insurance. Later on in the show, we'll tell you how you can get a 5% discount on your insurance just for listening to the show. We don't just dissect the official reports. In every episode, we identify safety issues and take the mystery out of accident investigations. So maybe pilots in their planes can have safer flights ahead . Well hello Todd, how are you today? Not bad at all. As we uh do the show it is uh thirty first of May, the last day of May. And uh well we're about to go into summertime where all kinds of traveling is going on and that's a good thing. But we're gonna be talking about a variety of things that have nothing to do with summer travel, but have everything to do with emerging issues in aviation safety. And before the show, uh John and I were talking about one of those emerging issues. Commentary out there in the Internet and elsewhere about the ongoing investigation of the UPS crash in Louisville. And uh let me let you take it from there, John. been on and in uh throwing out all sorts all kinds of an analysis of of uh UPS and other accidents. And on the UPS accident he has some from the very beginning claimed that the number two engine , which obviously uh was degraded , uh otherwise the airplane may have may have been able to get in the air, but claiming that the number two engine was degraded because of Flawed, FOD, foreign object dig estion. All it means is the engine was contaminated with bits and pieces that came into the engine and destroyed some of the blades and degrad ing the power . But if one loose and back and looks and listens to NTSB board member Todd Inman at the time, he has said publicly that the number two engine was not damaged by Ford. And I believe he was responding to a question from a reporter or somebody. And uh but he had said I heard it twice from different angles from him . So and it one of them may have been he also adjusts my class uh at Vaughn College. So in any event he's publicly stated that the engine was not farted and that's a disservice to everybody out there if in fact uh he continues to say that. fact checking, you cannot just uh speculate and then rely upon that speculation over and over. You know, we get caught sometimes ourselves with the speculation. We try to avoid it at all costs, but it's it's easy to get into . And it's really uh a disservice to the f flying community out there and particularly the pilots that just keep saying it's flooding because there's not much a pilot can do about fraud . It's just uh somebody else has to take care of the condition of the taxiways and the runways to keep the those objects out of your airplane engine. So I just wanted to to just point that out to everybody to be careful when people uh including us say things as facts uh that they aren't that they're just information and you should do your own job of of verifying. Trust but verify . Well that is a perfect segue into the main topic today. Uh talking about uh the use of artificial intelligence in aviation safety. Why are we talking about this? Because I've been using artificial intelligence in aviation safety. And I'm cutting to the middle of the story here. What you just said about talking about aviation incidents in general also works for AI. It is tempting to turn over your thinking to AI. It's tempting to think, oh my gosh, AI can do what I'm doing so much better. No. It does what you do so much differently. And your judgment and intelligence and experience are going to be key to using AI effectively and efficiently. Now , you think yo tourself , okay, what are you talking about here? Well, let me give you an example. I've been working for years with large amounts of aviation data. And typically when I talk about things on this podcast, I like to stick with publicly available data, NTSB reports and other things that are out there that anybody in the world could look at themselves and use their own analyses to either confirm or deny what I'm saying. And one of my favorite resources for years has been NASA's Aviation Safety Reporting System, which, for those of you who don't know, actually we've done several shows on ASRS and I'll put a uh I'll flash up one of the um one of the uh posters from that show so you can go back and get an idea of what that's about from our perspective. But what this database has is reports from people in the field, pilots, mechanics , air traffic controllers, flight attendants, ground personnel, even passengers, doesn't matter, of issues that might be of concern in aviation safety. Not accidents, not things that are officially reported or investigated by the NTSB, but other things that pop up. And the other day we were talking, and I said, hey, we could use this ASRS database to look at very specific things. And I'm skipping part of the story. One of the great resources NASA has is a monthly publication called Callback. What they do is they feature a handful of events from their database to illustrate an issue, a hazard, how people performed or behaved during that hazard. And they have several different styles of callback issues. And I thought, you know, some of what they do in callback might be absolutely perfect for what we talk about. And you mentioned, hey, can you do some stuff dealing with aviation maintenance? And you specifically said, well, let me let you say it. Specifically, what kind of events were you looking for You know just just where a mechanic maybe made a mistake in installation. And the part of the reason why I was looking at installations was was that there was a report out several years ago from General Electric, uh, the engine manufacturer, part of General Electric, that uh said that fifty percent of the turnbacks, that turn back is a term that refers to an airplane that returns to the airport or to the gate. It may not take off or it may take off and uh return to the airport. But in any event, GE at that time claimed that fifty percent of those events that happened were caused by the wrong installation but on the part of a mechanic. Now the wrong installation is a pretty wide net that's cast. 'Cause the installation could be the wrong part installed. People don't realize it that an aviation you may have a component on an engine and the part number may be two, three, four, five, six. But then as the airplane is in service , th that part gets changed a little bit, not physically. It still has the same mounting bolt holes, the same method of putting on the engine, but internally something would change. And it would end up with that two, three, four, five, six part number with a dash one, a dash two. I and sometimes you see But it sometimes can degrade performance if you put the wrong one in the wrong dash number. So I'm assuming the uh most of those callbacks with the because of that reason Discussed this before I thought, okay, I know kind of how the ASRS database is structured. And the basic structure is this. They have a whole bunch of coded variables for each event. Was it maintenance, was it flying, was it not flying? And they have two written parts. One a general narrative, that is whatever the person reported, how they wrote it, and second was a synopsis of the whole report. And what you were talking about, I thought okay, that's a whole bunch of different kinds of of maintenance installation errors. Let me just focus on one very, very narrow error. Something got put in backwards. And why did I focus on something like that? Because it's real easy to describe it, and I can use just a two or three search words for looking for anything that has the coded variable having to do with maintenance or several of those, or the word backwards somewhere in there. By the way, backward singular and backwards plural are two different words as far as searching is concerned, then you have two options. You can either do one word or the other word, or put a wild card character and do both words. Doing both words got me way too many events. I thought let me make it smaller. So I just did back words. And I got, I don't know, couple of hundred of these. And I went through them one after another. I thought, okay, this looks good, this looks this looks good. I had forty nine of them. Now, I just glanced at the narratives and glanced at the variables. I thought AI could help me out here. I've already used my intelligence and experience, well, not intelligence, experience with the database, understand how the database works. And that gave me a line on how to search for stuff. Then I hand over part of my job to AI, I said, look through these forty nine, look through the synopsis and the narrative, and tell me if all of these fit the general situation of something got installed backwards. It went through the forty nine and actually came back saying, okay, forty-four of these look fine, but five of these are worded kind of like it was backwards, but it really wasn't. It gave me a short explanation of why those five weren't good. I thought, oh, that makes sense. I was like rushing through trying to get a whole bunch to work with. Threw them out. I got forty-four. Now here's where the hard part comes. Before I did this for maintenance, I did this for a completely different subject, and this is gonna be subject of a hook another show entirely. As you know, we've talked on this podcast several times about unidentified anomalous phenomena. And some time ago a colleague of mine, we combined forces to come up with eighteen different events in the ASRS datab ase that were UAP related. I did a whole different thing with trying to tell my AI I use Claude, not exclusively, but I use Claude for this particular project. Take these eighteen, go through the entire ASRS database and see if anything else has a narrative or synopsis that's kind of written like these 18. It took myself and my colleagues weeks to come up with eighteen. It took Claude and me going back and forth about an afternoon to come up with fourteen more. Now here's where I have to step off and talk about the back and forth thing. We were talking before the show. I was saying to John, it's like, you know, I treat my AI like somebody I'm working with. I'm typing up messages, I'm saying, hey, why don't we do this? Why don't we do that? They send me back some stuff. But with a colleague it might be hours or weeks or days between you know my message going out and response coming back. With AI, it's seconds. But AI doesn't have my experience. AI hasn't made the mistakes I've made do zens of times. And I told John, it's kind of like having an intern. Very hard charging, intelligent, they're raring to go, but they don't have your insight. How do you have a process to bring that intern along to make them, you know, more expert in what they're doing? And I asked John, what about your most famous intern, Greg, who you were telling me how old was he when he first uh was an intern with you at the NTSB ? We think it was twenty one. He was in senior year of college. And right now Greg, like all of us on the show, are in a different kind of senior category. So we're talking a long time ago. But what was the process in general? Don't have to get specific . What was the best process you had for bringing Greg along, the getting him up to speed on how things really worked and how you should look at certain situations . Well, basically he had mentors at DNTSB which were the senior investigators uh who brought Greg along with uh how do we say it I'm trying to think the the right way to say it but they they were tough on him . Making him uh they will bring s Craig would bring something to the investigators that he thought was pertinent and they would push on him pretty hard to prove what he was saying. And it may not they may have known that it was not accurate and pushed on him to go dig himself to find out the most accurate information . And uh they were not easy on them . So from that point of view it was one on one with the knowledgeable expert, with a younger wannabe investigator, what they pushed and leaned on Enjoying your membership and the benefits that come with that, including getting early access to new episodes and other goodies to be determined later. And if you're interested in becoming a member, you'll see a graphic on the screen with a arrow point ing to the join button on the YouTube channel. Join, do what is necessary to become a member at whatever level you'd like, and enjoy those benefits. And if you have questions, whether or not you're a member or not, write to us at flight safety detectives at gmail dot com. And we're not leaving out sponsors. We love sponsors. We can always use more and we hope that you can use our content to attract our audience to your product or service. Contact us at FSDsponsors at gmail.com we'll have a conversation . This is something that you know is a completely applicable to working with any kind of AI. The AI has its own way of doing things. One of those ways of doing things is you ask it something, it'll give it an answer. It may give an accurate answer based on what you tell it, but it may not be the accurate answer you're expecting. So on the one hand, you might have to give the AI some instructions, hey, when you do this, I don't want you to do this, I want you to tweak it this way and then give me a result that I can look at. But there's another part of this. If you give the AI a command and it comes back with something that doesn't make sense, maybe the issue is you. You know so much about a subject that you make some assumptions that everybody should know this, but no, they may not. And here's where I bring in my favorite tool at the beginning of any process I have with AI . Blank piece of paper . I sit down, I sort of sketch out not the whole story, but what is it I want to do? What am I looking for? What are my inputs? What are my outputs? What is the problem I'm trying to solve? What will make me happy with a final result ? Now I take my scribbles on a piece of paper and I might type them up. Why do I type them up? Because AI is typically use the following format. You type in something for it to do something, something you call in general a prompt. So your prompt could be a one-liner thing like give me a chart showing the population changes in the United States over the last fifty years. That's fairly straightfor ward . What if it's more complicated than that? What if you wanted to do two or three different things in a particular way? That's when you have to sit down with yourself and say, what is it I want? What exactly is it that I want? And how do I want to see the result and in what way ? Now let's get back to those forty four um events maintenance events. Like I said before. Earlier I did something similar with UAP . We had 32 events and I said I looked at a whole bunch of well, I didn't say this to it. First I said, I would like to have something that looks like a callback document. Only instead of talking about whatever issue the NC at the NASA wanted to talk about, it's the issues I'm giving it right here. So the first thing I did is I went into NASA's site for the Aviation Sage Reporting System, went through a whole collection of their callbacks, and by the way, all those callbacks are online, they got them going at back years and years. In fact, I went through five or six years worth of them and found certain ways of writing a callback Like one of them was they took an issue and talked about how that issue manifested itself in different phases of flight. I thought perfect for UAP sort of sort of thing, because sometimes they see them take off like They had another style that I liked if you're of a certain age, who used to be a radio commentator named Paul Harvey. And his style of storytelling was he would begin the story talking about here's a man and his family who went through all these certain things and this massive problems happening in their life. And then they would go to commercial when they would come back, and he would tell the second half of the story how, the man and his family overcame everything they lived happily happily ever after. So I thought this'll be a great thing for various kinds of hazard. A Paul Harvey kind of setup. You start with the setup , then they come back with how did the person resolve it? Now, what I just said, anyone of you out there could probably sit down with a bunch of these , write stuff out, and over time come up with a way of recrafting a callback issue in that style. It might take you days or weeks. It took me a couple of hours. I fed my AI my very detailed description of what I wanted and how I wanted it. I fed the AI several examples of the different styles of callback, and this is a lot simpler than what I actually did. I basically said, look, I gave you a bunch of examples, I gave you a bunch of raw data, give me some reports that fit best with some of these styles. Came back and gave me several styles. The knowledge I gained from doing that, I said, hey, I can generalize this to any issue under the sun, using the same database, quarter million ASRS reports. So let's go back to what we're doing with maintenance. I got forty -four things being put in backwards, did the same process I did for UAP, and I came up with or rather with AI's help, came up with four different kinds of callback publications, which you've had a look see at . So what's your first impression having read callbacks, regular callbacks from NASA and reading these callbacks? Yeah they they are they're taking the same information but they are digesting it and presenting it very differently. And actually I like what they've done with that because it does get you engaged the way they write it and delivers the findings in a very straightforward way. So I like it. Now by they you mean the original people at NASA who put this together over the years. Yes. I basically locked stock and barrel said, I want to take this style, I want to take this way of doing it and apply it not to whatever they come out with every month, which is great stuff, but something that is of specific interest right now to us Right. And AI and maintenance is coming on strong right now . And uh it's a work in progress. I think that there's going to be a lot of improvements because of it, but not tomorrow. But in the in the maybe six months down the road we'll start to see some serious improvements because of what they the re-looking at all the past uh inform ation and scoop ing up all the current information that they can and seeing where the trends are, what was what was happening that we didn't see with our previous views. And it's coming. You know, and it I I T school as you know and and most of our audience should know, and the students are using AI more and more. But what we're seeing with them, again here comes the inexperience , uhh they're taking that AI question that they ask and then putting it in their assignments. And it's usually written assignments, not uh verbal . And it's wrong. Because the question wasn't asked the proper way or or uh the information wasn't there and AI made it up to give an answer. I don't know all it, but it's not accurate. So the students are uh uh tr maybe they're trying to take the easy way out and use an AI to give them some answers to put on paper. But those uh professors take the time to to go through that answers carefully are uncovering the f the problems with AI if you take that first or second brush at it and use it to answer a question that you don't have a good solid base of information I'm not a student in the same position as your students are. So I have to have my own process of okay. Does this make sense? Have I done this right? Do I read through this again to make sure that this makes sense? Do I go to the original source material that these callback styled things are made from? And uh by the way, how is this used in in maintenance? We happen to have here on the show a world renowned expert in safety management systems. How many books have you had contributed to about safety management systems? Uh four . So a safety management system and giving a thumbnail description . If I'm wrong, please tell me. This is an internal system that an organization would have to try and uncover problems by having people submit problems they come across in their in their work. Is that too simple of an ex explanation? Over the task was to raise issues that they see, problems they had maybe accomplishing something. I problems they see with materi al. You know, here we go with the the dash number on components. You know, maybe w they find out that that uh in the uh inventory for a particular airline, somehow uh components got into the system with the wrong dash number . And uh you know that stuff happens when you have large purchasing departments trying to find bits and pieces for airplanes from all kinds of sources. It's not easy sometimes getting what you need to keep an airplane running, so to speak. And uh you can get the wrong piece in there. And it's that mechanic, the installer has the obligation to make sure that the piece he puts on the airplane is the right piece. And so sometimes wrong parts do get on the airplane. So that all of those AI can help identify all of those many of those problems uh get lost in the system . And AI can ferret those out by identifying the those gaps that have occurred in all in all the the bits and pieces of nuts and bolts of an airplane when you're putting it together are in the procedures. That's a that's probably even a bigger problem than just the paths, is the procedures that the ma in my case, the mechanic it's supposed to be follow . Sometimes those procedures are wrong. Or they're they're they need to be tweaked a little bit because they're not quite accurate in certain certain places. Sometimes it's it's it's uh interesting when that when we say that let me give you an example. Uh on uh on an airbus airplane, there was a procedure and this pr and this this procedure uh was used in different areas and written differently in different areas of the air. So it's in the wheel well and it's to gain access to an area. And it says to remove uh component A uh first and then remove component B. The problem was that component B was the first component that you're looking at as you look into the wheel well . Component A was behind it. If you were to remove component A without removing component B, you have got a ja a real tiger by the tail because you can't reach it, you can't get it in and out easily. Right? So the sequence A and B in this case was wrong. The procedure should have said that you remove B first and then you get access to remove A . Elsewhere in the manual it had that step there pro perly. But in this particular one, the engineer who wrote those procedures wrote them wrong. That A. Now on the installation side it had it the correct way In the same document, the installation clearly said you put A in first and then put B in afterwards. But on the removal it says you remove A first and then remove B . So and the and the FAA who was very uh tough on documentation, they violated uh two coupler mechanics for failure to follow procedures. And they were correct, they failed to follow procedures because they removed B first. They looked at the job, realized that it would have been easier to remove component B and then re get to A and remove A. So imagine you're running an SMS program inside of an organization and you understand exactly where the problems are because you're the expert. You've been in that company for many however many years. You've seen the places that have cost your company money or cost the company by having the FAA come in and hassle you about it . How's what we've been talking about before, this whole AI thing going to help you out? Well your SMS database is kind of structured like the uh ASRS database. You have a bunch of things in some sort of organized fashion. Probably have some coded variables, probably has some narratives. Go through there and pick out ten, fifteen of them. And if you had this same kind of style program I'm talking about, which you could probably build on your own by using things like Claude and other AIs. You can create your own storytelling module, you can call it whatever you want, that takes let's say five, six, fifteen, twenty different examples. You say, hey , turn these into a learning tool that is of this particular structure. Maybe you give it the callback structure, maybe you have another structure, and you hand this out. Now you might have all the other training and professional education going on. But this is another way of informing a population about an issue. Telling stories. Maybe we tell stories, you know, over a couple drinks, maybe we tell a story during the company picnic, or we can tell a story in a fairly straightforward written thing that says, hey, here has what happened to colleagues like yourself. What can you learn from this? What takeaways can you take from this? Hello aviation enthusiasts, safety professionals, and curious minds. Welcome to Flight Safety Detectives. We are thrilled that you've dropped by . Here, we dive deep into the world of aviation safety. From incident investigations and technical analysis to lessons learned and what they mean for safer skies . As a member of our channel, you get exclusive member only perks , first access to new episodes, live stream discussions, and QA sessions. We are a community of fellow safety detectives who love aviation as much as you do. If you're passionate about aviation or just curious about how things work, we'd love to have you on board. Hit join, ring the bell to be notified, and jump into the discussions in the comments. Slap in because together we're going to explore aviation safety like never before Now this is all kinda high level what we're talking about using this to create a in it NASA style document. By the way, the NASA document is called callback. The ones I put together, I gave it a different title. When I was thinking of what am I going to call this thing? You won't believe this. What popped in my head was L O Cool Jay . He has a song called Mama said Knock You Out, where one of the first lines is Don't call it a comeb ack. I thought perfect . Don't call it a callback. That's the name of those documents. We have them on the page supporting the site. You can download the whole bunch and look at them yourself. Call 'em whatever you will. That's just my , you know, semi-imaginative way of of calling this a title that is not callback. In fact I say in the masthead, inspired by NASA, but this is NASA has nothing to do with this. This is all something I created radio and podcast for years. And very often we have to say something in a short period of time. So me personally I put together my own sound bites. That is, if I'm talking about a subject, I say to myself, what are the four or five things about this subject I want to get across to the audience? Now actually sit down and start writing out how is my soundbite gonna look and sound? I'd practice it and make sure I can say it normal tone of voice and I'm not gonna trip over any words. And I have a like some basic rules. No more than a couple of pieces of information, no more than eight seconds. So I actually put together a list of requirements into Claude and said make a soundbite process so that I can like feed in you know a paragraph or two or even a whole page of stuff. Hit a button and out comes a bunch of sound bites. Now I still have to correct them and make them my own, but if I'm running short on time , I'll use my regular sound bite process only a little bit and I might rely on this a little bit more. If I have a lot of time, I'll do both. My regular sound bite process, back it up with a little AI , mix the two together, and have the best output. So this will be downloadable again from the page that supports this website. You'll have a HTML file you can put on your own computer . You feed in whatever you want, hit a button, and it comes up with a prompt that is a big chunk of text that you paste inside a clog. Hit the button, boom, out it comes. He said to yourself, doesn't clog clause cost money? Yes and no. They have a free version? They have a version that costs money. And the instructions I have with this, I had it so that you had three options. Do the totally free version, do you know the paid version and do a version that's sort of like in between those. So no matter what your situation is with Claude, even if you never used it before, never trusted, you got everything you need to do with the free version. Now if you do the more complicated version, you'll probably have a different result because it includes a lot of the insights that I made when doing this thing, and that'll give you the full flavor of it. That is the full flavor of Claude. It'll give you slightly more complicated, sophisticated results. But enough talk. You'll have these for download, both the examples for maintenance, you'll have these for download the stuff I talked about making your own soundbites. And I know it's a lot in one show. And this is not going to be the last AI show. Why? John, in your opinion, is AI going away anytime soon? No, it's coming on strong. There I think there's definitely a role for AI in it in a number of different locations in society today, because it's an information system. It's an aggregator. And it's uh as as it matures it's gonna have uh a much improved view of all that data and the ability to extract what's meaningful from that data. So it's here to stay. It's gonna grow exponentially in the very near future. Let me ask you a question. I think you know the answer to this. I'm gonna give you three major professions within the aviation world. You tell me which is the most likely, which is the least likely to be taken over completely, not partially, completely by AI. Pilots, flight attendants, mechanics. Which one of those w can be possibly taken over completely by AI? In some maintenance probably f first , uh piloting for for um improving the training of pilots and uh and f and bringing up the rear would be flight attendants and completing their job. The the variable that puts the flight attendants back is much of their problem s come from individual passengers. So that's much harder for the for the for a tool to to uh extract the feelings or actions of But all three are going to have feel the impacts from AI . But since maintenance is a one hundred percent process, at least But when it comes to actually doing the job, that is turning the ranch or dealing with the fact that uh people around you might not be following a logical process and you have to figure out what's going on. There's still going to be a a role for the experience mechan ics to deal with those edge cases, in my opinion. So you know, in in today's environment, uh and pilots know this, that you you usually can tell uh the mechanic that that really knows whatever the subject is. I'm talking about when something's going wrong, not every day. And and uh I I'm just trying to think of a good example. I have one. I so we had an airplane uh at the gate that had a s a certain unit that wasn't working. And and a mechanic that was assigned to that gate was out and looking at it and it wasn't going well . And uh it was me in the inside. So I'm the supervisor, the lead, or crew chief, whatever the company calls them . And uh I see it not going well. I'm not seeing the problem, but the airplane's not going away . Uh so I go out and and we deal with the problem. And the captain says to me , right at the very end, he said, You know, I told the first officer uh when this wasn't going very well in the beginning, he said, just wait a couple of minutes and you'll see uh somebody can walk ing out from from the building someplace and he'll probably have a lot of gray hair and he'll come out here and and uh the problem will be dealt with. And I laughed because that's exactly what happened . Right there was the lack of of uh leadership. It wasn't it wasn't that the mechanic was unskilled, but he was afraid to take action on his own. Maybe that's maybe that's too general , but he was slow to take action on his own. And and uh when I come out I just said do this, do that, and we got rid of the problem and uh away they went. So AI can take care of a lot of that because it was sort of a routine process . It was based on my experience of seeing that problem many times before and remembering the course of action that we ta had taken to fix that problem , and and eliminating some of the steps that the m young mechanic was already taken. So he was taking what he thought was the obvious course of action. I was taking the course of action that was based upon years of of making the wrong the wrong selections for fixing the problem until get a handle on the right one. So that's a perfect scenario for AI to extract that information out of there. So in the very beginning, if we had a a robust AI tool in place at that time, we would ask them what's the most obvious cause of this system failing to work? Threw a switch and it didn't do anything. Alright, so the most obvious failure was really fifteen, I remember it well. And uh the the uh AI could have pointed him to that immediately because that was the common fix for that kind of problem. It's not absolute because it may not be that really. It's too fifteen and not two fourteen . It may not always fix that problem, but it's the one that would that 99% of the time does. So AI can bring you there almost instantly. Where if you're a young and inexperienced person for that particular system on the airplane, you would go through a bunch of other steps before you ended up at that relay. So time is money. And most of these cases you're working on an airplane that's on the gate that's full of passengers because the problem didn't show up until departure time . And now you've got people occupying an airplane on a gate that's not gonna go anywhere until the problem is solved So AI can save a whole bunch of money in that area by helping steer things to the to the uh cause that has previously been identified as the the the most likely. Well this is a perfect segu e in the second last word about this show about AI. AI is a tool. It's a very powerful tool. It's not going away. But AI by itself will not get the job done. That takes experience, insights, and wisdom, and that will only come from practice . And now for the last word Do it at home before you go to the airport other hotel room. Do it at the airport again, make sure they agree. Don't forget to check your weather here, there, and everywhere in between . I good pre flight. I just had a discussion last night with a with a uh a a corporate di uh director of oper ations at a s at a function I attended last n ight. And uh and he had that he agreed that too many of the too many of the pilots, including his pilots, don't do an adequate walk around on the airplane. And that he's been working towards uh increasing their awareness as they do walk-arounds. So it's not uh not just me that's saying that and seeing it. It's uh other people are noticing it now too. A lot of the You know, it's often said that the pilots are the last line of defense on the airplane before it takes off . And that's true. That there's a reason why they do walk around. There's a reason why w they're required . And if once you get in the air, please put that head on a swivel , especially on the general aviation side or general aviation airports. Uh they may have a lot of other activity in them besides general aviation. And there's an awful lot of students out there and s the whole purpose of a student license and and that