AI Fatigue: The Pros, Cons, and How We Prepare for More - Pat Wright

Great to be here. I'm happy to chat all about AI, the good, the bad, and the future is what I'm calling this presentation. Um, as I go through this presentation, please do me the favor of just dropping anything into the chat. There's a chat right there. I should be able to see it. If you have questions throughout, we do have some Q&A time at the end too as well. But if you want to ask in the middle of any of these things, if you want to dispute anything that I'm saying, please feel free. I like this to be as much of a discussion as possible. As AI is impacting all of us, it's important that we have a discussion about this as we move forward. So to get started, my name is Pat Wright. I am a PostgresQL advocate for Redgate software. Redgate software makes great monitoring and deployment processes for everything that you need for the DBAs in this modern world. And we have both Postgress uh Oracle and SQL Server um products that you can use and do those things. I've been doing database work now for about 25 years. Um, you can reach out to me at any of these different places right now to go and get these things. This presentation will be uploaded directly to my GitHub link that you see there. All of my presentations are up on my GitHub link. So, if you need anything there, please go and find it there. And also, you can reach out to me at any of these places. I'm more than happy to chat about this presentation or any of the others. So to get ready for this presentation or the way I I'm going to do this presentation is I asked AI what has it done good bad and what does it summarize that. So, we're going to go through what it came back to me and said, "These are the things that we thought were good and bad." And we're going to review the slides that it gave us, and then I'm going to put some of my own stories in there about how I've currently been working with the good and the bad. And then we're going to talk about the future and what you can do to really get better with AI projects and that sort of AI fatigue that we're all facing right now. So, I don't know about you, but I always take my prompts that I put into chat GBT or Claude or whatever AI I'm using and I kind of save them off. So this was my prompt about the good. This is what I asked it to do. I told it this is what I need out of the presentation. So this is what came out of it is what we're going to talk about here in a in a second. And I want to call out here that I did say promote yourself and talk about AI in a good light. So I did ask it to literally promote itself a little bit. So these are the slides that it came up with related to that. So it first brought up this idea that enterprise text to SQL has gotten very good. So, it's it's gotten a lot better at writing SQL statements and things like that. Where it kind of falls down, it mentions here is that where it doesn't do such a good job is the whole trans I mean the whole transformation of the code base and everything it needs to do. So, it's pretty good at making SQL statements for you when you ask it, hey, write me a SQL statement to go and pull this data and do this report or these sort of things. It's got that down, but it can't do everything you need it to do. And so that's what this slide talks about. So this was one of the things that it said, we shipped this. We can do this. AI can do this for you. Now, it's a good thing. I I think this is positive and it's something that I use pretty frequently if I wanted to write a SQL query query for me or something else. So this one is more about AI is saying that they found a way to actually impact the database and change things about Google's spanner database. This is specifically to Google Spanner database to make it faster, more efficient, more improvements on it without necessarily changing too many fundamental things. So it kind of said we made it more efficient, we made it better. Um it I like that it pulled out in here this little statement down here. Why this one matters to you. It's the only result in the deck where AI improved the layer underneath your database rather than queries on top of it. So it changed the fundamental nature of the database that it did this. Now it did this for Google Spanner. This doesn't help us if we're in Postgress. We're talking about Postgress right now. But it's another thing that AI has done in the recent time that has been a positive thing that we want to talk about. Okay, this one's really interesting to me because um this one impacts me a lot because I use Chrome a lot. So, but this one's talking about AI has gotten really really good at finding vulnerabilities in our systems. So, you can run different um analysis against it. You can run analysis against your code and it's finding a ton of different lines of codes that are vulnerable out there and that need to be patched. And it's even said that, hey, we also found 18 real, previously unknown vulnerabilities. Every team found at least one that was not the assignment. So, it's found this stuff even without being prompted to do so. So, this is a good thing that it's finding these things. And this is where I bring up the Chrome part of it for me is that I really like that it fixed Chrome things cuz I use Chrome all the time and I want to make sure that, you know, hopefully it's it's improved my Chrome experience. Um, I will say I hardly ever update my Chrome though, so that's another big problem that I need to fix in the future. But it's really good that it's showing that it can be very helpful when it comes to vulnerability, discovering these things. I hope other people have been doing this in their own um applications and their own systems to make it better for them. So, let's talk about some celebrations of real things that I've done. Now, those were the good that it came up with. Those were the key things. Now, let's talk about things that I celebrate. And obviously, um, if you're a Cubs fan, Cubs won recently. Um, I'm celebrating that Cubs fan wedding. So, when I started earlier in this year, I created an entire scanning app. Now, I am not an engineer. I've not really spent a ton of my time coding things in the past. I've done a lot of scripting. I've been a DBA, so I've done a put together scripting things. I've put together Terraform stuff. I've done a lot of different things like that, but I don't sit down as an engineer and build applications. So, I built an entire scanning app which was for the PG data conference in Chicago. Every sponsor had to be able to scan a badge and say, "This is stored back down into the system." And then they have to be able to retrieve that data after they've done all their scans. This is a common thing at all events. We didn't have one for the PG data event. So, I sat down with Vibe essentially with Claude and started writing that scanning app and built what needed to be done. Now, I've used one for a different event that I do put on locally here, but we wrote it many years ago and we haven't updated it. So, this gave us a new powerful app that we could use. We brought it all into AWS 2 and it worked really well. So, these sort of things are great things that we can do all the time. Another big one was the uh previous DBA team I just worked with. They created their own agent so they could answer a bunch of questions related to their local system. They could keep it in their system. they could do things very quickly with it and makes it a lot more efficient. And again, this is a DBA team that doesn't necessarily have years and years of coding experience, but they've been able to put this together because they have AI on their side. Um, that same company had in 2025, they released more features and more bugs fixed on their last release than ever before in the history of the of the company. And keep in mind that they started their path way back in 2023 and 2024. They started very early on the AI path. So they had a whole bunch of time to go through the pain and suffering that we see from a lot of companies now that are trying to adopt AI and trying to get it in there. It takes it's kind of hard sometimes. They went through all of that in 2023 and 2024. So 2025 became a very efficient way for them to get all of this out there. And so it was a huge success story that they created all of this work that got that got done so much better because of that 2025 in um putting AI into the system. Um I will say there's some ROI to that and everything. We're going to talk about that in a second. Don't worry, I'm getting there, but I will say that it was really a a good beneficial thing. And then the last one that I've done recently too is a load testing application. So um right now I primarily do presentations and demonstrations and things like that and one of the things that we were testing was our Redgate monitor product on different sort of workloads and we wanted to see those different workloads. Now I know those workloads very well because I've been a database administrator for a long time and so I wanted to recreate those workloads I saw in production and make them similar. So I sat down with Claude and I created basically good workloads that looked just like old production systems and we could then look at that and monitor and compare that and say what is working and what is not and how we do that. So that load testing application gave us the power and the flexibility to do that. So there's a lot of really really good cases with AI that says these are great things we can do and I'm sure that people on the call and people that are listening to this recording have a great examples of AI working very well for them and I hope that you can celebrate those as well too cuz now we're going to talk about some of the bad things that AI has done. Let's make sure we get into that. So let's talk about the bad. Here is what I kind of focused on for the bad. I um again this is the prompt that created the slides that you're about to see. I didn't modify any of the slides. I just um took the slides and I'm highlighting different things about them. This one also I kind of said focus on failures, crashes, and deletions by AI that have impacted organizations. So I wanted it to tell me about the bad things, the negative things it has done in the world in the last little while. So of course it came up with this one first. Now, I'm pretty sure most people on this recording or most people on this call have probably heard of the it deleted an entire production database. What I didn't know I I didn't know this is it only took 9 seconds to make this decision. And I'm like, that's that's really crazy to me. It's like we literally went from, oh, I didn't see this there. It didn't find a file. It didn't find a token as it mentions. And then it just deleted the issue. I mean, just deleted the backup. It's like there's no thought process, no no pause there to say, should I do this? And that's one of the key things. The other thing I find really funny about this is that it mentions down here another thing that it did is an agent wiping an entire drive past the recycle bin. Meaning it completely removed it and then completely removed it again so that it could not be brought back. This other one pardon mentioning there too hallucinating successful creation of a directory then overwriting a project directory. So it says, "Hey, I created the directory for you. Don't worry, it's there. There's no problems." And then it came along and said, "No, there's a problem. The project directory isn't there." [snorts] So those are those are really big things about that hallucination. Everything else, the other one I'll bring up on this is this statement at the bottom. The controls that failed were not the models. Token scope, backup, topology, environment separation. So this is like the junior DBA coming to you and saying, "Well, I didn't I I did delete the table, but you gave me permission as super user, so it's kind of your fault that you gave me permission, but I did it." Okay, so that's that's exactly what the AI is doing here. It's literally saying, "Well, you told me that I could do this, so I just did it. I didn't, you know, think about it or anything." So, we have to think about these things in terms of it can do this. It can cause these problems. We have to be very careful with it. And again, we're going to talk about these things in a second, but I wanted to call out that it wrote this itself. It literally wrote this itself saying, "You let me do this and that's why I did it." So, [gasps] yeah, I don't think it helps that company that lost their entire production database or anything. Okay. So, how about this one, too? Security is not solved. I I love this slide that it created because it's like, hey, we figured out how to write something. We we're really good at making code and we're good really good at writing things, but we're not really good at securing it, by the way. Just want you to know that that we're not really good with securing it. And so this should get all the security people on the call here recording. You should be very concerned. You should be making sure you put controls in place. And the one I call out is this one down here. Repositories with an assistant enabled um with an assistant enabled leak secrets 40% more often. Meaning there's a lot more secrets getting out there because of these agents. They don't care that your AWS encry cryptography key is out there somewhere else. They don't they don't think that's a big deal. They're like that's just another variable. I'm not worried about it. Let's just put this out there. So, it's really important that you worry about these things and you control these things. And this is definitely one of the bad things, and I think I've seen this a lot, is people talking about don't paste these things into AI. Don't do these things in there. And also be careful what your AI has access to because it will just push it out out to the world if you don't put some controls around it. So, this was a good one that I'm glad that it brought up and and again, it brought this up itself. Um, this one's really, really important because, and I love how it titled this. It's one of my favorite ones that it titled. We are shipping more and maintaining less. And what that means is that we're making lots and lots more code, but we're not using the old code that we still have out there. And this one kind of hits home and that it makes perfect sense that an AI would do this because I've known lots and lots of engineers over my times. And engineers typically don't like to go in and rebuild someone else's stuff. They like to write their own new thing. It's just that culture and that idea. It's like why would we try to, you know, make it work? Why don't we just rebuild something new and write our own thing? And that's exactly what AI is doing as well. It says right here, it's copied and pasted lines of code and it doesn't reference the old code. It doesn't it doesn't get into that at all. So it's really just making more code everywhere. And I saw this a lot in one of my previous companies that we went from um say like 10 million lines of code to 20 to 30 million lines of code in one year because AI was writing a lot of the code. So it really jumped high levels of steps. And that's a concern from your DevOps friends and people doing deployments and everything else. Not because you know there's a problem with it, but maybe because you have to bring that much more code out into the system. Sometimes app servers take longer to deploy to or web servers take longer to deploy to because you're pushing out more code. So keep these things in mind when you're doing this and maybe work with AI to say no, let's try to use what existing code was there instead of trying to find new stuff. Okay. Okay. And then this one is all talking about that it's not as reliable on the results with these estimates. These estimates were hilarious to me. And I bring this one up right here. Developers were 19% slower with AI and believe they were 20% faster. Now, the funny thing that I have here is once again knowing developers for a long, long time, we always say there's two really hard things. Naming things and estimating for developers. So they can't estimate very well. And so I'm like, okay, did this really just have the developers were estimating wrong or did they actually get 20% faster or what happened here? And again, this is all a slide created by AI. So in other words, it's not like it went out and surveyed a bunch of people and asked this. It's just using its own knowledge and data that it has to make these numbers up. So the thing that I pull out of this though is that is that we really do not know fully how much we're using AI and how impactful it is. We're still trying to figure that out in good measurements. Some of those things are working and some of those things are not. So we have to be very careful about what is it actually doing for us and what is it not doing for us. Again, measuring becomes a harder thing. And this one in this case it's saying we're not very good at measuring this and you need to get better at measuring it. I do call out this bottom one too. I still like this one. Debugging real SQL is still a human job. Meaning that database experts do it 78% of the time and the model only did it 38% of the time to figure out the issue. Now, that makes perfect sense to me because a really good DBA or any DBA that's been in their system for 6 to 7 months knows the data, knows how the system works, knows everything about it. A model is simply going to know, hey, you have a foreign key here. I should be worried about a foreign key or you have an index here and maybe I should look into this index or something. It doesn't know the knowledge behind it or what was created there. That's where your database experts come in and that's where it becomes more important for them to be there. So it's very important that you still have those database experts and helping you out with those sort of things. Okay. All right. So the costs that land outside the engineering or so. So this one's actually not about we're not talking dollar cost here. We're talking about how does it impact the other things that are going on. And this is really about skills gap. Um this is really about the problem is is that AI has taken away a large skills gap from everything else. meaning that we're not learning as much as we did before. And we're going to talk about this more in a minute. But the the key pieces to bring out here is that the more you trust the model, the less you check it. So once you get your agents out there working and they've been doing stuff for 6 months and they've been doing it well, you really don't ask them questions anymore. You don't say, "Hey, are you sure this is right? Are you sure you're making sure this is checked?" They don't do that anymore. They don't need to. So the problem is is then you have just this trust inside it and you don't know that you can always trust it. So the skills gap is a big concern moving forward and again we're going to talk about this more in a second. Um I will point out this one in the bottom though too. The junior role is where DBAs learn judgment. We are automating the training ground. So meaning we are not learning our judgment nearly as much. new DBAs coming in, new database professionals coming in are just asking AI to solve the problem and they are not solving the problem themselves. So, you're having a lot harder time with teaching them and bringing them up to up to um skills that they need. So, important thing to know. Okay. So, the real thing to uh real stories to not celebrate and so you're asking why do I have a cash out ticket from Vegas on this picture? Yeah, I I was actually in Vegas last weekend. I didn't do very well. I only made 40 cents out of it. So, that's why I'm saying don't celebrate these things. That's why that's there. Um, but let's talk about the skills gap problem. This is the big problem. And this picture up in the upper right was from a presentation at PJD Hydrobad, um, an AI keynote that was really well done. And they talked about what was scarce in the past. In the past, the scarcity was technical expertise, development capacity, data analysis, and access to people. We didn't have DBAs. We didn't have anybody out there. So we were lacking that. But now what we're missing because AI is there. Now we have those things through AI because it can do it very easily. But we're missing judgment and we're missing trust and governance and we're missing the ability to learn and measurable outcomes. We're losing that. That's the skills gap. People are not learning anymore with AI. They're just letting it do it and then they're not going to have the ability to problem solve in the future. So that's a big problem that we have moving forward. And that's one of the reasons I consider this uh don't celebrate this and this is a bad thing. This is one of the bad parts of AI. Other examples that we have is so deleting files. I was working on a project um with AI. It was scraping website information and it created a little simple CSV file. I asked it to update the CSV file. Lo and behold, it deleted the entire file. And then of course it told me it looks like your file has been deleted or corrupted in some way. And I'm like yeah, you just did that. So the funny part was is that I I went and I scraped it all down again, got all the data back. I made a backup of it into another directory as I do as a DBA. I just make a backup of it and then along it came and said and deleted again. But then it immediately asked, "Let me see if there's a backup out there." And it found my backup and copied it over. And I'm like, "That's good and helpful, but at the same time, you've now just, you know, found my backup and I'm now concerned." So keep those things in mind when you're when you're working on this. Um the higher costs I talked about earlier the company that did a lot of work to release everything release everything that um was so much beneficial for it that ended up costing about 5% higher in cost. And we're not talking about hundreds of K here. We're talking a lot more than that. The point is is that you have to understand that when you put that much into AI and everything else, then you definitely are going to have higher costs. And that may be perfectly valid for your company. That's a good question for your company to ask. We want to improve. We want to move forward faster. We're going to pay the cost. But it needs to be a decision your company makes. You need to think about this no matter what. Um, I see one question comes in. I'll get to that in one second. Um, wiring a house. It's a good friend of mine right now is rewiring his basement a little bit. He's doing work on it. He went to Chat GBT and asked kind of the proper wires for what he was doing. It was a sound setup. It was um HDMI sort of things, different sound things. And he pulled all these cables and got them in there and then he tested it and it didn't work. And >> [clears throat] >> um Tat GPT later came back and said, "No, no, that won't work with the situation. You need these cables." And so he had to redo the entire thing. So again, these are the bad things that come up to to make problems. You need to work on that. And so we're going to talk about that in just a second. Um the question that's come in, do you find that asking a follow-up question of the AI like are you sure still creates significant improvements in the quality of the response from the models? Um Dave, excellent question is that yes, sometimes if you ask are you sure or sometimes if you say check your work again, another way to do it is say test and validate what you just created. I use a Claude MD um model and we'll talk about it in a minute, but I use one that actually in there says you have to write a test that looks like your value. You have to make sure that it comes back the way it looks like. So, not just the are you sure, go further than that and say test what you've actually done and show me that it produces the result that you expect. Okay, so that that hopefully answers the question. If not, feel free to drop some more in the chat. So, finally, I asked it to summarize. This was the the the prompt for summarization. Just tell me what's important and come in there. And this is what it gave for the job thing. Now, now the the funniest part about this is that verification is the job now. So, it's telling us we need to start verifying what AI does. That's good. That's it's true. We need to make sure that. But the funny thing to me is that it's like, okay, the leading model can only do 38% alone. Database experts can do 78. But expert working with 80 with act with with AI can do 87. How does that math work? Like like 78 to 87. I'm like I don't know why it thinks that it only gained certain amount. Like I'm like why did you come up with this number? I'm like okay. The point is is that what it's trying to say is that we want to work together to make it a better system which I somewhat agree. I agree that we should be doing things with it but I don't think we should be going too far with it. [snorts] And this last part is this last sentence down here at the bottom. That was the funniest sentence to build that habit in and this is the most useful colleague you have ever had. It's saying AI is the most useful colleague you've ever had. Now I have had some amazing amazing expert DBAs that I have worked with and I guarantee you AI is not the best [clears throat] colleague I have ever had. It is a very good colleague. It does a lot of great things for me. the scanning app, all the things I talked about earlier, the the doing load testing, that's all great, but would I tell it to be the best database colleague I've ever had? Absolutely not. I would call one of my friends as a best colleague and find out from them before I really went into AI to do that. So, but this is how it sees it. This is how it sees the summary, that sort of thing. So, that's its summary. We've gone through that. Now, let's talk about some of the things that I want to add on to this and where we're going in the future, where we're looking at, and some of the things that you can do to improve as you move forward with these things. So, moving forward, one of the biggest problems we have right now is AI slop. And this is a term in LinkedIn and and I'm talking mostly on LinkedIn and mostly social media and things like this. This is a term out there that just says you have a lot of different things out there built by AI. And the big problem with things like this is that I want to make it clear. I love infographics. I love visualizations like this that really translate something for me. And years ago, it used to take people a lot of time to put these things together and figure that out and make it really informative. But now you can just ask AI to do it and just spits these things out. But it's not always accurate. It's not always right. And years ago, somebody in LinkedIn, somebody out there published and said, "If you post to LinkedIn every day or every week or something, you're more likely to be seen." So now you can have AI bots just out there generating these things, just pushing them out there and doing whatever, and it doesn't really matter what the content is. You're just trying to better the algorithm so you get seen. And that's what we have a problem with AI slop. That's what's really the issue out there in the LinkedIn world is that people are just publishing and publishing and publishing and they're not really valuable for what they publish. So, it's important that you know that this is out there. It's happening. The best solutions, we're going to talk about that in just a second. But just understand that a lot of these things are out there and and keep in mind I actually this picture I found this and I like this. If you're using this to explain something and you're saying, "Hey, this is how we're building a car and this is the ideas from it." That's great. That's wonderful. Please use it. But don't just put it out there with whatever was generated from the AI and just literally turn it into a LinkedIn post. That's not going to help anybody. Write your post, talk about it, and then use the image for that. Helpful. Earlier this year, I actually did another AI presentation and I had it generate all the images for me. And I made it clear that it generated all the images, but it was me talking the entire time. And the images were fun. They were a lot of fun, but we did it specifically for that reason. So try to be clear on when you're using this and when you're not. Okay. So let's just talk about the future. We've got a few minutes left. Again, we're going to have some time for questions and everything else too, but please feel free to drop those questions in there if you've got any. Drop them in at any time. So the future and moving forward. I've broken this into two slides. One is about software and agents and then one is about the human fatigue side of things. So we're going to talk about both of these. Um really when you talk about software and agents, we cannot say the first word enough, control. you have to start setting up your cloud environments, your co-pilot environments, everything else with controls. Um, there's a bunch of different blog posts out there. I've done a demo on it before. Um, I I believe on my GitHub profile, there is a cloud.md file in there if you want to see how I set up my controls. You can go out there and you can take a look at it. But you should have controls on your AI every single time you talk to it. And that control should say something like, "Ask me before you delete a file. ask me before you go to do this drop database statement or delete table statement or something like this. You are not allowed to do this or say in this environment if I tell you we are in a development environment you can do these things. If I tell you we are in a production environment you can't do these things. You can set that inside there and you can control that. But you need controls in place. Um you should be doing nothing with your AI if you don't have controls in place. Um, testing. I I've said that um earlier. I love to tell my AI to test itself. I love to say write a test that will make sure that what you're doing is going to give you the exact same results. I do it all the time. I just tell it, okay, this looks good. Now go and test and make sure that this says the exact number that I'm looking for. Um, instructions and teach. You should do this more often. ask AI to guide you through the steps that it did and then read through those steps or possibly even do them yourself. Um, I was doing this recently with Docker is that it was building Dockers for me and I had a hard time rerunning the Dockers and stuff. Instead of asking it to do it, I had it tell me the steps and then I created my own scripts to do that so that I could get more practice creating those Docker scripts because you never know when you can't talk to your AI. As a matter of fact, I know all the time I run out of online access. Um, I travel for a living and so sometimes I don't have Wi-Fi access and I can't talk to AI. I can build a local one, but I still don't always run a local one as well. So, it's important to know that. And then problem solving. Have it create problems for you and then use that time to actually solve those problems and help it figure it out. Don't say solve this for me. Say give me a problem, create this problem and then actually work through the solution. That's an important to cover that skills gap. Um, okay. And then the fatigue and human side of it. I can't say this enough. Connection, connection, connection. We We need to still be able to call someone else and say, "How would you do this database backup?" Or, "How would you do this replication scenario?" Or sit down with a person at an event and say, "I'm trying to set up replication to do these things. What is the best way for me to set it up?" And talk to an expert or someone else. Yes, AI will give you the steps. They'll give you the tasks, but they won't tell you, "Oh, by the way, when you set up replication, you may have to worry about the 30 gig wall file it's going to create because it's going to get backed up and it's going to stop running." These things are what humans have figured out, what the DBA experts have figured out and things like that. So, make those connections. Um, make the engagement. When I say engagement, I mean on AI, I mean on LinkedIn posts, stop just liking things, especially when they're AI or something else. comment, engage people, talk to them about it, tell them, "Hey, I don't feel like this is a good idea or hey, this is a great thing that you've created and celebrate that." Um, don't just hit the little buttons because that's not nearly as much of engagement. Talk to people and engage with them. And I know as an introvert, we are all introverts, this is very hard to do. [snorts] I don't disagree, but I'm going to tell you one reason why right here at the bottom that it's critical that you do this in the future. Attend events, mentor other people. These are all important, but the most important one is referral. Referral is basically right now, I've heard from lots and lots of people in the industry, if they get laid off or something else, they have a hard time getting another job because getting the AI to see your resume is very difficult right now. All the AIs are just saying, "No, this person doesn't fit. This person doesn't fit." You can always get around an AI if you have a referral. If you have somebody you know in the company, if I go to a company that I know and I know that John has worked there or somebody else or Joe has worked there and I can talk to them and I can say I'm want to come and get hired here, they can give a referral because they worked with me in the past or they've done an event where I've been at or they've seen me speak or present. That referral is your golden ticket to any future job or any future thing because you can then make that connection and you can get around AI and everything else. I've hired literally hundreds of people in my many years of experience. And if you give me a referral, I'm going to go and hire that person so much faster because that's a person telling me, "Hey, this person is good to work with. They're wonderful. I've done this before. You need to hire them." And just to show these pictures up here, this was I was up in India um a lot. I got to spend um time with wonderful human beings that I spent the whole weekend with. And that's how I make connections and engagement. this is just a great way to do this. Hari does an amazing job at the HydroAB PG days and it was so fun to connect with him and I love to see him at all the other events we go to and so it's just a way that you can really connect and meet people by doing those sort of things. So I can't stress this enough. If you want to deal with the AI fatigue, if you want to step away from the AI a little bit, let it keep doing tools for you. Let it make applications and software and MCP servers do all of these things. But don't forget that the wisdom that you need to put into these things are from humans, not from AI. All the wisdom right now comes from humans, not from AI. So you need to be out there speaking to them, talking to them, and finding events to attend. Um, and also I I do put on a lot of events. I help with a lot of events. We need to see more people out at our events. You need to make these connections out there. So please come out and visit us at events. Um, okay. That's really all I had for today. Um, I'm going to go ahead and leave my slide up. Again, we've got some questions. Uh, if you want any ask any questions, please feel free to drop them in there. But also, if you want to head out, that's fine, too, as well. This recording will be available very soon, so we'll have it all out there. But let me go ahead and answer this other question that came up. Uh, given the evolution of new technologies from experimental to useful to scalable in the database, specifically, where would you rank AI currently, and why? Ooh, that's a good one. Um, oh man, I I'm sorry. I've got to give you the database answer of it depends because it depends on the company and what they're doing with it. Um I will say that my last company that I worked with with using AI, they were fully in the useful space and they were in the scalable space even really they had they had spent two to three years in that space getting it there. They left experimental and they went into that space. the current company I'm at is closer into the experimental and getting into the useful and going from there and moving forward. So in the database you know it's too hard to say for one thing if we're just talking about Postgress it's useful and scalable 100%. It's not experimental. You have PG Vector. You have a lot of great things out there. It's all available to you right out there. If you're talking about SQL Server, I would say the same thing. Useful and scalable. It's built into it. It's got stuff that you need. It's got all the pieces. Um MCP servers, they're a lot more, they have a lot more experimental and useful there. You're a little farther down. It depends on who's made the MCP server, how much is out there and available and that sort of thing. So, yeah, I I think it's still coming up, but yeah. So scalable if the data and team are ready, right? Yes. Yes, that's the way to put it. If the data and team are ready for it, they can go to scalable um time, but a lot of people are still in experimental. If people are out there telling you that we've fully got an AI situation running and solution and everything done, if they've only been doing it for 6 months, it's not scalable in my opinion yet. We again at the past company we started two years ago and we got to scalable somewhere around a year and a half to two years in. It takes time. Don't don't just think that 6 months is enough to really get you there to the scalable side. So definitely be working on it. Um if you're working on it in your company just know that it takes longer than you think. It's not just a simple thing. You've got to really give it time and energy to figure these things out. So excellent questions. Thank you Dave for those questions. Um, hopefully this starts a lot of discussions for you. I really hope that this presentation was a good way for you to talk to your companies or other people about these sort of things. If they bring up AI fatigue, bring up the last little section about it and talk about things that you can do those human connections. I I want this to be as much a discussion. I plan to post this on my LinkedIn and then have more discussion there and hopefully people will bring up more discussion about it. We need to be talking about these things and you need to be reaching out to others whenever you Thank you everybody. There.