[00:00:01] Speaker A: Hello. Thank you for joining us. This is what Counts, a podcast created by Trailblazer Consulting. Here we highlight proven solutions developed through our experience working with companies across various industries, and we talk about how you can apply these solutions to your company. We share our experience solving information management challenges like creating and implementing a records retention schedule, creating an asset data hierarchy, or helping with email management. This is Lee, and in this episode, more and I will talk about artificial intelligence AI as a tool. More. This is an interesting topic, that's for sure, especially today's day and age. Some people will welcome it, some people will fear it. What do you think about AI?
[00:00:44] Speaker B: Well, I have a lot of thoughts about AI actually.
And you Some people may welcome it, some people may fear it. I think you and I had a little mini version of that a couple weeks ago. You said you were going to use, I don't know, one of the tools out there. Chat to write.
[00:01:02] Speaker A: Yeah. Yes, right, yeah.
[00:01:04] Speaker B: To write a script. And I said, kill me now. So that was a little overreaction. But my feeling about AI, once I kind of sorted through that, was it could be a very useful tool. What computers have always been good at is processing data faster than a human. They are able to search across hundreds of rows of data, thousands of rows of data, looking for patterns or doing calculations. It's not that a person couldn't do that, but the volume would just take so long and be so slow that it would not be worth it. Same thing from a language perspective, which is where I think the AI tools that are out there right now are shining is in looking at words across many sources. And I say it that way, words as opposed to content. Because where I start to really doubt AI is does the AI understand the meaning of everything? Because you have content and you also have context and you have a question of the source. You know, is something being written as sarcasm? Is something being written as a factual statement? Is something being written to persuade or to sell something?
All of the same words could be used in each of those situations, but the context of it is important in understanding the validity of the statement. So, for instance, you could say, oh, you're on fire today, in that sort of a tone. And you might mean that I was being kind of a brat. Or you could say, maura, you're on fire, meaning, whoops, you put your arm across a candle and you set your sleeve on fire. I have never done that, but my sister has.
Then it might be that we often say, I'm caught in a fire drill. And what we mean is there's somebody asking for data or another version of the same slide deck, or an answer to a question that we've answered 50 times, but they need a slightly different turn on it. And it's not. We're in school and we're having a fire drill for how to exit safely in the event of a fire. So the same words, the same content, if you don't understand the context, you are going to misinterpret. So that is a problem with AI, in my opinion, when you're looking across, say, thousands of pieces of writing on the Internet, so you ask a question. And I've noticed lately that some of the search engines, before giving you links to websites, they will give you their own AI summary of it. And some of them are very good and some of them are not.
But you have to really look at it and you have to understand where did this come from and what was the source before you can know, how do I interpret this? So it's neither good nor bad, but it could be dangerous. And to me, that makes it that. To me, that makes it a lot like other tools. There are always safety requirements around tools. So I recently been trying to hang shelves. It's been going on for months. I bought an electric. I had an electric screwdriver that was not up to the task. I bought an electric drill. Turns out I don't know how to use an electric drill.
I did not hurt myself, but I did put a lot of holes in the wall. So I needed something else to get those shelves installed. And in fact, what I needed was a professional, and that is what I found. Hired a handyman. He hung the shelves. He did not put any extraneous holes in the wall.
[00:05:17] Speaker A: Well, AI is not going to help you there.
[00:05:20] Speaker B: It's not, but it's an analogy. Whereas I feel that I am an expert on words and I'm an expert, I've been portraying myself anyway, as an expert here on this podcast for three years now in information governance. So when I'm talking about how to organize your files or how to manage your data in accordance with your business process so that you make it easier to file your data, to capture records as part of your business process, so that you can apply information classification requirements or regulatory reporting requirements or retention requirements, I'm. I'm coming at that with a lot of experience and authority, and I can point to the sources for my expertise.
If some AI tool uses the same words that I'm using because they've picked them up in a large search, but they're trying to apply it to how to drill a hole in the wall. It's not going to work. So that's, that's what I think, is that AI can be a powerful tool and if you understand how to use it safely, and if you don't, then you can get yourself in trouble. We've heard some stories. We heard about lawyers who submitted fake case citations in briefs. That was an early story in the CHAT GPT space. There's been a lot of reporting in the academic world of college students in particular using ChatGPT to write essays and papers. And like, the stakes are kind of high there. You could be suspended or expelled for academic ethics violations. You could get a bad grade.
At the very least, you're not learning anything if you're doing it that way. So those are sort of unsafe ways to use AI. But the safe ways to use AI could be. Okay, I have a set of data that I am confident of its validity. I understand where it came from, and I would like to use an AI tool to help me synthesize some of that data into a draft of an essay or a script. So actually your original I'm going to try ChatGPT to write this script for myself was not a bad use of it, even though I overreacted. But it. You knew the sources you were looking at and I'm assuming. But you can confirm you didn't just accept what that AI tool gave you as the script. You read it and you compared it to the work we've done and the requirements we've produced for our clients and other things. Because it was a script about requirements gathering. And so you looked at it in the context of I know where my sources are and I know what this topic is about. I understand it. I'm not just relying on this AI tool to tell me what's real.
[00:08:35] Speaker A: Yeah, that's absolutely true. And I think you're kind of dancing around something in my head. You're dancing around something which is if you use it for something sophisticated, it's the best word I can come up with. But maybe it's scholastic report or something like that, your thesis, whatever it may be, I think you're going to get yourself in trouble if you use it for something simple. I think it can be very valuable and speed up your time to getting your task completed. This instance. Yes. I said I want an introduction to requirements gathering. And it was. It gave me a lot of information.
I used about 40% of it.
[00:09:21] Speaker B: Right.
[00:09:21] Speaker A: So that's pretty good to get 40% of your paper kind of written and then the 60. The rest of it you're able to do. I thought it gave me a good path forward and so forth. So that was a good use of it. If you get into more sophisticated things, I think you got to be a little leery. I'm trying to do. I'm trying to figure something out and we can use this or not, but when did. What's it called when you compared documents? Way back when, like in 2, 2010, the Information Governance we had this big compare documents to see if they were, you know, the same or if they were ready for retention or something like that. And that sort of AI has been used. It has to been increased in the abilities to do this searching. What was it called though?
[00:10:17] Speaker B: I'm not sure I can think of the exact.
The exact phrase you're looking for. But it was like a side by side comparison. There were elements of keyword in context was one of the things. There were relevance rankings. How often did a certain word or synonyms of a word appear in a.
In a passage which raised the relevance of this passage and it also raised. So raised the relevance in response to your query. It also raised the confidence level that this was an accurate result. So you asked a question and the computer, the system gave you back 10 answers and it said these 10 documents we think are relevant to your question. And we think these three have a very high relevance rating and a very high confidence rating. And it was based mostly on the occurrence of certain words, the frequency of certain words in the passages, words or synonyms. And these other seven might be relevant, but we have a lower relevance rating and a lower confidence rating. So there were. So there was like those combinations of tools. And I actually think that's a great point that I should have made. So thank you. Which is AI isn't new. It kind of feels like it burst on the scenes in the last two years. It's actually been around for a couple of decades. We had, we had automated classification tools, early automated automated classification tools in records management. That was the one you were looking for. Okay. Where based on the content, again the keywords in context of the documents, a tool would read it and attempt to file it in your filing system or match it to a record category. And it worked better when there were more documents for the tool to review. And it got better the more it reviewed. And it also worked better when there were fewer categories. It was. It kind of came along around the same time as the big bucket theory of records retention and moving away from the very Detailed granular retention schedules with hundreds, if not a thousand record categories, which had really grown out of the world of filing paper records because you needed all that granularity in order to find things in a file drawer or a file room. So these tools have been building on each other based on practical needs and also the fact that the more information became available electronically as opposed to in paper, the easier it was for a computer to do what it's good at, which is read a lot of words quickly and compare them to each other. So I think those early iterations, those have become very strong, like that automated classification is actually much stronger today than it used to be. Fuzzy logic is one of the foundations of ediscovery tools. So that ediscovery tools are no longer just searching for a specific word list. They are searching for concepts that are represented by specific words. And they can find things that are, you know, within rankings of relevance and confidence. And they can tell you what they've done, the tools can tell you. So I think those are all good examples of how a person can harness some of the power of AI to help you with your job, which you. I went down this whole path because that's what the question you asked me. But at the beginning of when you were talking, you said something about more sophisticated uses might be more risky. And I think that's not untrue. But I think really the point is how much you rely on the AI tool. So if you rely on it as a starting point to help you kind of do a big perusal of a lot of data and focus in on some key sources or the key points, and you're going to apply your own knowledge and your own test and you're going to do your due diligence and add your thoughts to it as you're building your thesis, then it's a great tool. If you're gonna set, you know, one of those tools loose on the Internet and say, write me a 10 page paper on, you know, photosynthesis, then you're gonna get caught and it might not be right. You might find some wrong sentences in there because somebody wrote a sarcastic or false statement about photosynthesis that showed up in your paper and you didn't know any better. So I think that's, that's the message for me is we're not going to leave, we're not going to be left behind by AI, we're not going to be replaced by AI. There is still a need for a human person to review and, and understand what they are then taking Ownership for and putting out into the world. But there's no reason to ignore it and there's no reason to pretend that it's all bad because it's not.
[00:15:35] Speaker A: So I agree with that. I think that's a good close too, if you're finished. Yeah. Auto classification.
[00:15:45] Speaker B: Auto classification, yes.
[00:15:46] Speaker A: That was definitely what I was trying to figure out. Yeah. And it's come so far right now.
[00:15:52] Speaker B: Because it was really bad. When we first looked at it in 2003 or four. It was really bad.
[00:15:58] Speaker A: Sure. Now you can take a 500 page contract, throw it in to, you know, a scanner or whatever and it populates the correct fields.
So that's much farther than what we were talking about way back when.
[00:16:14] Speaker B: Exactly. That, that ingest of a document like a contract. It's the. It used to be we had intelligent, we had optical character recognition where the a scanner could just tell you what the words were. Then we had intelligent character recognition which if you had forms, the scanner could read the form and put the data into the right fields. But now it's come to the scanner can read a sentence that says the parties of, you know, Lee Karris and Maura Dunn have agreed to enter into this agreement on such and such date and it can read that sentence even if it's slightly out of order. On today's date, Lee and Maura entered into an agreement and it can fill in the data fields based on understanding the meaning of that sentence.
And that's a huge leap forward from where we were in the beginning of auto classification. Yep.
Cool. So don't be afraid of AI, but use it wisely and also hire professionals for your shelving needs.
[00:17:21] Speaker A: If you have any questions, please send us an email at
[email protected] or look us up on the web at www.trailblazer.us.com. Thank you for listening and please tune into our next episode. Also, if you like this episode, please be a champion and share it with people in your social media network. As always, we appreciate you the listeners. Special thanks goes to Jason Blake created our music
Speaker B: Also, don't hire me for your shelving needs.
Okay, yeah, you can end it.
