Ep. 330: GTD and AI  ## Metadata - Author: [[Getting Things Done]] - Full Title: Ep. 330: GTD and AI - Category: #podcasts - URL: https://share.snipd.com/episode/a27fef52-ffa9-4f8e-9f39-101f427c994b ## Document Notes ## Highlights - **Transition To Remote And AI Work** - Deb describes shifting her company from multiple offices to remote work after COVID and keeping one office on her property. - Her firm now runs training, coaching, and special projects including AI implementations. Transcript: Deb Smith-Hemphill We used to have seven offices around the U.S. And then COVID came and everybody wanted to work from home. And so we didn't need them anymore. So we kept the office that's on the same property as my home is, but we got rid of all the other ones and everybody works remotely. So I own a management productivity consulting firm. We deal with productivity. We deal with executive readiness. So we've got a group that deals with training and a group that does speaking engagements and a group that heads up our executive coaching team. And then we've got a sort of a special projects group that works with clients on longer term, sometimes cultural issues, or as we're going to talk about AI implementations. ([Time 0:00:54](https://share.snipd.com/snip/ba2d603f-f727-4dd1-a35a-6f822ac03aab)) - **Four Distinct Types Of AI** - Deb categorizes AI into generative, boundaried, genic (transformational), and embedded types. - Choose the AI type that matches the work you want it to do for best results. Transcript: Deb Smith-Hemphill I think it's important to note that there are several kinds of AI, so they do different things for you. There's generative AI, which lots of people are aware of. You ask chat GPT to write something for you or you go to perplexity to get it to write something for you or generate a table or generate a report or whatever. There is also what we call boundaried AI. Boundaried AI means that it's operating as it normally would, but it has boundaries around it. So we have a client, their legal department, their corporate legal department across the U.S. Is the client. Is boundaried. And it means that they can search outside the system, but no one can search inside their system. So their system doesn't release any information, but they can go outside to search or grab information if they care to. But it allows them to assemble and repurpose their information so easily and so well and so fast. One of those clients this morning was telling me that a function that used to take her four hours to piece together information from presentations, from documents, from court documents, Client documents, whatever, now takes her about 10 minutes. So it's a really useful tool in terms of productivity for her. Yeah. John Forrester It almost sounds like she would be spending 10 minutes checking the results as opposed to spending four hours compiling all of that. Exactly. ([Time 0:05:06](https://share.snipd.com/snip/6c91c71e-2a2e-4dd2-9ffc-ea88c3aea5fc)) - **Offload Repetitive Tasks Intentionally** - Identify repetitive or low-value tasks you dread and train AI on the specific criteria you use to evaluate them. - Use that AI to free time for higher-level strategic and coaching work. Transcript: Deb Smith-Hemphill And I said, okay. I said, here's what I want you to answer for me. What tasks on your list make you say, oh, if I have to do one more of those, I'm going to scream. And she said, well, it's not that I would scream, she said, but I review a lot of presentations. She heads up a sales and marketing team. She said, I review a lot of presentations. And she said, I said, so if you had a clone with your brain of brilliance attached to it, who could do those for you, you would give it to the brain, right? And she said, absolutely. And I said, well, I think that's a great thing for AI, as long as you can teach it the criteria that you use. Because it doesn't work. Just like as a manager, it doesn't work to say, I'll know it when I see it and send people out with a half-baked request. And they are constantly guessing, did I get it right? Did I get it right? Did I get it right? Right. Until they come back and get a bunch of responses, they're really not learning anything. John Forrester It's all just trial and error. Deb Smith-Hemphill And pretty much frustration along with it. So I said, what we really have to do is go back and figure out that criteria, because even though you know it when you see it, you weren't born knowing that. You weren't born knowing how to evaluate those presentations. So you form that judgment some way and we can teach that to your AI partner as long as we can kind of capture what it is. And so that's really the thinking. I hear people say, oh, it's going to make humans stop thinking. Oh, no, no, not at all. It's that much more crucial that you think through very carefully what you want or how it is you do something to be able to get the model to take that on. And then it frees her up to spend more one-on time with her folks, to spend more time strategically planning for their future, all the things that she wants to do at her level of work or Above, the model, her AI partner can really start to pick up some of the stuff that really is below her thinking level. And she's capable of so much more if she could just get the time. ([Time 0:10:24](https://share.snipd.com/snip/521f1042-093f-4cb7-b518-7efdc5e471ed)) - **Master The Prompt Window** - Treat the prompt window as your new work interface and invest time in learning prompt design. - Take basic courses (e.g., Google Essentials) to understand what to feed different models. Transcript: Deb Smith-Hemphill And I think they've done a really nice job. So it's a great baseline course to take. The thing to remember is that our new work interface is the prompt window. That's how we're interfacing today. And years ago, you used to collect information, you put it in a file folder, you would file it a certain way, you'd put it in a digital folder and you'd name it a certain thing so you could Find it later files and folders were that was king you know and yes we saw clients build these massively you know intricate systems of files and folders. Right. And we have them as well as productivity consultants. And I remember very clearly, we had a Windows server that crashed twice in the same year. And if you're a productivity consultant and you can't see your own calendar, that's sort of not good. John Forrester Nope. Deb Smith-Hemphill So decided we would move everything to the cloud. And Shenentech is a client of ours and was at that time as well. And they had just standardized on Google. And I thought, well, if they did it, I could do it. So we moved everything to the cloud. ([Time 0:15:33](https://share.snipd.com/snip/b768447f-01fb-477e-8ce0-ee1ae0fe2bc5)) - **From Folders To Search Conversion** - Deb recounts switching her company from folders to cloud search after a server crash and learning to rely on search. - She forced herself to retrieve only via search for two months and never went back to complex folders. Transcript: Deb Smith-Hemphill So decided we would move everything to the cloud. And Shenentech is a client of ours and was at that time as well. And they had just standardized on Google. And I thought, well, if they did it, I could do it. So we moved everything to the cloud. And then I realized we didn't have to have windows computers anymore we could have whatever we wanted and I'd always wanted to try a mac so I bought a mac and I bought some one-on training And I said to my folks we were 40 people at that 44 people at that time today we're 84 but they're about half as big as we are now. And I just said to everybody, I'm going to try this. And then if it works for our clients and our business, let me know if you want to switch. Not realizing there were 44 people who wanted to switch. They were all like, oh, I can't wait. John Forrester So they were already waiting for you to say, let's do this. Deb Smith-Hemphill And they were, okay no persuasion required we really weren't able to do 44 at all at the same time we had to kind of roll it in but we did but um anyway i went and i took my class with my wonderful Trainer from apple and i said okay the first thing i need to know is how to build files and folders and build the structure on a Mac. And he looks at me and he goes, why do you think that's your job? And I said, I don't understand. And he said, I said, why are you asking me? Of course, it's my job. I have to know where I put stuff. He goes, no, he said, you're supposed to apply your brilliance to creating the document. He said, then the computer is supposed to take it from you, park it someplace safe and bring it back when you need it. And I said, doesn't that sound fun? And I don't trust that as far as I could throw a grand piano. He said, OK, I'm going to make you a deal. He said, I'll teach you how to build files and folders and you can file things that way. He said, but I need you to promise me your end of this bargain is for the next two months. You're only going to retrieve things using search. And I said, okay, I'll make that deal. John Forrester Wow. So he let you have the files and folders because that would make you comfortable, but he seemed pretty confident that you wouldn't even bother using them eventually. Deb Smith-Hemphill Oh, it took like a couple weeks of seeing how wonderful search was. And it brought back to me stuff that even I didn't realize I had written. It saved me so much time and rework of stuff that had already been parked there in the previous files we moved over. And before I knew it, I was just using search for everything. And I thought, this is absolutely brilliant. ([Time 0:16:35](https://share.snipd.com/snip/c424e70a-ef16-4486-848f-6527860d61c4)) - **Context Beats Simple Search** - Context and iterative conversation beat simple search or one-shot prompts every time. - Invite the model to ask clarifying questions to improve outputs. Transcript: Deb Smith-Hemphill Well, I think if you're using AI as a search tool, it's kind of like using a Stradivarius to teach your five-year how to play the violin. It's kind of using a massively brilliant, wonderful tool for something that's way too small. Context is what beat search today. Yeah. So giving it really good context. Here's the situation. Here's what I'm looking for. Here's the outcome that I want to generate. But also here's, you know, we have a whole, we use text expander, which is, you know, an auto typer. You put in a little abbreviation. It gives you the whole thing you put in. And so we have a text expander abbreviation for us as consultants. Each one of us has one. So we'll say, this is who I am. This is my expertise. This is my client base. This is the kind of work I do. This is to give it the whole background. Because context is the thing that's going to beat out search. If it's, you know, I wish I could give credit to the right person for this. And I, I don't know their name because I met them at a cocktail party at a conference. And you know how your ID tag gets slipped around and, you know, I don't, I don't know who they were. It's not my brilliance. It's their brilliance. But he was saying to me, it's the difference between lazy and leveraged AI. Lazy AI, you're an eighth grader and you have an assignment to write a report about Abraham Lincoln. So you go to AI and you say, write report write me a report on abraham lincoln end of story lazy ai right and the teacher reads it and says sorry you're flunked you didn't write this because John Forrester There was no context i'm an eighth grader writing report or i'm i'm an eighth grader in the bottom half of my class writing a report, and AI could write a respectable report based on that Context. Deb Smith-Hemphill So if you were using leveraged AI, you would say, I'm an eighth grader, I have an assignment to write a report on Abe Lincoln, what kinds of things should I be asking for in this report or Asking about in this report? And AI might come back and say, well, you might want to know how he became president. You might want to know about moral conflicts he had during the Civil War. You might want to know the timeline. You might want to. So you're having a conversation asking questions. Tell me more about the moral dilemmas and it would tell you more about that. Or could you lay out the timeline for me and tell me about that? So you really have that conversation, that iteration with AI. ([Time 0:20:50](https://share.snipd.com/snip/0be26054-c308-4259-b56b-b5b3c6ce230f)) - **Use C-R-I-T Prompt Structure** - Use the C-R-I-T prompt structure: Context, Role, Interview (ask questions), Task. - Let the model ask three to five clarifying questions before executing the task. Transcript: Deb Smith-Hemphill I think it's amazing. It's context. So C is the context. C-R is the set of initials. So context. And then the R is the role that you want AI to take in this. The I is interview stands for interview. And in essence, what you're saying is ask me three to five questions or however many questions you want. We usually use three to five. Yeah. That will help you one by one individually, that will help you understand the context and the role better. So you're literally inviting the language model to ask you questions so that it can get more of what it needs in context and role. And then the T is the actual task that you want. Oh, I like that a lot. Or whatever. It's a wonderful structure. ([Time 0:25:03](https://share.snipd.com/snip/84e5b589-ff77-4e8c-8268-d57089e6783c)) - **AI Elevates Thinking Levels** - AI amplifies thinking levels: it can accelerate first-to-third level thinking when guided well. - Those who remain at first-level tasks risk obsolescence unless they shift to higher-level work. Transcript: Deb Smith-Hemphill Exactly right. Exactly right. And if you give it the right information to start with, if you approach it in the right way as a partner, it can move you from first level thinking, second level thinking, sometimes third Level thinking so quickly, so much more quickly than you could have made it on your own. And I think that sometimes the thing that frightens people because they say, well, I'm going to be out of a job. Well, if you only do first level thinking and you're not capable of second level thinking, then that is a possibility. You haven't been able to grow in your work and be able to move up. But there are a lot of folks who I think and, you know, I'm thinking the population like first-level managers and sometimes second-level managers get so caught in the focus of today And solving problems that happened yesterday that they really don't have the time to be doing what they're supposed to be doing, which is coaching folks, strategic planning for their Future, trying to figure out what their folks are going to need in six months so they can start implementing those things now. Right. They go, I haven't got time for that. I got 12, you know, crises to solve. Yeah. And I spend my whole day fighting fires. So I think for folks who get caught in that, AI is going to either fight or eliminate a lot of those. And they have to be ready to jump to that second and third level thinking. John Forrester Yeah. Yeah. Deb Smith-Hemphill Or they may get lost in the shuffle and you hate to see that. You want it to be a tool that really augments and helps people build skills and perform at higher levels rather than eliminating their opportunity to do what they do. John Forrester Right, right. Well, you're very clear there about here's how you would keep it from eliminating by doing higher level thinking and ask it to help you with that. Yeah. Deb Smith-Hemphill Yeah. Sometimes we'll ask clients to, you know, within a job, lay out their ideal day. I don't mean go sit on the beach and go over your ideal day. I mean, what would be the ideal work day? You know, you would come in, you would get a chance to meet with your people. You might have a chance to interface really face to face with some clients. Might, you know, be able to research this. I might have time to plan this. I might, you know, so lay out your ideal day and then let's look at all the stuff that bumps that off course and see how we can get you some help from this partner. John Forrester And that reminds me of the GTD model about the threefold nature of work. Most of the people, when they lay out their ideal day, it's going to have much more to do with planning their work and doing preplanned work instead of responding to latest and loudest Unplanned work. But in reality, I bet a lot of them are saying, oh, I spend most of my time in reactive mode instead of proactive. Deb Smith-Hemphill Exactly. Exactly. So I think there's so much leverage to be had there if we can help people set up on the right track. John Forrester How do you see GTD applying to AI that way? How does GTD thought process apply to different tools? Deb Smith-Hemphill I think it's absolutely, first of all, the thought process, being very, very clear about the outcome you're looking for. Because if you can't be clear about that, then you'll just get something that's kind of good. You know, it's good enough. John Forrester It's better than keeping it in your head, but it doesn't necessarily relieve all of your stress. Yeah. Deb Smith-Hemphill No, not at all. It's interesting to me that some people look at it as it's just, you know, instead of a blank piece of paper, I got something. So that's good enough. Instead of having a very clear idea about what a really outstanding outcome would look like, you know, what would this be if it were just outrageously successful? And moving through that thought process, whether it's a natural planning model for a project or whether it's the outcome of a phone call you need to make this morning or a meeting that You're setting up and you want to have. I'm always stunned at people who pick up the phone, go to type an email, go into a meeting and no clue about outcome they wanted to generate. It's just, you know, what's going to happen is going to happen. No, you go in by design. And so I think outcome thinking is one of the most valuable things that there is. And the natural planning model is a wonderful way to think 360 degrees around something. So it's very useful as a way to think through even how you might approach something through AI. John Forrester Yeah. Deb Smith-Hemphill I ran an experiment knowing that I was going to get the chance to do this interview with you. I went to one of our AI tools and I said, you are David Allen. And you are using the natural planning model. And then I laid out information about a project. I said, would you write, go through all that thinking, lay it out for me and write the plan? It did an excellent job. So it pretended to be David Alan. And it's interesting. There's a guy named Jensen Huang, who is the CEO of Intuit. I think. Is that right? No, that's not right. He's the CEO of NVIDIA. John Forrester Yes, that rings a bell now. Yeah. Yeah. Deb Smith-Hemphill And he gave a speech where he talked about AI as being the great equalizer. Because he said years ago, if you wanted a computer to do something for you, you had to know programming language. He said, now you can just ask it in your natural language and there you go. We'll do it. He said but you have to be nice so start with flattery he said you're the world's expert on whatever you are David Allen using the natural planning model you are whatever so give it some Compliments use some flattery and then ask it for what you want and then when it gives you back stuff you still have to be nice. So tell it it did great. Could I have a little more of this? Could I have a little less of that? Could you focus it more like this? If I don't understand what ([Time 0:26:45](https://share.snipd.com/snip/430402e9-2fa4-4d3c-8352-8d68b9ccb729)) - **Verify AI Outputs Rigorously** - Treat AI like a very smart intern: give context, check work, and don't blindly trust outputs. - Verify sources and guard against hallucinations, especially for legal or academic uses. Transcript: Deb Smith-Hemphill And, you know, when we first start with clients, we suggest to them that they think of ai as a really really scary smart intern who's been with you for two weeks they don't know your business They don't have any context around what you're asking them to do but they're really smart and they really want to please you um more and more we're seeing it die down. But when AI first came on the scene, there was the problem with what they called hallucinations. John Forrester Right, right. I want to please you so much that I'm going to give you something that isn't even real. Deb Smith-Hemphill Yeah, I made it up. There's an attorney somewhere back east, I don't remember where, who had ChatGPT write his brief, which he submitted to the court. And the court went back to him and said, you know, these cases that you've cited don't exist. And he goes, oh, that's my bad. I had AI write it for me. And they said, yeah, that is your bad. And you won't have a law license anymore. And he evidently lost his law license. So it's reducing, but I think that's because in part they recognized that was a problem, but also in part because people are learning how to use it in a much better way than they did before. ([Time 0:34:48](https://share.snipd.com/snip/dc3515df-f897-4e60-84f3-a69f334acb86)) - **Teach Thinking Not Banning Tools** - Instead of banning AI in education, teach students to think and use AI well. - Teaching thinking and prompt craft prepares students for tools they'll use in every career. Transcript: Deb Smith-Hemphill Well, you risk triggering my bugaboo about education, which is we should be teaching people how to think. Oh, really? John Forrester That's radical. Deb Smith-Hemphill I really, I tell you, when I went to school, which was like 150 years ago, so keep that in mind. But when I went to school as a kid, the school that I went to was very unusual because all the other kids around me were learning things by rote. They were learning multiplication tables by rote. They were memorizing all this stuff. Where I went to school, they taught you how to think and search for stuff. And curiosity was kind of the key. And it turned out to be I wasn't I didn't love the school the whole time I was there is there from pre kindergarten to 12th grade. So I was there the whole time with a lot of kids that were way, way, way smarter than I was. I have no idea how I got in and I have no idea how I lasted through, but I did. And it turned out to be a great education because they taught you how to think. And so you could literally conquer anything. You knew how to analyze. You knew how to think. You knew how to approach a problem from all different angles. But we never did the memorization kind of stuff. And I think it was a wonderful thing. And today, when I look at folks who are saying, oh, my students are going to use AI. Yeah, let's teach them to use it well. And teach them the thought process that they need, because if you teach them how to think, the results they're going to generate are great. But it's going to be a tool. You know, it's like when the calculator came out and people said, oh, you can't use that, that's cheating. For crying out loud, I don't know anybody, youwrites an Excel spreadsheet today. ([Time 0:36:39](https://share.snipd.com/snip/afebc88c-5fa2-4bf0-a104-275c35fa5054))