How Your Personal Prompting Style Affects Your AI Experience

Episode Notes

Transcript

Announcer:
Web 2.0. Innovation, collaboration. Got the world turning as fast as it can. Hear how technology can help, legally speaking, with two of the top legal technology experts, authors, and lawyers, Dennis Kennedy and Tom Mighell. Welcome to the Kennedy Mighell Report here on the Legal Talk Network.

Dennis Kennedy:
And welcome to episode 427 of the Kennedy Mighell Report. I’m Dennis Kennedy in Garrett, Indiana.

Tom Mighell:
And I’m Tom Mighell in Dallas.

Dennis Kennedy:
In our last episode, we continued our Fresh Voices on Legal Tech interview series with Ariana Paulino, technical services and reference librarian at Ponza Maur Law Library at Nova Southeastern University, exploring her law librarian and law student skills from on the ground, getting her perspectives on tech competence, collaboration and AI. It was a great illustration why focusing on the work right in front of us can be the best way to learn how to use tech and AI. In this episode, we are stepping back to compare our personal interaction styles with AI, and I mean specifically how Tom and I actually talk to our AI in fundamentally different ways and looking at a new experiment in audience profiling using AI. Tom, what’s all on our agenda for this episode?

Tom Mighell:
Well, Dennis, in this edition of the Kennedy Mighell Report, we will be taking a look under the hood at our contrasting personal prompting philosophies and specific dialogue tactics from sideline play cards to directive tasking, whatever those things mean. Then in segment B, we’ll look at an experiment Dennis Ran using AI to infer our listener profiles and this podcast’s jobs to be done in the total absence of traditional survey data. And as usual, we’ll finish up with our parting shots, that one tip website or observation that you can start to use the second that this podcast is over. But first up, we want to talk about what our… We’ve been talking as we prepare for our other podcasts. We always talk about the things that we’ve been doing and how we’ve been using AI. And the one common theme that we’ve been figuring out during our pre-recording sessions is how different our styles are when we are working with AI.
So we thought maybe there’s something there, maybe there’s something to talk about. So we both use AI daily. I guess maybe we start with our philosophies of interacting with AI. Would that be the right place to start? And if so, how would you describe your philosophy when you sit down to talk with an AI tool?

Dennis Kennedy:
Well, Tom, first of all, Best, I have a little bit of fear on this episode and it goes back to our PTI shows where we have these segments and you always win. I just have this feeling at the end of this, I’m going to be converted to your approach, but I’m going to give a strong try.

Tom Mighell:
No challenge here. No challenge here.

Dennis Kennedy:
So I would set the stage this way. When I talk to people, most of the time people assume there’s sort of one correct way to prompt and you can learn it in an hour or so in a CLE session. But I actually have found over the last, whatever it is, four years, that your interaction style with AI directly dictates the depth, the tone and the rigor, the answers that you get back. And you learn more about how AIs actually work and what they can be used for as you vary your approaches to prompting. And I think it’s a lot more fun that way. Although I got to tell you, Tom, I can’t believe the number of times, even in the last year, people have told me like, “Oh, you don’t know exactly how to prompt. The problems you’re seeing must be because you don’t know how to prompt very well.” Even though I’ve been doing this for a long time and written several, what I though were influential articles, but apparently they weren’t because people come back and tell me to do the thing that I’m doing all wrong and that I should do exactly what I wrote about, which unfortunately was good advice three, four years ago, but AI tools have changed and it doesn’t work as well anymore.
So I want to start with the basics. There’s no magic bullets on this stuff. What I can totally confirm from my approach to prompting is that your typos don’t matter because I feel like I’m setting world’s records on typos and the AI does just fine. So my method, my philosophy, I would call it is collaborative, it’s challenging, and I shift styles a lot. So I treat the AI like it is a high level thought partner that needs to be challenged. And the key is adapting my voice and communication style depending on the specific type of session that I’m having. So I might shift my tone, I might shift personas, I have different protocols that I use and stylistically I do different things. So that’s the one part of the philosophy. And then the other thing is that I’m not so concerned. Actually, I’m not very concerned at all with the answers that I get because I rely on a harvesting approach.
And so I do a lot of things that are very unusual in sessions and I explore tangential ideas, I test the AI, test edge cases, I gather different things across really long sessions. And then at the end I synthesize the usable crop that I harvest at the very end. And so why I do that most of the time is that it uncovers unexpected nuances, keeps the sessions organic, keeps it fun for me and forces the AI pass its generic consensus answers into doing something that feels closer to actual, I don’t know if that’s the right word, but to a reasoning that I find really helpful. Tom, what about you?

Tom Mighell:
I will be honest, Dennis, when I hear you say intentionally non-linear sprawling threads, I synthesize the crop at the very end, my first thought is, oh my gosh, that sounds like it takes forever. And I guess for someone who is retired forever isn’t necessarily a bad thing. I don’t have that kind of time. I have objectives for my AI to meet and I don’t have the time to explore or test edge cases or having long conversations. Now, I think you definitely understand with the way that you approach it, I think you definitely understand how these tools actually think better than I understand them because you poke at them. I don’t, I just want it to work. I just want it to do the job. And so my style is pretty straightforward. I treat AI like a partner that I’ve onboarded. I set the rules, I set the protocols, I make sure it understands the job, and then I turn it loose to do that specific job.
I build the thing, I build the framework, I ask the AI to build the thing, I correct it, I use it. I then go back and reuse it and keep improving it. Every project that I have has what I call a state of play for each one so it knows where we left off. It has standing rules that we update as things change. So it always is knowing what I want and what I need that are based on the way that we’ve worked together in the past. So I think of it less like prompting and more like onboarding a new hire who I am slowly getting used to my way of working.

Dennis Kennedy:
So I think this is the key contrast. So Tom sets clear bounds, gives instructions and gets something that’s a linear deliverable. And I think that is sometimes what I want to do, but I also find it increasingly difficult to achieve in the current versions of the AIs, which is a topic for another show, frankly. So what I find really useful is that I’m doing exploration sessions where I make a lot of shifts to see what the model can actually do for me. And that may mean breaking boundaries and I want to see what new things emerge. And so Tom says, yeah, it’s going to take more time than if I say, give me the three cases that most fit my fact situation. I agree, but I’ll do something where at the end I’ll have a much clearer understanding, a breakthrough idea, a completely different approach to what I want.
And that’s where I feel that AI gives me the most value. And it’s part of, as I’ve tell some people, these days I see AI as my art and my median. And so what I’m doing has less to do with practice of law than it ever has. So I would sum up Tom, by saying that you treat the AI like a well-managed employee who gets clear directions from a great manager like you. I treat it like panning for gold in a river stream and collecting the flakes that might turn out to be gold or might not want to examine more carefully. And the B segment I think will actually show that when we get to that. And I have the suspicion, Tom, that when I describe what I’m doing, it’s horrifying to you.

Tom Mighell:
I’m not horrified. It’s one of those things that I just can’t picture myself necessarily doing. So here’s my question for you. You pan for gold, you synthesize the crop at the end. So what do you actually do with the crop? What do you use the flakes for?

Dennis Kennedy:
So what then that gets turned into is ideas for articles into new things I do with AI, into brainstorming things, into paths forward, into things I want to explore, into, oh, I didn’t realize there was connection between this and that. How would I learn more about that? And so I’m using it. I mean, it really is the brainstorming notion of divergent thinking, convergent and that double diamond where I’m saying I need to get more and more ideas and then narrow them down and then to go to bring out and then to also use techniques that I might be able to use if I’m in the right setting, which I’m not, but I could use an AI. So I could do the Toyota 5Ws. I could do jobs to be done. I can do all sorts of stuff using AI and see where it takes me.
And so that’s what I find is really intriguing because I’m not in a world these days or in a job where I need to say, oh, I need to do X. It has one answer. Can AI give me that answer? I’m really looking in a completely different way of doing things. So my approach to AI is I will always tell people, my approach to AI is way different than what most people’s really is. And so my approach to prompting has to be much different.

Tom Mighell:
Right. And I think that’s the difference. You talk about mining for flakes of gold. I would make the argument that I’m mining actual chunks of real gold with the products that I’m creating. Last week, I built an entire change management and communications plan for a client in just under an hour, I think is what it took. And it used to take something that took 10 to 12 hours. To me, that’s real gold that we’re creating there. So I will take well-managed employee, but to be fair to my well-managed employee, one of my rules is that it’s not a yes person. It’s got to push back. It needs to tell me when I’m being stupid. It needs to stop telling me I’m right all the time. And I definitely want an employee who’s allowed to say, Tom, that’s a bad idea because I need it to be reasonable with me and I’m always asking it to poke the edges of my arguments and find out what’s right.
So it’s a well-managed employee who nevertheless is not afraid to talk back.

Dennis Kennedy:
My original process with AI was designing a bunch of protocols and saying, okay, now protocol based on these things, generate exactly what I want. And over the last year, especially that started to break down. And so what I’ve found is that I can do something or I can get an answer or a draft of something in a very short time. And it might be good enough, but if I keep persisting with it, I’ll go in some different directions and what I’ll end up with is almost infinitely better. And that’s what’s intriguing to me about AI. I would also say that obviously sometimes it’s incredibly better and other times it’s a fail. And I just did something this week that I thought really had a lot of potential with AI that I had to fail completely. So I’m taking a far more risk with AI to get what I hope is better return.

Tom Mighell:
All right. We’ve talked a little bit about our philosophies. It’s time to actually put things into practice. We’ll do that after a quick break for a word from our sponsors. And now let’s get back to the Kennedy Mighell Report. I’m Tom Mighell.

Dennis Kennedy:
And I’m Dennis Kennedy. We want to remind you to share the podcast with a friend or two that helps us out. In this part we wanted to look, as Tom said, we want to say, what does this actually look like in practice? And so I’ve kind of talked about that. I want to move from a philosophical approach and a practical approach, what I used to call a protocol approach to saying, can I use specific conversational moves in harvesting tactics to get closer to what I want? So what I’ve found is that I’m going to use this term called default attractors. So AIs tend to, in the simplest form, there are the average consensus answers that they tend to move toward. I call those default attractors. I think that’s a good term for it. I think it’s also used elsewhere, but it sort of gives you the average answer and it sort of flattens the response.
And so you get this sort of averagey thing out of AI. That’s exactly what I don’t want. And so what I’ve been doing lately is to say, in an AI session, I will give a protocol. You can call it a persona if you want, and then I will start to use that and then I’ll make shifts and do improvisation within the session. And the whole idea is to get away from the default attractors. So I could say something like, “Here, give me this and AI will give me the top five reasons to do this and I’ll read like every other listicle that you would see.” So I’m trying to get away from that. And so what I’m trying to do is use the prompts that I’m doing now to kind of shift the AI off of the averageness that it has. And I don’t mean it that it’s something that’s thinking.
I’m just trying to change the probabilities and the math behind the AI by the prompting that I do. So it’s more like brainstorming and I just sort of keep testing, see how does the AI’s perspective adapt and can I get it out of the average way of doing things? Okay. So this sounds more complicated than it really is. So I told somebody the other day was asking me about prompting. I said the prompt I use more often than anything else these days is three words and it is tell me more. And so the notion here is I use brief open-ended nudges during the deep dives to just let the AI expand on what it’s saying without oversteering it. And then I sometimes will say, tell me more, go deeper and that will go into more depth. And again, I’m just looking to see if it will give me a bit more that would be useful and kind of move off the average answer.
Then I started to play more and I said, “Well, what if I had a whole list of these that I could start to use?” And I used the notion of football coaches have this play card that they look at during games that tell you in this situation you’ve called this play. And so I would say, if I have a list of these and AI gives me a response, can I just prompt and say, “Where’s this argument at the weakest? What am I missing?” Other things like that. Then I also do just standard conversational improv. So sort of like yes, and can I take the model’s output, let it wander? Can I do the classic, just take the last sentence that it gave me and put a question mark at the end of it and turn that into a prompt? And then I also do a lot lately of saying, “What did you mean by this?” And then it will expand on that.
So it’s really become a more conversational, improvisational, and then I’m not really specifying too much what it’ll do. I’m letting it go to see what comes out and then I harvest later. So again, Tom, I probably horrified you with that, but that’s my current approach.

Tom Mighell:
So question for you though. How do you interact primarily? Are you typing or are you talking? Would you type your prompts mostly?

Dennis Kennedy:
I type it. I’m going to switch to voice here soon-ish because it makes more sense.

Tom Mighell:
And I wonder whether that is part of the difference as well, because I use WhisperFlow for everything, for almost every single thing. So I’m talking to it, my prompts are by nature, conversational, like I’m briefing a colleague. I’m just thinking totally off the top of my head. I think your prompts tend to be more surgical. You craft them. At least when you’re typing them, you have better opportunity to do that anyway. I ramble. I’m horrified when I see what it puts into the AI, but it’s still, like you say, it gets it all right. And so what I ramble and then I let it ask me questions until it understands. And I think that is one main difference.

Dennis Kennedy:
Yeah. So I did this prompt earlier today where the entire prompt was and question mark, and the AI ran with that, which actually shows a danger of AIs in how they’re set up to always complete an answer and to move toward completion. So one of the things I’ve been trying to do in prompting is to get AIs to say, “Hey, look, if you have a question, ask me that question.” So I will just flat out say, “Ask me any clarified questions.” But there’s a prompt that I do that I’m using now that says, “What did you assume that I wanted that you didn’t ask me?” And it will tell me some of those things and it will kind of riff off of those and that will give me some new paths to go. So I’m looking at the look on Tom’s face, and I’m realizing that at the end here, I’m going to go in a direction that he absolutely cannot predict.
So Tom, tell us about your approach in comparison to my, what I’ll call very jazz approach.

Tom Mighell:
Frankly, I think I’ve said most of my approach already. There’s not a lot more to it. My main tactic is that I’m not starting from scratch, and I think that to a certain extent, you’re not necessarily either. You build protocols. I don’t know how much you still use the protocols. I build skills. I have skills that are basically saved instructions for things that I do over and over. I ask it to load a skill. I had it built a skill for one that writes in my voice. I fed it all of my writing samples and it writes in my voice. I have a skill for preparing for this podcast. I have a skill for work operations, and it already knows the rules. I have a standing instruction in all of my skills to ask questions before it starts. And frankly, for me, that one habit fixes most of the problems of it didn’t understand me, is that it’s asking enough things to where it gets most of the things that I need to think about and it asks them.
Granted, it won’t get everything, but it does a pretty good job. And for anything ongoing, I mentioned this earlier, all of my projects keep a running state of play, what we decided, what’s open, what’s next. I’m never re-explaining myself to say, “Oh, by the way, we’ve talked about this before.” Even if I open up a new chat, it can then still look at the state of play to know what’s going on. I’ve also created a way to… I have instructions in each of my projects to use Todoist to create tasks, and they’ve created a way to talk to each other, to have the different projects talk with each other and make sure that everybody is kind of on the same page. I’ve created similar to you, a chief of staff. So my chief of staff, if one project has something for another project to work on, it leaves a task and a special folder in Todoist that the other project will pick up on and go run with it.
And so I guess it’s a roundabout way of using agents to do it and not quite the same type of agent, but I definitely will have those things set up. That’s pretty much it. My style is, like I said, I talk to it. I say, “Here’s the thing that I want to accomplish. Let’s have a conversation back and forth about what it is so that you are satisfied with what you need, and then I want you to go build it and then let’s work from there.” So I’m pretty straightforward with mine, very vanilla, but I get what I need.

Dennis Kennedy:
Yeah. So what I found is that in my original approach was similar to yours, but without using the sort of built in the skills things because they didn’t do some of the things that I wanted, so I didn’t go in that direction, was that they tended to drift so much that a lot of what I was doing was trying to fight that drift, that flattening and other things. And so I can hear it now in our conversation, but my insight lately has been, and this comes up with agentic AI, because I will do things in AI where it will be drifting all over the place, going in its own directions, making assumptions, all kinds of things where I go like, “There’s no way agentic AI based on LLMs can possibly work.” So that’s my thought. But my new thinking is that there is a realm where AI makes sense, where you need the variation, you need some approaches.
There’s things that AI does really well, but a lot of things have to be converted to software. And I think this will be a topic for future podcasts, Tom, and we can stop here unless you have some thoughts on this. But I think what we want to do is we want to minimize where we rely on AI as it currently exists and maximize how we turn things into code and software. And AI can help us. I mean, that’s code, that’s other things. It can help us with some of those things. But I think there is a realm that requires software and there’s a realm where AI can really help us. And if we keep really clean lines on that, we’re going to be happy with AI. And if we blur those lines, we’re going to get the hallucinations and everything that everybody fears.

Tom Mighell:
This may be a little bit off topic, but I read a piece recently online. The website Every is something that I subscribe to because they’re using AI in very interesting ways, and I follow what they do. It was called How to Create Your Own Personal AI Benchmark. His point is that every time a new model comes out, they say, “You should switch to this, and this is now the new place we want to move to, and this is the best.” But his point is that benchmark that everybody’s arguing about testing don’t actually tell you whether a model can help you with your job. One of the big tests asks things like the title of a cheap trick live album, which is totally useless to me in my job. His practical tip that he offered is something that he calls a back pocket eval. So for those of you out there, evals are short for evaluations and it’s ways to test the AI to see if it’s doing what it should be doing.
And a back pocket eval is keep a list of the things, the tasks, the things, whatever, that AI failed for you. And every time a new model comes out, try it again. And if it can suddenly do it now, is that a model for you to be using? Is that showing progress? And here’s the embarrassing part for me. I tried to come up with a back pocket list and I couldn’t. I couldn’t come up with anything. I’ve always gotten something out of AI, even when I wasn’t thrilled with it, which means that either AI is amazing or I’m not asking it to do anything hard, and I suspect that it’s the second. The guy in the article, the guy that wrote the article, he calls that the discernment horizon. If all your tasks are easy, every model looks exactly the same, and maybe that’s my issue.
If you ask a genius to make coffee, you’ll never find out that he also knows about quantum physics. I guess for me, that’s sort of the answer to our own approaches to AI. None of us gets to declare our way as the right way. I think, and I said it before a minute ago, I’m paraphrasing the Rolling Stones, the test is your own work. Does the way get you what you need? And that’s kind of how I end up on this.

Dennis Kennedy:
Yeah, I think you’re absolutely right, Tom, that they say to each his or her own, and you need to do what makes sense. You need to focus on what job you want to accomplish and whether the AI fits that job, and are you asking AI to do something that it’s just not good at or can’t do, and to do less of that and do more of what it is. Somebody asked me recently whether they could just watch me while I was prompting AI. I was like, “Oh my God, this would be the scariest thing you can imagine.” Terrifying. But if I showed you what I came up with at the end of it, you would go like, “Oh, wow. I would’ve never thought you had gotten there and it’s so clear and it’s so insightful.” But to get there, it was a whole lot of work in its own way.
But it’s something that for me, I really enjoy doing in a way that somebody else would say, “Geez, why are you doing this when you could just have the AI create a little bit of code and then structure your way down there?” It’s like Tom, you always say, “There are pilers and filers, and this is another way that I’m a piler and you’re a filer.”

Tom Mighell:
Yep. All right. We’ve got a lot more to talk about, but before we move on to our next segment, we need to take another quick break for a word from our sponsors. And welcome back to the Kennedy Mighell Report. I’m Tom Mighell.

Dennis Kennedy:
And I’m Dennis Kennedy. I wanted to try something with the AI that I’ve done in the past, but I wanted to do it in really specific way. So traditional audience surveys for podcasts are notoriously slow, they’re low yield, and they’re prone to response bias. So people select whether they participate or not. Tom and I were talking about, “Well, who is our audience?” I was like, “Well, what if we just do an experiment or I prompt an LLM to analyze our content history, our topic vectors, our industry positioning and other factors and try to infer who our listeners actually are and why they tune in to see if that might work just directionally as a stand-in for an actual survey.” So that’s the experiment, and now we’re going to see what it came up with, and then Tom gets to react and say, “Wow, that makes sense,” or, “That doesn’t make sense,” or, “How Would we learn more about those?
It was really the purpose of an experiment like this is to say, what will we do next to learn more, to confirm or disconfirm, I guess, what it came up with? So the AI said that we have three types of listener profiles. So the first one is the pragmatic innovator. So that’s a mid to senior lawyer or legal tech leader who needs practical hype-free advice on what technology to adopt today versus what to ignore. The second one is this strategic technologist, might be an IT director, Lawyer Librarian or legal ops person looking for high level frameworks and trend synthesis to bring back to firm leadership. And the third is the curious explorer. It could be a solo practitioner, a law student, an individual lawyer seeking a low friction conversational way to keep up with the internet and AI space, legal tech in general, without drowning in technical jargon.
I kind of like these, Tom, what do you think?

Tom Mighell:
Well, first, I have to point out that we did a version of this. I prepared for this podcast using my AI second brain, which pointed out to me that we did a version of this back in episode 388 where we had ChatGPT guess which AI tools you were using at the moment. And it got Gemini and Claw exactly backwards. So I know from experience that AI can make very confident guesses about people from what they say publicly and still get the important stuff wrong. So here’s my problem. The listener that I picture when I talk is a solo or small firm lawyer who does their own tech because nobody else is going to do it. That person shows up in only one third of one persona that you had lumped in under the curious explorer. Neither of us has ever worked in big firm legal tech.
I work in information governance. You taught, you worked in-house before that. I don’t think an IT director at a large firm needs us to tell them what Copilot does. We certainly don’t talk about legal technology for law firms to adopt unless it has an internet focus, but I will be fair to the AI here. I think it described our Fresh Voices guest list really well. I went back and I counted and I think more than half of our recent episodes are interviews mostly with big firm innovation folks and academics. So maybe the AI didn’t profile our audience, maybe it was profiling our booking list instead, just a guess. I personally think the AI did a great job in describing who you want as our listener, but I still go by my grassroots experience of who has ever reached out to me after a podcast or who has given a shout out that we appreciated what we talked about.
And it’s almost always a solo or small firm lawyer. So I am skeptical of the result, but interested to know the truth and we’ll talk about that in just a minute.

Dennis Kennedy:
These categories feel right to me based on the limited amount of feedback that I get, especially recently. But I think that what you’ve also shown our listeners is a great way to think about the answers AI gives you because you can understand how it would look at this Fred Voice’s series and draw the conclusions that it did, which is why I would always want to dig deeper. So the second thing I asked us to do is say, because we always talk about job to be done here on the podcast. And I said to the AI, “What do you think are the job to be done of this podcast is?” And it says, “Oh, the main one is to filter out legal tech noise so I can spend my limited time only on practical tools, clear risk and actionable strategies.” And then there’s a secondary job.
It says, “Give me an enjoyable collegiate dialogue between two trusted experts that reassures me I’m not falling behind on internet and AI trends.” Tom, your reaction to that.

Tom Mighell:
So I buy this more than I buy the personas. Reassure me that I’m not falling behind is honestly the most accurate thing I think that I think about with our podcast. But to me, that’s still a solo lawyer. That’s who we’re talking to in my opinion. I sort of feel like the jobs to be done part is describing a solo practitioner and the personas part that we just talked about describes a law firm innovation department. So it feels to me like the AI is a little bit contradicting itself when it’s talking about both of these, but I buy the jobs to be done piece a little bit more.

Dennis Kennedy:
Yeah, I like this because this is something that I think both of us think about as we say, well, and when we’re coming up with new topics, we go, well, what is practical? What would be good tools to use? I mean, that’s what our collaboration tools book was about. We are concerned about risks, also benefits. I mean, that’s a whole tech competence thing, but we always want to say, well, what can people do with this? And so I think there is this notion to say, I can listen to these two people who’ve been around talking about this for 20 years and I feel that I’m keeping up to speed with that. So the last part is, and we’ve kind of hit on this, but maybe sum up, is I also asked the AI to say, how well did it do? And it said that it felt that it was spot on, on the jobs to be done filtering mandate.
So we’re filtering out the legal tech noise and on the professional personas. And I guess where Tom and I differ on this is I think that it is right that our audience personas are individuals and individual practitioners, but I’m not as convinced the type of organization they’re in, but I think they are individuals. And then it said what it missed, and this is kind of interesting, is that there’s a subtle community element that the listeners tune in for our longstanding host dynamic and the personal banter, got to love banter, that doesn’t show up when you just look at the list of topics. So Tom, how did AAI do about assessing how well it did itself?

Tom Mighell:
I love when you asked something to grade itself, but I’ll bring my benchmark article that I was talking about back into the conversation here is it graded it, but it graded against itself. We’ve never surveyed our listeners, so there’s no right answer to check it against. It’s grading itself on vibes. I mean, it’s just like, “Hey, I felt I did really good on this part.” So the advice that the benchmark article gave is to skip scoring and ask yes or no questions. So question, did it identify the solo and small firm lawyer as our core listener? Not really. Did it get the job right? Jobs to be done? Yes. Did it catch that people come for the banter? No. My verdict is it’s a great starting point, but I would argue that it’s a hypothesis and not an answer. And so the fix for that is easy.
Ask the actual humans. So listeners, I would say help us settle this. Find us on LinkedIn and tell us three things. Tell us what you do, the size of your firm or organization and why you listen. A DM on LinkedIn is 30 seconds to answer those questions. I would love to hear from you all so that we can see how well Dennis’s AI did in solving this issue or predicting and estimating who our audience

Dennis Kennedy:
Is. Once again, we go back to our early school days and it all comes down to the scientific method. We got to get some data and figure out what’s going on. But now it’s time for our parting shots at OneTip website or observation you can start using the second this podcast is over. Tom, take it away.

Tom Mighell:
So my parting shot is an AI tool from Google that I just discovered and it is called Google Weather Lab. It is part of the Google Deep Mind set of projects. Think of it as a public window into Google’s experimental AI hurricane forecasting. You pull up a world map, you can click on a storm and you’ll see where the AI thinks that it’s headed, how strong it’s going to get up to 15 days out. And instead of one line on a map, it shows you 50 possible paths so you can actually see how confident or not the forecast is. But it’s not just tracking hurricanes and cyclones. You can see precipitation predictions all over the world for up to a month in advance. And it’s very interesting. It is accurately predicting the fact that we are about to get rain from Hurricane Polo here for the next couple of days.
So it’s a very intriguing use of AI to predict the weather. Just remember that this is a research tool, not an official forecast. So if a storm is coming your way, please listen to the National Weather Service and not a website and definitely not me, but this is still kind of a pretty cool tool to use. And Dennis, no harvesting or gold mining required.

Dennis Kennedy:
The weather stuff is really interesting because this morning I made plans for tomorrow based on the fact that it was going to be essentially a perfect day. And tonight I looked at the weather again and it’s supposed to rain tomorrow. So there we are. That’s the weather. So my party chat this week is something completely different than what I’ve been talking about the whole podcast. And it’s using AI as a consultant for something right in front of you, and that’s rearranging your physical desktop. And this is something that I think AI has gotten a lot better at over the last couple of years. And so you just take a photo of your physical desk, your monitor, cables, notebooks, whatever’s on your desk, papers strewn all over. You upload it into an AI as a photo, and then you just do a prompt like act as an ergonomic and productivity expert, analyze my physical workspace layout, identify friction points or clutter bottlenecks and suggest three concrete layout changes to optimize comfort and workflow.
And it gives you an objective, fresh perspective on your physical working space that you haven’t noticed just because you’re sitting there every day. And I did this and I kind of changed things from one side to the other on my desk at the advice of it, and it just made a world of difference. So totally cool practical use of AI in an advanced way that you might not have even thought of.

Tom Mighell:
I did a version of this a couple of months ago before we moved out of our house where I have a terrible cord management problem. I just have so many cords and it’s the jungle down there. It’s so terrible. I’m just so bad at it. So I took a picture of the back of my desk, the backside, and I said, “Look at all the cords. Here they are. Please recommend a cable and cord management system for me to put into my office.” And I will tell you that the system that it recommended to me was so expensive and so time-consuming to put in that I was immediately discouraged and have never done it. Now I was getting ready to move out, so I wasn’t going to do it at that point. I probably need to, now that I’m settled in another place for the time being, probably need to do that again.
But sometimes you get what you ask for and it’s a lot more than you were planning on. And so that wraps it up for this edition of the Kennedy Mighell Report.

Dennis Kennedy:
Thanks to the Legal Talk Network team for producing the show. You can find show notes and transcripts on the Legal Talk Network website.

Tom Mighell:
If you like what you hear, please subscribe in your favorite podcast app and leave us a review.

Dennis Kennedy:
You can also connect with us on LinkedIn with your questions and whatever you want to tell us about your audience profile and listening habits because that will help us out.

Tom Mighell:
So until the next podcast, I’m Tom Mighell.

Dennis Kennedy:
And I’m Dennis Kennedy and you’ve been listening to the Kennedy Mighell Report, reporting from the leading edge of technology and legal work for more than 20 years.

Announcer:
Thanks for listening to the Kennedy Mighell Report. Check out Dennis and Tom’s book, The Lawyer’s Guide to Collaboration Tools and Technologies: Smart Ways to Work Together from ABA Books or Amazon, and join us every other week for another edition of the Kennedy Mighell Report only on the Legal Talk Network.