WEBVTT - Are you a GenAI Gunslinger? Why your approach is costing you results.    

0:00:00.560 --> 0:00:06.000
<v Speaker 1>Today, I am back with Neo Applan, Inventium's Genai expert

0:00:06.400 --> 0:00:11.320
<v Speaker 1>to help unpack a very cool concept that he recently

0:00:11.360 --> 0:00:16.400
<v Speaker 1>shared with me. It's about gunslinging versus architecting, and it's

0:00:16.400 --> 0:00:20.079
<v Speaker 1>all about how you work with AI. So if you

0:00:20.239 --> 0:00:24.800
<v Speaker 1>have ever found yourself wrestling with chat, GPT or one

0:00:24.800 --> 0:00:27.080
<v Speaker 1>of the other jenai is to get it just right,

0:00:27.600 --> 0:00:36.200
<v Speaker 1>then this episode is for you. Welcome to How I Work,

0:00:36.400 --> 0:00:40.560
<v Speaker 1>a show about habits, rituals, and strategies for optimizing your day.

0:00:41.159 --> 0:00:45.400
<v Speaker 1>I'm your host, doctor Amantha Imber. Two years ago, I

0:00:45.520 --> 0:00:49.960
<v Speaker 1>completely overhauled how I work with Genai and I'm now

0:00:50.040 --> 0:00:54.680
<v Speaker 1>saving over forty hours every single week. That is no exaggeration,

0:00:55.480 --> 0:00:59.120
<v Speaker 1>and that's exactly why my company, Inventium created the Genai

0:00:59.280 --> 0:01:03.120
<v Speaker 1>Productivity Upgrade. It's a twelve week course designed to move

0:01:03.200 --> 0:01:07.720
<v Speaker 1>you from AI doubler to productivity machine. No fluff, just

0:01:07.840 --> 0:01:11.040
<v Speaker 1>practical strategies that will pay off from week one, saving

0:01:11.080 --> 0:01:14.559
<v Speaker 1>you at least ten hours every single week. You'll learn

0:01:14.640 --> 0:01:17.760
<v Speaker 1>how to make AI sound exactly like you use AI

0:01:17.920 --> 0:01:21.319
<v Speaker 1>as your second brain to excel at your job, and

0:01:21.640 --> 0:01:25.560
<v Speaker 1>so much more. Whether you're a complete beginner or already dabbling.

0:01:25.640 --> 0:01:28.760
<v Speaker 1>We've got you covered, starting with prompting fundamentals and going

0:01:28.800 --> 0:01:31.959
<v Speaker 1>all the way through to advanced automations and agentic AI.

0:01:32.520 --> 0:01:35.720
<v Speaker 1>We kick off on July fourteen, and spots are limited.

0:01:36.160 --> 0:01:40.080
<v Speaker 1>Visit inventium dot com, dot au, forward slash Jenai hyphen

0:01:40.200 --> 0:01:43.360
<v Speaker 1>cohort to secure your place now, there's a link to

0:01:43.400 --> 0:01:45.760
<v Speaker 1>that in the show notes, and you've got nothing to

0:01:45.800 --> 0:01:49.080
<v Speaker 1>lose because there's a seven day money back guarantee, So

0:01:49.320 --> 0:01:51.640
<v Speaker 1>head to the link in the show notes and check

0:01:51.640 --> 0:01:56.320
<v Speaker 1>out the program today. So, Neo, you say that most

0:01:56.360 --> 0:02:00.760
<v Speaker 1>people are gunslingers when they prompt Jenai, Can you tell

0:02:00.800 --> 0:02:02.000
<v Speaker 1>me what you mean by that.

0:02:03.280 --> 0:02:06.200
<v Speaker 2>Yeah, it's a metaphor. It's not a technical term. It's

0:02:06.320 --> 0:02:09.280
<v Speaker 2>how some people work with jen Ai. It's effectively, get

0:02:09.320 --> 0:02:12.000
<v Speaker 2>a decent enough prompt, shoot from the hip and hope

0:02:12.040 --> 0:02:14.760
<v Speaker 2>and pray that it's going to go pretty well. Just

0:02:14.800 --> 0:02:17.280
<v Speaker 2>as gun slinging isn't the most accurate way that you

0:02:17.360 --> 0:02:21.800
<v Speaker 2>can hit a target, same thing with prompting gun slinging,

0:02:21.840 --> 0:02:24.440
<v Speaker 2>where you've got a pretty average prompt, you've shot it

0:02:24.480 --> 0:02:27.120
<v Speaker 2>out there, chances are it won't hit the bull's eye

0:02:27.320 --> 0:02:29.840
<v Speaker 2>every time. And so what you're finding is you then

0:02:29.919 --> 0:02:32.280
<v Speaker 2>have to have a discussion with the AI to improve

0:02:32.320 --> 0:02:33.800
<v Speaker 2>the output. I mean a little bit more of this,

0:02:33.919 --> 0:02:36.000
<v Speaker 2>a little less so that it can go from a

0:02:36.040 --> 0:02:39.200
<v Speaker 2>discussion to feeling very much like a debate and argument

0:02:39.600 --> 0:02:42.320
<v Speaker 2>with AI to get what you're really after. So the

0:02:42.360 --> 0:02:45.239
<v Speaker 2>first prompt might be fast, but then the cleaning up

0:02:45.400 --> 0:02:48.880
<v Speaker 2>is a very long task and can get frustrating after time,

0:02:48.960 --> 0:02:51.280
<v Speaker 2>lots of back and forth refining, so it can be

0:02:51.360 --> 0:02:54.560
<v Speaker 2>time consuming and inefficient if you're doing anything beyond kind

0:02:54.560 --> 0:02:56.560
<v Speaker 2>of basic prompting or basic tasks.

0:02:57.000 --> 0:03:00.760
<v Speaker 1>Okay, so then there is what you call architect What

0:03:00.800 --> 0:03:02.840
<v Speaker 1>does that look like instead.

0:03:03.080 --> 0:03:05.720
<v Speaker 2>Well, instead of shooting from the hip, it's much more deliberate,

0:03:05.800 --> 0:03:09.320
<v Speaker 2>it's much more collaborative with the AI, and it's more

0:03:09.440 --> 0:03:12.760
<v Speaker 2>likely to get you a great result and certainly faster

0:03:13.400 --> 0:03:16.880
<v Speaker 2>than gunslinging. And that there's a lot less refinement and

0:03:16.919 --> 0:03:19.560
<v Speaker 2>a lot less arguing with AI. So what you do is,

0:03:19.760 --> 0:03:22.160
<v Speaker 2>first off, you need to agree with AI and what

0:03:22.200 --> 0:03:24.520
<v Speaker 2>your goals are. Now, notice I say to agree, So

0:03:24.560 --> 0:03:26.800
<v Speaker 2>you need to say, here's what my goals are, and

0:03:26.840 --> 0:03:29.760
<v Speaker 2>you need to make sure that AI understands those goals.

0:03:29.760 --> 0:03:32.880
<v Speaker 2>So here's what I'm trying to achieve. Then you also

0:03:32.919 --> 0:03:36.520
<v Speaker 2>work with it on things like best practice or structure

0:03:36.960 --> 0:03:40.080
<v Speaker 2>or all those kind of things before it starts writing

0:03:40.120 --> 0:03:42.920
<v Speaker 2>anything else. So it's effectively you're working with it as

0:03:42.960 --> 0:03:46.320
<v Speaker 2>a planning partner and getting it to challenge your ideas,

0:03:46.400 --> 0:03:49.120
<v Speaker 2>getting it to challenge its own ideas, so it knows

0:03:49.120 --> 0:03:51.880
<v Speaker 2>the structure, what you're trying to achieve, what best practice is,

0:03:52.280 --> 0:03:54.960
<v Speaker 2>before you get it to write the thing. So it's

0:03:54.960 --> 0:03:57.160
<v Speaker 2>almost like designing a house before you build it. You

0:03:57.160 --> 0:03:59.000
<v Speaker 2>don't want to just build a house first. You want

0:03:59.040 --> 0:04:01.920
<v Speaker 2>to design it first, and then when it's already designed,

0:04:02.120 --> 0:04:03.520
<v Speaker 2>then you go and get the builders to come on

0:04:03.600 --> 0:04:04.400
<v Speaker 2>and to build the thing.

0:04:05.600 --> 0:04:10.000
<v Speaker 1>Okay, so I'm obviously not going to use architecting when

0:04:10.040 --> 0:04:13.480
<v Speaker 1>I'm writing a simple email. So I'm wondering what are

0:04:13.520 --> 0:04:17.719
<v Speaker 1>the kinds of tasks that architecting is best suited for.

0:04:18.680 --> 0:04:21.400
<v Speaker 2>Anything it's complex or that you need to get a

0:04:21.440 --> 0:04:25.320
<v Speaker 2>great result at where good enough isn't isn't enough, isn't

0:04:25.360 --> 0:04:27.400
<v Speaker 2>good enough? You need to get something it's excellent. So

0:04:27.920 --> 0:04:30.560
<v Speaker 2>things like if you're working on a strategy paper or

0:04:30.640 --> 0:04:34.120
<v Speaker 2>training sessions or workshop design or any kind of report

0:04:34.200 --> 0:04:37.520
<v Speaker 2>or analysis and all those kind of things, something where

0:04:37.800 --> 0:04:40.320
<v Speaker 2>you need to make sure that it's grounding it in

0:04:40.640 --> 0:04:44.240
<v Speaker 2>background and science and industry best practices and things like that,

0:04:44.480 --> 0:04:46.839
<v Speaker 2>where you want to get a great result, This is

0:04:46.880 --> 0:04:49.520
<v Speaker 2>where I would use architecting. So if you need a

0:04:49.560 --> 0:04:53.000
<v Speaker 2>structured output or some thinking through it, it's really good

0:04:53.080 --> 0:04:55.479
<v Speaker 2>for that. So if you wanted to have something which

0:04:55.560 --> 0:04:58.400
<v Speaker 2>is average or it's quick, fine, gunsling, like if I'm

0:04:58.400 --> 0:05:01.160
<v Speaker 2>doing emails off gunsling thing is fine. But if I

0:05:01.200 --> 0:05:03.080
<v Speaker 2>want to get a document or a structure, or I

0:05:03.160 --> 0:05:05.880
<v Speaker 2>want to, as I said, a team workshop. If I

0:05:05.920 --> 0:05:08.599
<v Speaker 2>want a great team workshop, then why don't I get

0:05:08.680 --> 0:05:11.800
<v Speaker 2>architecting to help me to build something even better. It

0:05:11.800 --> 0:05:14.240
<v Speaker 2>works great with all the different ais, I should say,

0:05:14.760 --> 0:05:17.919
<v Speaker 2>but in some of them it works better. So copilot

0:05:18.080 --> 0:05:21.120
<v Speaker 2>architecting is brilliant. It does such a much better job

0:05:21.400 --> 0:05:23.760
<v Speaker 2>than just average copilot gun slinging.

0:05:24.760 --> 0:05:27.040
<v Speaker 1>And if I'm using chat ept, as I know a

0:05:27.080 --> 0:05:29.279
<v Speaker 1>lot of listeners do, should I be choosing one of

0:05:29.320 --> 0:05:32.799
<v Speaker 1>their logic models for this or any models? Fine?

0:05:33.160 --> 0:05:35.200
<v Speaker 2>Yes, the logic models, and these are the ones that

0:05:35.279 --> 0:05:39.120
<v Speaker 2>start in O so three to four, etc. The logic

0:05:39.160 --> 0:05:41.800
<v Speaker 2>models are excellent for these kind of things. But they

0:05:41.839 --> 0:05:46.000
<v Speaker 2>do a better job when you do architecting, because what

0:05:46.040 --> 0:05:48.479
<v Speaker 2>you're still doing is you're working with the GENAI on

0:05:48.560 --> 0:05:51.200
<v Speaker 2>what the goals are and things like best practice rather

0:05:51.200 --> 0:05:54.560
<v Speaker 2>than letting it think. You're working with the GENAI to

0:05:54.600 --> 0:05:57.680
<v Speaker 2>be able to architect this together, and those logic models

0:05:57.680 --> 0:06:00.720
<v Speaker 2>do give you a better result. So yeah, do use

0:06:00.760 --> 0:06:03.719
<v Speaker 2>the architecting, and those models you over next level the result.

0:06:04.560 --> 0:06:07.239
<v Speaker 1>So NIO, I would love you to walk me through

0:06:07.800 --> 0:06:10.720
<v Speaker 1>an example, like, let's just say I don't know you're

0:06:10.760 --> 0:06:16.440
<v Speaker 1>designing a team building session for a client off site

0:06:16.560 --> 0:06:19.080
<v Speaker 1>or maybe your own team's off site. How would you

0:06:19.200 --> 0:06:22.320
<v Speaker 1>use architecting to go through that task?

0:06:23.480 --> 0:06:25.800
<v Speaker 2>I would first start with the context, So here, I'm

0:06:25.839 --> 0:06:28.360
<v Speaker 2>planning on building a team off site, and I'd like

0:06:28.400 --> 0:06:31.120
<v Speaker 2>to build a structure with you. Tell it what your

0:06:31.160 --> 0:06:35.280
<v Speaker 2>goals are. After that, I'd get it to ask itself

0:06:35.520 --> 0:06:39.120
<v Speaker 2>what best practice is? Are there ways that I can

0:06:39.160 --> 0:06:42.839
<v Speaker 2>achieve this goal even better? Get it to even review

0:06:42.880 --> 0:06:44.800
<v Speaker 2>its own work to see whether it can improve it.

0:06:45.440 --> 0:06:48.240
<v Speaker 2>What's the science behind team performance? All those kind of

0:06:48.240 --> 0:06:51.360
<v Speaker 2>things to get it to think through all the best

0:06:51.360 --> 0:06:54.279
<v Speaker 2>practice paths on how to build a great team workshop,

0:06:54.600 --> 0:06:56.440
<v Speaker 2>I might work with it on say, structure, how would

0:06:56.440 --> 0:06:58.760
<v Speaker 2>I structure these kind of things? What kind of activities

0:06:58.800 --> 0:07:02.120
<v Speaker 2>would work. I'm not getting it to build the workshop

0:07:02.640 --> 0:07:05.800
<v Speaker 2>or even the final output yet. I'm getting it to

0:07:05.880 --> 0:07:08.880
<v Speaker 2>work through best practice, how it aligns with my goals?

0:07:08.920 --> 0:07:11.720
<v Speaker 2>Are there better ways to do it? All those kind

0:07:11.720 --> 0:07:14.840
<v Speaker 2>of things. Once I'm comfortable it's nailed what I'm trying

0:07:14.840 --> 0:07:18.360
<v Speaker 2>to achieve, it understands best practice, it's refined as much

0:07:18.360 --> 0:07:21.120
<v Speaker 2>as I think it needs to, then I'll go, Okay,

0:07:21.240 --> 0:07:24.040
<v Speaker 2>let's build these things together. Then it will go through

0:07:24.120 --> 0:07:26.200
<v Speaker 2>and actually build me a workshop, and it will say

0:07:26.440 --> 0:07:29.440
<v Speaker 2>how and Wyatt's built the workshop for that best practice

0:07:29.720 --> 0:07:31.760
<v Speaker 2>and how it will deliver to what I'm trying to

0:07:31.800 --> 0:07:33.200
<v Speaker 2>achieve with those goals.

0:07:33.640 --> 0:07:35.800
<v Speaker 1>And then I imagine you could break it down even

0:07:35.920 --> 0:07:38.920
<v Speaker 1>further to go, okay, can we work on the morning

0:07:38.960 --> 0:07:41.520
<v Speaker 1>segment first? And then let's work on the segment that

0:07:41.960 --> 0:07:44.240
<v Speaker 1>is just before lunch and so on and so forth,

0:07:44.240 --> 0:07:46.920
<v Speaker 1>as opposed to okay, build out the full agenda is

0:07:47.560 --> 0:07:48.760
<v Speaker 1>Have I got my thinking right on that?

0:07:48.800 --> 0:07:50.760
<v Speaker 2>Absolutely? Yeah? And this is really good for you. You've

0:07:50.800 --> 0:07:53.680
<v Speaker 2>got a big document, say I'd be working on it

0:07:53.720 --> 0:07:56.760
<v Speaker 2>with the goals the strategies what you need to do

0:07:56.800 --> 0:07:59.560
<v Speaker 2>in the document, and then once you're comfortable with we'll

0:07:59.600 --> 0:08:02.360
<v Speaker 2>call it like index or a table of contents, then

0:08:02.440 --> 0:08:05.640
<v Speaker 2>you'd build out chapter by chapter, section by section, and

0:08:05.680 --> 0:08:07.920
<v Speaker 2>that's much better to do it that way. That way

0:08:08.040 --> 0:08:10.040
<v Speaker 2>it will give you a better result and it won't

0:08:10.040 --> 0:08:12.520
<v Speaker 2>go off track at all. So yes, structure first and

0:08:12.560 --> 0:08:15.520
<v Speaker 2>then section by section is the best way to architect

0:08:15.880 --> 0:08:17.240
<v Speaker 2>and best way to get the results.

0:08:17.960 --> 0:08:20.680
<v Speaker 1>Neo, thank you so much for coming on again. I

0:08:20.760 --> 0:08:24.600
<v Speaker 1>know that we have recently in our own GENAI workshops

0:08:24.600 --> 0:08:29.200
<v Speaker 1>been talking about these terms architecting and gunslinging, and I

0:08:29.240 --> 0:08:32.400
<v Speaker 1>feel like there is that aha moment when people start

0:08:32.440 --> 0:08:34.960
<v Speaker 1>to actually put this into action. So if this is

0:08:35.040 --> 0:08:40.360
<v Speaker 1>resonated with you, I strongly recommend give architecting a go

0:08:40.600 --> 0:08:43.440
<v Speaker 1>today when you are working with the AI to work

0:08:43.480 --> 0:08:47.199
<v Speaker 1>on a complex task. If you like today's Joe, make

0:08:47.240 --> 0:08:50.000
<v Speaker 1>sure you hit follow on your podcast app to be

0:08:50.040 --> 0:08:53.920
<v Speaker 1>alerted when new episodes drop. How I Work was recorded

0:08:53.960 --> 0:08:56.599
<v Speaker 1>on the traditional land of the Warrangery people, part of

0:08:56.640 --> 0:08:59.280
<v Speaker 1>the Cooler Nation. A big thank you to Martin Nimber

0:08:59.440 --> 0:09:00.600
<v Speaker 1>for doing this sound mix